Commit | Line | Data |
---|---|---|
a0d0e21e LW |
1 | /* regexec.c |
2 | */ | |
3 | ||
4 | /* | |
5 | * "One Ring to rule them all, One Ring to find them..." | |
6 | */ | |
7 | ||
61296642 DM |
8 | /* This file contains functions for executing a regular expression. See |
9 | * also regcomp.c which funnily enough, contains functions for compiling | |
166f8a29 | 10 | * a regular expression. |
e4a054ea DM |
11 | * |
12 | * This file is also copied at build time to ext/re/re_exec.c, where | |
13 | * it's built with -DPERL_EXT_RE_BUILD -DPERL_EXT_RE_DEBUG -DPERL_EXT. | |
14 | * This causes the main functions to be compiled under new names and with | |
15 | * debugging support added, which makes "use re 'debug'" work. | |
166f8a29 DM |
16 | */ |
17 | ||
a687059c LW |
18 | /* NOTE: this is derived from Henry Spencer's regexp code, and should not |
19 | * confused with the original package (see point 3 below). Thanks, Henry! | |
20 | */ | |
21 | ||
22 | /* Additional note: this code is very heavily munged from Henry's version | |
23 | * in places. In some spots I've traded clarity for efficiency, so don't | |
24 | * blame Henry for some of the lack of readability. | |
25 | */ | |
26 | ||
e50aee73 AD |
27 | /* The names of the functions have been changed from regcomp and |
28 | * regexec to pregcomp and pregexec in order to avoid conflicts | |
29 | * with the POSIX routines of the same names. | |
30 | */ | |
31 | ||
b9d5759e | 32 | #ifdef PERL_EXT_RE_BUILD |
54df2634 | 33 | #include "re_top.h" |
9041c2e3 | 34 | #endif |
56953603 | 35 | |
a687059c | 36 | /* |
e50aee73 | 37 | * pregcomp and pregexec -- regsub and regerror are not used in perl |
a687059c LW |
38 | * |
39 | * Copyright (c) 1986 by University of Toronto. | |
40 | * Written by Henry Spencer. Not derived from licensed software. | |
41 | * | |
42 | * Permission is granted to anyone to use this software for any | |
43 | * purpose on any computer system, and to redistribute it freely, | |
44 | * subject to the following restrictions: | |
45 | * | |
46 | * 1. The author is not responsible for the consequences of use of | |
47 | * this software, no matter how awful, even if they arise | |
48 | * from defects in it. | |
49 | * | |
50 | * 2. The origin of this software must not be misrepresented, either | |
51 | * by explicit claim or by omission. | |
52 | * | |
53 | * 3. Altered versions must be plainly marked as such, and must not | |
54 | * be misrepresented as being the original software. | |
55 | * | |
56 | **** Alterations to Henry's code are... | |
57 | **** | |
4bb101f2 | 58 | **** Copyright (C) 1991, 1992, 1993, 1994, 1995, 1996, 1997, 1998, 1999, |
b94e2f88 | 59 | **** 2000, 2001, 2002, 2003, 2004, 2005, 2006, by Larry Wall and others |
a687059c | 60 | **** |
9ef589d8 LW |
61 | **** You may distribute under the terms of either the GNU General Public |
62 | **** License or the Artistic License, as specified in the README file. | |
a687059c LW |
63 | * |
64 | * Beware that some of this code is subtly aware of the way operator | |
65 | * precedence is structured in regular expressions. Serious changes in | |
66 | * regular-expression syntax might require a total rethink. | |
67 | */ | |
68 | #include "EXTERN.h" | |
864dbfa3 | 69 | #define PERL_IN_REGEXEC_C |
a687059c | 70 | #include "perl.h" |
0f5d15d6 | 71 | |
54df2634 NC |
72 | #ifdef PERL_IN_XSUB_RE |
73 | # include "re_comp.h" | |
74 | #else | |
75 | # include "regcomp.h" | |
76 | #endif | |
a687059c | 77 | |
c277df42 IZ |
78 | #define RF_tainted 1 /* tainted information used? */ |
79 | #define RF_warned 2 /* warned about big count? */ | |
ce862d02 | 80 | #define RF_evaled 4 /* Did an EVAL with setting? */ |
a0ed51b3 LW |
81 | #define RF_utf8 8 /* String contains multibyte chars? */ |
82 | ||
eb160463 | 83 | #define UTF ((PL_reg_flags & RF_utf8) != 0) |
ce862d02 IZ |
84 | |
85 | #define RS_init 1 /* eval environment created */ | |
86 | #define RS_set 2 /* replsv value is set */ | |
c277df42 | 87 | |
a687059c LW |
88 | #ifndef STATIC |
89 | #define STATIC static | |
90 | #endif | |
91 | ||
32fc9b6a | 92 | #define REGINCLASS(prog,p,c) (ANYOF_FLAGS(p) ? reginclass(prog,p,c,0,0) : ANYOF_BITMAP_TEST(p,*(c))) |
7d3e948e | 93 | |
c277df42 IZ |
94 | /* |
95 | * Forwards. | |
96 | */ | |
97 | ||
33b8afdf | 98 | #define CHR_SVLEN(sv) (do_utf8 ? sv_len_utf8(sv) : SvCUR(sv)) |
53c4c00c | 99 | #define CHR_DIST(a,b) (PL_reg_match_utf8 ? utf8_distance(a,b) : a - b) |
a0ed51b3 | 100 | |
3dab1dad YO |
101 | #define HOPc(pos,off) \ |
102 | (char *)(PL_reg_match_utf8 \ | |
52657f30 | 103 | ? reghop3((U8*)pos, off, (U8*)(off >= 0 ? PL_regeol : PL_bostr)) \ |
3dab1dad YO |
104 | : (U8*)(pos + off)) |
105 | #define HOPBACKc(pos, off) \ | |
07be1b83 YO |
106 | (char*)(PL_reg_match_utf8\ |
107 | ? reghopmaybe3((U8*)pos, -off, (U8*)PL_bostr) \ | |
108 | : (pos - off >= PL_bostr) \ | |
8e11feef | 109 | ? (U8*)pos - off \ |
3dab1dad | 110 | : NULL) |
efb30f32 | 111 | |
53c4c00c | 112 | #define HOP3(pos,off,lim) (PL_reg_match_utf8 ? reghop3((U8*)pos, off, (U8*)lim) : (U8*)(pos + off)) |
1aa99e6b | 113 | #define HOP3c(pos,off,lim) ((char*)HOP3(pos,off,lim)) |
1aa99e6b | 114 | |
1a4fad37 AL |
115 | #define LOAD_UTF8_CHARCLASS(class,str) STMT_START { \ |
116 | if (!CAT2(PL_utf8_,class)) { bool ok; ENTER; save_re_context(); ok=CAT2(is_utf8_,class)((const U8*)str); assert(ok); LEAVE; } } STMT_END | |
117 | #define LOAD_UTF8_CHARCLASS_ALNUM() LOAD_UTF8_CHARCLASS(alnum,"a") | |
118 | #define LOAD_UTF8_CHARCLASS_DIGIT() LOAD_UTF8_CHARCLASS(digit,"0") | |
119 | #define LOAD_UTF8_CHARCLASS_SPACE() LOAD_UTF8_CHARCLASS(space," ") | |
120 | #define LOAD_UTF8_CHARCLASS_MARK() LOAD_UTF8_CHARCLASS(mark, "\xcd\x86") | |
51371543 | 121 | |
3dab1dad YO |
122 | /* TODO: Combine JUMPABLE and HAS_TEXT to cache OP(rn) */ |
123 | ||
5f80c4cf | 124 | /* for use after a quantifier and before an EXACT-like node -- japhy */ |
e2d8ce26 JP |
125 | #define JUMPABLE(rn) ( \ |
126 | OP(rn) == OPEN || OP(rn) == CLOSE || OP(rn) == EVAL || \ | |
cca55fe3 JP |
127 | OP(rn) == SUSPEND || OP(rn) == IFMATCH || \ |
128 | OP(rn) == PLUS || OP(rn) == MINMOD || \ | |
3dab1dad | 129 | (PL_regkind[OP(rn)] == CURLY && ARG1(rn) > 0) \ |
e2d8ce26 JP |
130 | ) |
131 | ||
cca55fe3 | 132 | #define HAS_TEXT(rn) ( \ |
3dab1dad | 133 | PL_regkind[OP(rn)] == EXACT || PL_regkind[OP(rn)] == REF \ |
cca55fe3 | 134 | ) |
e2d8ce26 | 135 | |
a84d97b6 HS |
136 | /* |
137 | Search for mandatory following text node; for lookahead, the text must | |
138 | follow but for lookbehind (rn->flags != 0) we skip to the next step. | |
139 | */ | |
cca55fe3 | 140 | #define FIND_NEXT_IMPT(rn) STMT_START { \ |
3dab1dad YO |
141 | while (JUMPABLE(rn)) { \ |
142 | const OPCODE type = OP(rn); \ | |
143 | if (type == SUSPEND || PL_regkind[type] == CURLY) \ | |
e2d8ce26 | 144 | rn = NEXTOPER(NEXTOPER(rn)); \ |
3dab1dad | 145 | else if (type == PLUS) \ |
cca55fe3 | 146 | rn = NEXTOPER(rn); \ |
3dab1dad | 147 | else if (type == IFMATCH) \ |
a84d97b6 | 148 | rn = (rn->flags == 0) ? NEXTOPER(NEXTOPER(rn)) : rn + ARG(rn); \ |
e2d8ce26 | 149 | else rn += NEXT_OFF(rn); \ |
3dab1dad | 150 | } \ |
5f80c4cf | 151 | } STMT_END |
74750237 | 152 | |
acfe0abc | 153 | static void restore_pos(pTHX_ void *arg); |
51371543 | 154 | |
76e3520e | 155 | STATIC CHECKPOINT |
cea2e8a9 | 156 | S_regcppush(pTHX_ I32 parenfloor) |
a0d0e21e | 157 | { |
97aff369 | 158 | dVAR; |
a3b680e6 | 159 | const int retval = PL_savestack_ix; |
b1ce53c5 | 160 | #define REGCP_PAREN_ELEMS 4 |
a3b680e6 | 161 | const int paren_elems_to_push = (PL_regsize - parenfloor) * REGCP_PAREN_ELEMS; |
a0d0e21e | 162 | int p; |
40a82448 | 163 | GET_RE_DEBUG_FLAGS_DECL; |
a0d0e21e | 164 | |
e49a9654 IH |
165 | if (paren_elems_to_push < 0) |
166 | Perl_croak(aTHX_ "panic: paren_elems_to_push < 0"); | |
167 | ||
a01268b5 | 168 | #define REGCP_OTHER_ELEMS 6 |
4b3c1a47 | 169 | SSGROW(paren_elems_to_push + REGCP_OTHER_ELEMS); |
3280af22 | 170 | for (p = PL_regsize; p > parenfloor; p--) { |
b1ce53c5 | 171 | /* REGCP_PARENS_ELEMS are pushed per pairs of parentheses. */ |
cf93c79d IZ |
172 | SSPUSHINT(PL_regendp[p]); |
173 | SSPUSHINT(PL_regstartp[p]); | |
3280af22 | 174 | SSPUSHPTR(PL_reg_start_tmp[p]); |
a0d0e21e | 175 | SSPUSHINT(p); |
40a82448 DM |
176 | DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, |
177 | " saving \\%"UVuf" %"IVdf"(%"IVdf")..%"IVdf"\n", | |
178 | (UV)p, (IV)PL_regstartp[p], | |
179 | (IV)(PL_reg_start_tmp[p] - PL_bostr), | |
180 | (IV)PL_regendp[p] | |
181 | )); | |
a0d0e21e | 182 | } |
b1ce53c5 | 183 | /* REGCP_OTHER_ELEMS are pushed in any case, parentheses or no. */ |
3280af22 NIS |
184 | SSPUSHINT(PL_regsize); |
185 | SSPUSHINT(*PL_reglastparen); | |
a01268b5 | 186 | SSPUSHINT(*PL_reglastcloseparen); |
3280af22 | 187 | SSPUSHPTR(PL_reginput); |
41123dfd JH |
188 | #define REGCP_FRAME_ELEMS 2 |
189 | /* REGCP_FRAME_ELEMS are part of the REGCP_OTHER_ELEMS and | |
190 | * are needed for the regexp context stack bookkeeping. */ | |
191 | SSPUSHINT(paren_elems_to_push + REGCP_OTHER_ELEMS - REGCP_FRAME_ELEMS); | |
b1ce53c5 | 192 | SSPUSHINT(SAVEt_REGCONTEXT); /* Magic cookie. */ |
41123dfd | 193 | |
a0d0e21e LW |
194 | return retval; |
195 | } | |
196 | ||
c277df42 | 197 | /* These are needed since we do not localize EVAL nodes: */ |
a3621e74 | 198 | # define REGCP_SET(cp) DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, \ |
faccc32b | 199 | " Setting an EVAL scope, savestack=%"IVdf"\n", \ |
02db2b7b | 200 | (IV)PL_savestack_ix)); cp = PL_savestack_ix |
c3464db5 | 201 | |
a3621e74 | 202 | # define REGCP_UNWIND(cp) DEBUG_EXECUTE_r(cp != PL_savestack_ix ? \ |
c3464db5 | 203 | PerlIO_printf(Perl_debug_log, \ |
faccc32b | 204 | " Clearing an EVAL scope, savestack=%"IVdf"..%"IVdf"\n", \ |
02db2b7b | 205 | (IV)(cp), (IV)PL_savestack_ix) : 0); regcpblow(cp) |
c277df42 | 206 | |
76e3520e | 207 | STATIC char * |
097eb12c | 208 | S_regcppop(pTHX_ const regexp *rex) |
a0d0e21e | 209 | { |
97aff369 | 210 | dVAR; |
b1ce53c5 | 211 | I32 i; |
a0d0e21e | 212 | char *input; |
b1ce53c5 | 213 | |
a3621e74 YO |
214 | GET_RE_DEBUG_FLAGS_DECL; |
215 | ||
b1ce53c5 | 216 | /* Pop REGCP_OTHER_ELEMS before the parentheses loop starts. */ |
a0d0e21e | 217 | i = SSPOPINT; |
b1ce53c5 JH |
218 | assert(i == SAVEt_REGCONTEXT); /* Check that the magic cookie is there. */ |
219 | i = SSPOPINT; /* Parentheses elements to pop. */ | |
a0d0e21e | 220 | input = (char *) SSPOPPTR; |
a01268b5 | 221 | *PL_reglastcloseparen = SSPOPINT; |
3280af22 NIS |
222 | *PL_reglastparen = SSPOPINT; |
223 | PL_regsize = SSPOPINT; | |
b1ce53c5 JH |
224 | |
225 | /* Now restore the parentheses context. */ | |
41123dfd JH |
226 | for (i -= (REGCP_OTHER_ELEMS - REGCP_FRAME_ELEMS); |
227 | i > 0; i -= REGCP_PAREN_ELEMS) { | |
1df70142 | 228 | I32 tmps; |
097eb12c | 229 | U32 paren = (U32)SSPOPINT; |
3280af22 | 230 | PL_reg_start_tmp[paren] = (char *) SSPOPPTR; |
cf93c79d IZ |
231 | PL_regstartp[paren] = SSPOPINT; |
232 | tmps = SSPOPINT; | |
3280af22 NIS |
233 | if (paren <= *PL_reglastparen) |
234 | PL_regendp[paren] = tmps; | |
a3621e74 | 235 | DEBUG_EXECUTE_r( |
c3464db5 | 236 | PerlIO_printf(Perl_debug_log, |
b900a521 | 237 | " restoring \\%"UVuf" to %"IVdf"(%"IVdf")..%"IVdf"%s\n", |
9041c2e3 | 238 | (UV)paren, (IV)PL_regstartp[paren], |
b900a521 | 239 | (IV)(PL_reg_start_tmp[paren] - PL_bostr), |
9041c2e3 | 240 | (IV)PL_regendp[paren], |
3280af22 | 241 | (paren > *PL_reglastparen ? "(no)" : "")); |
c277df42 | 242 | ); |
a0d0e21e | 243 | } |
a3621e74 | 244 | DEBUG_EXECUTE_r( |
bb7a0f54 | 245 | if (*PL_reglastparen + 1 <= rex->nparens) { |
c3464db5 | 246 | PerlIO_printf(Perl_debug_log, |
faccc32b | 247 | " restoring \\%"IVdf"..\\%"IVdf" to undef\n", |
4f639d21 | 248 | (IV)(*PL_reglastparen + 1), (IV)rex->nparens); |
c277df42 IZ |
249 | } |
250 | ); | |
daf18116 | 251 | #if 1 |
dafc8851 JH |
252 | /* It would seem that the similar code in regtry() |
253 | * already takes care of this, and in fact it is in | |
254 | * a better location to since this code can #if 0-ed out | |
255 | * but the code in regtry() is needed or otherwise tests | |
256 | * requiring null fields (pat.t#187 and split.t#{13,14} | |
daf18116 JH |
257 | * (as of patchlevel 7877) will fail. Then again, |
258 | * this code seems to be necessary or otherwise | |
259 | * building DynaLoader will fail: | |
260 | * "Error: '*' not in typemap in DynaLoader.xs, line 164" | |
261 | * --jhi */ | |
bb7a0f54 | 262 | for (i = *PL_reglastparen + 1; (U32)i <= rex->nparens; i++) { |
097eb12c AL |
263 | if (i > PL_regsize) |
264 | PL_regstartp[i] = -1; | |
265 | PL_regendp[i] = -1; | |
a0d0e21e | 266 | } |
dafc8851 | 267 | #endif |
a0d0e21e LW |
268 | return input; |
269 | } | |
270 | ||
02db2b7b | 271 | #define regcpblow(cp) LEAVE_SCOPE(cp) /* Ignores regcppush()ed data. */ |
a0d0e21e | 272 | |
a687059c | 273 | /* |
e50aee73 | 274 | * pregexec and friends |
a687059c LW |
275 | */ |
276 | ||
76234dfb | 277 | #ifndef PERL_IN_XSUB_RE |
a687059c | 278 | /* |
c277df42 | 279 | - pregexec - match a regexp against a string |
a687059c | 280 | */ |
c277df42 | 281 | I32 |
864dbfa3 | 282 | Perl_pregexec(pTHX_ register regexp *prog, char *stringarg, register char *strend, |
c3464db5 | 283 | char *strbeg, I32 minend, SV *screamer, U32 nosave) |
c277df42 IZ |
284 | /* strend: pointer to null at end of string */ |
285 | /* strbeg: real beginning of string */ | |
286 | /* minend: end of match must be >=minend after stringarg. */ | |
287 | /* nosave: For optimizations. */ | |
288 | { | |
289 | return | |
9041c2e3 | 290 | regexec_flags(prog, stringarg, strend, strbeg, minend, screamer, NULL, |
c277df42 IZ |
291 | nosave ? 0 : REXEC_COPY_STR); |
292 | } | |
76234dfb | 293 | #endif |
22e551b9 | 294 | |
9041c2e3 | 295 | /* |
cad2e5aa JH |
296 | * Need to implement the following flags for reg_anch: |
297 | * | |
298 | * USE_INTUIT_NOML - Useful to call re_intuit_start() first | |
299 | * USE_INTUIT_ML | |
300 | * INTUIT_AUTORITATIVE_NOML - Can trust a positive answer | |
301 | * INTUIT_AUTORITATIVE_ML | |
302 | * INTUIT_ONCE_NOML - Intuit can match in one location only. | |
303 | * INTUIT_ONCE_ML | |
304 | * | |
305 | * Another flag for this function: SECOND_TIME (so that float substrs | |
306 | * with giant delta may be not rechecked). | |
307 | */ | |
308 | ||
309 | /* Assumptions: if ANCH_GPOS, then strpos is anchored. XXXX Check GPOS logic */ | |
310 | ||
3f7c398e | 311 | /* If SCREAM, then SvPVX_const(sv) should be compatible with strpos and strend. |
cad2e5aa JH |
312 | Otherwise, only SvCUR(sv) is used to get strbeg. */ |
313 | ||
314 | /* XXXX We assume that strpos is strbeg unless sv. */ | |
315 | ||
6eb5f6b9 JH |
316 | /* XXXX Some places assume that there is a fixed substring. |
317 | An update may be needed if optimizer marks as "INTUITable" | |
318 | RExen without fixed substrings. Similarly, it is assumed that | |
319 | lengths of all the strings are no more than minlen, thus they | |
320 | cannot come from lookahead. | |
321 | (Or minlen should take into account lookahead.) */ | |
322 | ||
2c2d71f5 JH |
323 | /* A failure to find a constant substring means that there is no need to make |
324 | an expensive call to REx engine, thus we celebrate a failure. Similarly, | |
325 | finding a substring too deep into the string means that less calls to | |
30944b6d IZ |
326 | regtry() should be needed. |
327 | ||
328 | REx compiler's optimizer found 4 possible hints: | |
329 | a) Anchored substring; | |
330 | b) Fixed substring; | |
331 | c) Whether we are anchored (beginning-of-line or \G); | |
332 | d) First node (of those at offset 0) which may distingush positions; | |
6eb5f6b9 | 333 | We use a)b)d) and multiline-part of c), and try to find a position in the |
30944b6d IZ |
334 | string which does not contradict any of them. |
335 | */ | |
2c2d71f5 | 336 | |
6eb5f6b9 JH |
337 | /* Most of decisions we do here should have been done at compile time. |
338 | The nodes of the REx which we used for the search should have been | |
339 | deleted from the finite automaton. */ | |
340 | ||
cad2e5aa JH |
341 | char * |
342 | Perl_re_intuit_start(pTHX_ regexp *prog, SV *sv, char *strpos, | |
343 | char *strend, U32 flags, re_scream_pos_data *data) | |
344 | { | |
97aff369 | 345 | dVAR; |
b7953727 | 346 | register I32 start_shift = 0; |
cad2e5aa | 347 | /* Should be nonnegative! */ |
b7953727 | 348 | register I32 end_shift = 0; |
2c2d71f5 JH |
349 | register char *s; |
350 | register SV *check; | |
a1933d95 | 351 | char *strbeg; |
cad2e5aa | 352 | char *t; |
a3b680e6 | 353 | const int do_utf8 = sv ? SvUTF8(sv) : 0; /* if no sv we have to assume bytes */ |
cad2e5aa | 354 | I32 ml_anch; |
bd61b366 SS |
355 | register char *other_last = NULL; /* other substr checked before this */ |
356 | char *check_at = NULL; /* check substr found at this pos */ | |
1df70142 | 357 | const I32 multiline = prog->reganch & PMf_MULTILINE; |
30944b6d | 358 | #ifdef DEBUGGING |
890ce7af | 359 | const char * const i_strpos = strpos; |
30944b6d | 360 | #endif |
a3621e74 YO |
361 | |
362 | GET_RE_DEBUG_FLAGS_DECL; | |
363 | ||
a30b2f1f | 364 | RX_MATCH_UTF8_set(prog,do_utf8); |
cad2e5aa | 365 | |
b8d68ded | 366 | if (prog->reganch & ROPT_UTF8) { |
a3621e74 | 367 | DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, |
b8d68ded JH |
368 | "UTF-8 regex...\n")); |
369 | PL_reg_flags |= RF_utf8; | |
370 | } | |
371 | ||
a3621e74 | 372 | DEBUG_EXECUTE_r({ |
0df25f3d YO |
373 | RE_PV_DISPLAY_DECL(s, len, PL_reg_match_utf8, |
374 | PERL_DEBUG_PAD_ZERO(0), strpos, strend - strpos, 60); | |
375 | ||
2a782b5b JH |
376 | if (!PL_colorset) |
377 | reginitcolors(); | |
b8d68ded | 378 | if (PL_reg_match_utf8) |
a3621e74 | 379 | DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, |
b8d68ded | 380 | "UTF-8 target...\n")); |
2a782b5b | 381 | PerlIO_printf(Perl_debug_log, |
a0288114 | 382 | "%sGuessing start of match, REx%s \"%s%.60s%s%s\" against \"%s%.*s%s%s\"...\n", |
e4584336 | 383 | PL_colors[4], PL_colors[5], PL_colors[0], |
2a782b5b JH |
384 | prog->precomp, |
385 | PL_colors[1], | |
386 | (strlen(prog->precomp) > 60 ? "..." : ""), | |
387 | PL_colors[0], | |
388 | (int)(len > 60 ? 60 : len), | |
389 | s, PL_colors[1], | |
390 | (len > 60 ? "..." : "") | |
391 | ); | |
392 | }); | |
cad2e5aa | 393 | |
c344f387 JH |
394 | /* CHR_DIST() would be more correct here but it makes things slow. */ |
395 | if (prog->minlen > strend - strpos) { | |
a3621e74 | 396 | DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, |
a72c7584 | 397 | "String too short... [re_intuit_start]\n")); |
cad2e5aa | 398 | goto fail; |
2c2d71f5 | 399 | } |
a1933d95 | 400 | strbeg = (sv && SvPOK(sv)) ? strend - SvCUR(sv) : strpos; |
1aa99e6b | 401 | PL_regeol = strend; |
33b8afdf JH |
402 | if (do_utf8) { |
403 | if (!prog->check_utf8 && prog->check_substr) | |
404 | to_utf8_substr(prog); | |
405 | check = prog->check_utf8; | |
406 | } else { | |
407 | if (!prog->check_substr && prog->check_utf8) | |
408 | to_byte_substr(prog); | |
409 | check = prog->check_substr; | |
410 | } | |
411 | if (check == &PL_sv_undef) { | |
a3621e74 | 412 | DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, |
33b8afdf JH |
413 | "Non-utf string cannot match utf check string\n")); |
414 | goto fail; | |
415 | } | |
2c2d71f5 | 416 | if (prog->reganch & ROPT_ANCH) { /* Match at beg-of-str or after \n */ |
cad2e5aa JH |
417 | ml_anch = !( (prog->reganch & ROPT_ANCH_SINGLE) |
418 | || ( (prog->reganch & ROPT_ANCH_BOL) | |
7fba1cd6 | 419 | && !multiline ) ); /* Check after \n? */ |
cad2e5aa | 420 | |
7e25d62c JH |
421 | if (!ml_anch) { |
422 | if ( !(prog->reganch & (ROPT_ANCH_GPOS /* Checked by the caller */ | |
423 | | ROPT_IMPLICIT)) /* not a real BOL */ | |
3f7c398e | 424 | /* SvCUR is not set on references: SvRV and SvPVX_const overlap */ |
7e25d62c JH |
425 | && sv && !SvROK(sv) |
426 | && (strpos != strbeg)) { | |
a3621e74 | 427 | DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "Not at start...\n")); |
7e25d62c JH |
428 | goto fail; |
429 | } | |
430 | if (prog->check_offset_min == prog->check_offset_max && | |
cce850e4 | 431 | !(prog->reganch & ROPT_CANY_SEEN)) { |
2c2d71f5 | 432 | /* Substring at constant offset from beg-of-str... */ |
cad2e5aa JH |
433 | I32 slen; |
434 | ||
1aa99e6b | 435 | s = HOP3c(strpos, prog->check_offset_min, strend); |
653099ff GS |
436 | if (SvTAIL(check)) { |
437 | slen = SvCUR(check); /* >= 1 */ | |
cad2e5aa | 438 | |
9041c2e3 | 439 | if ( strend - s > slen || strend - s < slen - 1 |
2c2d71f5 | 440 | || (strend - s == slen && strend[-1] != '\n')) { |
a3621e74 | 441 | DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "String too long...\n")); |
2c2d71f5 | 442 | goto fail_finish; |
cad2e5aa JH |
443 | } |
444 | /* Now should match s[0..slen-2] */ | |
445 | slen--; | |
3f7c398e | 446 | if (slen && (*SvPVX_const(check) != *s |
cad2e5aa | 447 | || (slen > 1 |
3f7c398e | 448 | && memNE(SvPVX_const(check), s, slen)))) { |
2c2d71f5 | 449 | report_neq: |
a3621e74 | 450 | DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "String not equal...\n")); |
2c2d71f5 JH |
451 | goto fail_finish; |
452 | } | |
cad2e5aa | 453 | } |
3f7c398e | 454 | else if (*SvPVX_const(check) != *s |
653099ff | 455 | || ((slen = SvCUR(check)) > 1 |
3f7c398e | 456 | && memNE(SvPVX_const(check), s, slen))) |
2c2d71f5 | 457 | goto report_neq; |
c315bfe8 | 458 | check_at = s; |
2c2d71f5 | 459 | goto success_at_start; |
7e25d62c | 460 | } |
cad2e5aa | 461 | } |
2c2d71f5 | 462 | /* Match is anchored, but substr is not anchored wrt beg-of-str. */ |
cad2e5aa | 463 | s = strpos; |
2c2d71f5 | 464 | start_shift = prog->check_offset_min; /* okay to underestimate on CC */ |
2c2d71f5 | 465 | end_shift = prog->minlen - start_shift - |
653099ff | 466 | CHR_SVLEN(check) + (SvTAIL(check) != 0); |
2c2d71f5 | 467 | if (!ml_anch) { |
a3b680e6 | 468 | const I32 end = prog->check_offset_max + CHR_SVLEN(check) |
653099ff | 469 | - (SvTAIL(check) != 0); |
a3b680e6 | 470 | const I32 eshift = CHR_DIST((U8*)strend, (U8*)s) - end; |
2c2d71f5 JH |
471 | |
472 | if (end_shift < eshift) | |
473 | end_shift = eshift; | |
474 | } | |
cad2e5aa | 475 | } |
2c2d71f5 | 476 | else { /* Can match at random position */ |
cad2e5aa JH |
477 | ml_anch = 0; |
478 | s = strpos; | |
2c2d71f5 JH |
479 | start_shift = prog->check_offset_min; /* okay to underestimate on CC */ |
480 | /* Should be nonnegative! */ | |
481 | end_shift = prog->minlen - start_shift - | |
653099ff | 482 | CHR_SVLEN(check) + (SvTAIL(check) != 0); |
cad2e5aa JH |
483 | } |
484 | ||
2c2d71f5 | 485 | #ifdef DEBUGGING /* 7/99: reports of failure (with the older version) */ |
0033605d | 486 | if (end_shift < 0) |
6bbae5e6 | 487 | Perl_croak(aTHX_ "panic: end_shift"); |
2c2d71f5 JH |
488 | #endif |
489 | ||
2c2d71f5 JH |
490 | restart: |
491 | /* Find a possible match in the region s..strend by looking for | |
492 | the "check" substring in the region corrected by start/end_shift. */ | |
cad2e5aa | 493 | if (flags & REXEC_SCREAM) { |
cad2e5aa | 494 | I32 p = -1; /* Internal iterator of scream. */ |
a3b680e6 | 495 | I32 * const pp = data ? data->scream_pos : &p; |
cad2e5aa | 496 | |
2c2d71f5 JH |
497 | if (PL_screamfirst[BmRARE(check)] >= 0 |
498 | || ( BmRARE(check) == '\n' | |
499 | && (BmPREVIOUS(check) == SvCUR(check) - 1) | |
500 | && SvTAIL(check) )) | |
9041c2e3 | 501 | s = screaminstr(sv, check, |
2c2d71f5 | 502 | start_shift + (s - strbeg), end_shift, pp, 0); |
cad2e5aa | 503 | else |
2c2d71f5 | 504 | goto fail_finish; |
4addbd3b HS |
505 | /* we may be pointing at the wrong string */ |
506 | if (s && RX_MATCH_COPIED(prog)) | |
3f7c398e | 507 | s = strbeg + (s - SvPVX_const(sv)); |
cad2e5aa JH |
508 | if (data) |
509 | *data->scream_olds = s; | |
510 | } | |
f33976b4 | 511 | else if (prog->reganch & ROPT_CANY_SEEN) |
3baa4c62 JH |
512 | s = fbm_instr((U8*)(s + start_shift), |
513 | (U8*)(strend - end_shift), | |
7fba1cd6 | 514 | check, multiline ? FBMrf_MULTILINE : 0); |
cad2e5aa | 515 | else |
1aa99e6b IH |
516 | s = fbm_instr(HOP3(s, start_shift, strend), |
517 | HOP3(strend, -end_shift, strbeg), | |
7fba1cd6 | 518 | check, multiline ? FBMrf_MULTILINE : 0); |
cad2e5aa JH |
519 | |
520 | /* Update the count-of-usability, remove useless subpatterns, | |
521 | unshift s. */ | |
2c2d71f5 | 522 | |
a0288114 | 523 | DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "%s %s substr \"%s%.*s%s\"%s%s", |
2c2d71f5 | 524 | (s ? "Found" : "Did not find"), |
33b8afdf | 525 | (check == (do_utf8 ? prog->anchored_utf8 : prog->anchored_substr) ? "anchored" : "floating"), |
2c2d71f5 | 526 | PL_colors[0], |
7b0972df | 527 | (int)(SvCUR(check) - (SvTAIL(check)!=0)), |
3f7c398e | 528 | SvPVX_const(check), |
2c2d71f5 JH |
529 | PL_colors[1], (SvTAIL(check) ? "$" : ""), |
530 | (s ? " at offset " : "...\n") ) ); | |
531 | ||
532 | if (!s) | |
533 | goto fail_finish; | |
534 | ||
6eb5f6b9 JH |
535 | check_at = s; |
536 | ||
2c2d71f5 | 537 | /* Finish the diagnostic message */ |
a3621e74 | 538 | DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "%ld...\n", (long)(s - i_strpos)) ); |
2c2d71f5 JH |
539 | |
540 | /* Got a candidate. Check MBOL anchoring, and the *other* substr. | |
541 | Start with the other substr. | |
542 | XXXX no SCREAM optimization yet - and a very coarse implementation | |
a0288114 | 543 | XXXX /ttx+/ results in anchored="ttx", floating="x". floating will |
2c2d71f5 JH |
544 | *always* match. Probably should be marked during compile... |
545 | Probably it is right to do no SCREAM here... | |
546 | */ | |
547 | ||
33b8afdf | 548 | if (do_utf8 ? (prog->float_utf8 && prog->anchored_utf8) : (prog->float_substr && prog->anchored_substr)) { |
30944b6d | 549 | /* Take into account the "other" substring. */ |
2c2d71f5 JH |
550 | /* XXXX May be hopelessly wrong for UTF... */ |
551 | if (!other_last) | |
6eb5f6b9 | 552 | other_last = strpos; |
33b8afdf | 553 | if (check == (do_utf8 ? prog->float_utf8 : prog->float_substr)) { |
30944b6d IZ |
554 | do_other_anchored: |
555 | { | |
890ce7af AL |
556 | char * const last = HOP3c(s, -start_shift, strbeg); |
557 | char *last1, *last2; | |
be8e71aa | 558 | char * const saved_s = s; |
33b8afdf | 559 | SV* must; |
2c2d71f5 | 560 | |
2c2d71f5 JH |
561 | t = s - prog->check_offset_max; |
562 | if (s - strpos > prog->check_offset_max /* signed-corrected t > strpos */ | |
1d86a7f9 | 563 | && (!do_utf8 |
0ce71af7 | 564 | || ((t = (char*)reghopmaybe3((U8*)s, -(prog->check_offset_max), (U8*)strpos)) |
2c2d71f5 | 565 | && t > strpos))) |
6f207bd3 | 566 | NOOP; |
2c2d71f5 JH |
567 | else |
568 | t = strpos; | |
1aa99e6b | 569 | t = HOP3c(t, prog->anchored_offset, strend); |
6eb5f6b9 JH |
570 | if (t < other_last) /* These positions already checked */ |
571 | t = other_last; | |
1aa99e6b | 572 | last2 = last1 = HOP3c(strend, -prog->minlen, strbeg); |
2c2d71f5 JH |
573 | if (last < last1) |
574 | last1 = last; | |
575 | /* XXXX It is not documented what units *_offsets are in. Assume bytes. */ | |
576 | /* On end-of-str: see comment below. */ | |
33b8afdf JH |
577 | must = do_utf8 ? prog->anchored_utf8 : prog->anchored_substr; |
578 | if (must == &PL_sv_undef) { | |
579 | s = (char*)NULL; | |
a3621e74 | 580 | DEBUG_EXECUTE_r(must = prog->anchored_utf8); /* for debug */ |
33b8afdf JH |
581 | } |
582 | else | |
583 | s = fbm_instr( | |
584 | (unsigned char*)t, | |
585 | HOP3(HOP3(last1, prog->anchored_offset, strend) | |
586 | + SvCUR(must), -(SvTAIL(must)!=0), strbeg), | |
587 | must, | |
7fba1cd6 | 588 | multiline ? FBMrf_MULTILINE : 0 |
33b8afdf | 589 | ); |
a3621e74 | 590 | DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, |
a0288114 | 591 | "%s anchored substr \"%s%.*s%s\"%s", |
2c2d71f5 JH |
592 | (s ? "Found" : "Contradicts"), |
593 | PL_colors[0], | |
33b8afdf JH |
594 | (int)(SvCUR(must) |
595 | - (SvTAIL(must)!=0)), | |
3f7c398e | 596 | SvPVX_const(must), |
33b8afdf | 597 | PL_colors[1], (SvTAIL(must) ? "$" : ""))); |
2c2d71f5 JH |
598 | if (!s) { |
599 | if (last1 >= last2) { | |
a3621e74 | 600 | DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, |
2c2d71f5 JH |
601 | ", giving up...\n")); |
602 | goto fail_finish; | |
603 | } | |
a3621e74 | 604 | DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, |
2c2d71f5 | 605 | ", trying floating at offset %ld...\n", |
be8e71aa | 606 | (long)(HOP3c(saved_s, 1, strend) - i_strpos))); |
1aa99e6b IH |
607 | other_last = HOP3c(last1, prog->anchored_offset+1, strend); |
608 | s = HOP3c(last, 1, strend); | |
2c2d71f5 JH |
609 | goto restart; |
610 | } | |
611 | else { | |
a3621e74 | 612 | DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, " at offset %ld...\n", |
30944b6d | 613 | (long)(s - i_strpos))); |
1aa99e6b IH |
614 | t = HOP3c(s, -prog->anchored_offset, strbeg); |
615 | other_last = HOP3c(s, 1, strend); | |
be8e71aa | 616 | s = saved_s; |
2c2d71f5 JH |
617 | if (t == strpos) |
618 | goto try_at_start; | |
2c2d71f5 JH |
619 | goto try_at_offset; |
620 | } | |
30944b6d | 621 | } |
2c2d71f5 JH |
622 | } |
623 | else { /* Take into account the floating substring. */ | |
33b8afdf | 624 | char *last, *last1; |
be8e71aa | 625 | char * const saved_s = s; |
33b8afdf JH |
626 | SV* must; |
627 | ||
628 | t = HOP3c(s, -start_shift, strbeg); | |
629 | last1 = last = | |
630 | HOP3c(strend, -prog->minlen + prog->float_min_offset, strbeg); | |
631 | if (CHR_DIST((U8*)last, (U8*)t) > prog->float_max_offset) | |
632 | last = HOP3c(t, prog->float_max_offset, strend); | |
633 | s = HOP3c(t, prog->float_min_offset, strend); | |
634 | if (s < other_last) | |
635 | s = other_last; | |
2c2d71f5 | 636 | /* XXXX It is not documented what units *_offsets are in. Assume bytes. */ |
33b8afdf JH |
637 | must = do_utf8 ? prog->float_utf8 : prog->float_substr; |
638 | /* fbm_instr() takes into account exact value of end-of-str | |
639 | if the check is SvTAIL(ed). Since false positives are OK, | |
640 | and end-of-str is not later than strend we are OK. */ | |
641 | if (must == &PL_sv_undef) { | |
642 | s = (char*)NULL; | |
a3621e74 | 643 | DEBUG_EXECUTE_r(must = prog->float_utf8); /* for debug message */ |
33b8afdf JH |
644 | } |
645 | else | |
2c2d71f5 | 646 | s = fbm_instr((unsigned char*)s, |
33b8afdf JH |
647 | (unsigned char*)last + SvCUR(must) |
648 | - (SvTAIL(must)!=0), | |
7fba1cd6 | 649 | must, multiline ? FBMrf_MULTILINE : 0); |
a0288114 | 650 | DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "%s floating substr \"%s%.*s%s\"%s", |
33b8afdf JH |
651 | (s ? "Found" : "Contradicts"), |
652 | PL_colors[0], | |
653 | (int)(SvCUR(must) - (SvTAIL(must)!=0)), | |
3f7c398e | 654 | SvPVX_const(must), |
33b8afdf JH |
655 | PL_colors[1], (SvTAIL(must) ? "$" : ""))); |
656 | if (!s) { | |
657 | if (last1 == last) { | |
a3621e74 | 658 | DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, |
33b8afdf JH |
659 | ", giving up...\n")); |
660 | goto fail_finish; | |
2c2d71f5 | 661 | } |
a3621e74 | 662 | DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, |
33b8afdf | 663 | ", trying anchored starting at offset %ld...\n", |
be8e71aa | 664 | (long)(saved_s + 1 - i_strpos))); |
33b8afdf JH |
665 | other_last = last; |
666 | s = HOP3c(t, 1, strend); | |
667 | goto restart; | |
668 | } | |
669 | else { | |
a3621e74 | 670 | DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, " at offset %ld...\n", |
33b8afdf JH |
671 | (long)(s - i_strpos))); |
672 | other_last = s; /* Fix this later. --Hugo */ | |
be8e71aa | 673 | s = saved_s; |
33b8afdf JH |
674 | if (t == strpos) |
675 | goto try_at_start; | |
676 | goto try_at_offset; | |
677 | } | |
2c2d71f5 | 678 | } |
cad2e5aa | 679 | } |
2c2d71f5 JH |
680 | |
681 | t = s - prog->check_offset_max; | |
2c2d71f5 | 682 | if (s - strpos > prog->check_offset_max /* signed-corrected t > strpos */ |
1d86a7f9 | 683 | && (!do_utf8 |
0ce71af7 | 684 | || ((t = (char*)reghopmaybe3((U8*)s, -prog->check_offset_max, (U8*)strpos)) |
1aa99e6b | 685 | && t > strpos))) { |
2c2d71f5 JH |
686 | /* Fixed substring is found far enough so that the match |
687 | cannot start at strpos. */ | |
688 | try_at_offset: | |
cad2e5aa | 689 | if (ml_anch && t[-1] != '\n') { |
30944b6d IZ |
690 | /* Eventually fbm_*() should handle this, but often |
691 | anchored_offset is not 0, so this check will not be wasted. */ | |
692 | /* XXXX In the code below we prefer to look for "^" even in | |
693 | presence of anchored substrings. And we search even | |
694 | beyond the found float position. These pessimizations | |
695 | are historical artefacts only. */ | |
696 | find_anchor: | |
2c2d71f5 | 697 | while (t < strend - prog->minlen) { |
cad2e5aa | 698 | if (*t == '\n') { |
4ee3650e | 699 | if (t < check_at - prog->check_offset_min) { |
33b8afdf | 700 | if (do_utf8 ? prog->anchored_utf8 : prog->anchored_substr) { |
4ee3650e GS |
701 | /* Since we moved from the found position, |
702 | we definitely contradict the found anchored | |
30944b6d IZ |
703 | substr. Due to the above check we do not |
704 | contradict "check" substr. | |
705 | Thus we can arrive here only if check substr | |
706 | is float. Redo checking for "other"=="fixed". | |
707 | */ | |
9041c2e3 | 708 | strpos = t + 1; |
a3621e74 | 709 | DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "Found /%s^%s/m at offset %ld, rescanning for anchored from offset %ld...\n", |
e4584336 | 710 | PL_colors[0], PL_colors[1], (long)(strpos - i_strpos), (long)(strpos - i_strpos + prog->anchored_offset))); |
30944b6d IZ |
711 | goto do_other_anchored; |
712 | } | |
4ee3650e GS |
713 | /* We don't contradict the found floating substring. */ |
714 | /* XXXX Why not check for STCLASS? */ | |
cad2e5aa | 715 | s = t + 1; |
a3621e74 | 716 | DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "Found /%s^%s/m at offset %ld...\n", |
e4584336 | 717 | PL_colors[0], PL_colors[1], (long)(s - i_strpos))); |
cad2e5aa JH |
718 | goto set_useful; |
719 | } | |
4ee3650e GS |
720 | /* Position contradicts check-string */ |
721 | /* XXXX probably better to look for check-string | |
722 | than for "\n", so one should lower the limit for t? */ | |
a3621e74 | 723 | DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "Found /%s^%s/m, restarting lookup for check-string at offset %ld...\n", |
e4584336 | 724 | PL_colors[0], PL_colors[1], (long)(t + 1 - i_strpos))); |
0e41cd87 | 725 | other_last = strpos = s = t + 1; |
cad2e5aa JH |
726 | goto restart; |
727 | } | |
728 | t++; | |
729 | } | |
a3621e74 | 730 | DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "Did not find /%s^%s/m...\n", |
e4584336 | 731 | PL_colors[0], PL_colors[1])); |
2c2d71f5 | 732 | goto fail_finish; |
cad2e5aa | 733 | } |
f5952150 | 734 | else { |
a3621e74 | 735 | DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "Starting position does not contradict /%s^%s/m...\n", |
e4584336 | 736 | PL_colors[0], PL_colors[1])); |
f5952150 | 737 | } |
cad2e5aa JH |
738 | s = t; |
739 | set_useful: | |
33b8afdf | 740 | ++BmUSEFUL(do_utf8 ? prog->check_utf8 : prog->check_substr); /* hooray/5 */ |
cad2e5aa JH |
741 | } |
742 | else { | |
f5952150 | 743 | /* The found string does not prohibit matching at strpos, |
2c2d71f5 | 744 | - no optimization of calling REx engine can be performed, |
f5952150 GS |
745 | unless it was an MBOL and we are not after MBOL, |
746 | or a future STCLASS check will fail this. */ | |
2c2d71f5 JH |
747 | try_at_start: |
748 | /* Even in this situation we may use MBOL flag if strpos is offset | |
749 | wrt the start of the string. */ | |
05b4157f | 750 | if (ml_anch && sv && !SvROK(sv) /* See prev comment on SvROK */ |
a1933d95 | 751 | && (strpos != strbeg) && strpos[-1] != '\n' |
d506a20d IZ |
752 | /* May be due to an implicit anchor of m{.*foo} */ |
753 | && !(prog->reganch & ROPT_IMPLICIT)) | |
754 | { | |
cad2e5aa JH |
755 | t = strpos; |
756 | goto find_anchor; | |
757 | } | |
a3621e74 | 758 | DEBUG_EXECUTE_r( if (ml_anch) |
f5952150 | 759 | PerlIO_printf(Perl_debug_log, "Position at offset %ld does not contradict /%s^%s/m...\n", |
e4584336 | 760 | (long)(strpos - i_strpos), PL_colors[0], PL_colors[1]); |
30944b6d | 761 | ); |
2c2d71f5 | 762 | success_at_start: |
30944b6d | 763 | if (!(prog->reganch & ROPT_NAUGHTY) /* XXXX If strpos moved? */ |
33b8afdf JH |
764 | && (do_utf8 ? ( |
765 | prog->check_utf8 /* Could be deleted already */ | |
766 | && --BmUSEFUL(prog->check_utf8) < 0 | |
767 | && (prog->check_utf8 == prog->float_utf8) | |
768 | ) : ( | |
769 | prog->check_substr /* Could be deleted already */ | |
770 | && --BmUSEFUL(prog->check_substr) < 0 | |
771 | && (prog->check_substr == prog->float_substr) | |
772 | ))) | |
66e933ab | 773 | { |
cad2e5aa | 774 | /* If flags & SOMETHING - do not do it many times on the same match */ |
a3621e74 | 775 | DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "... Disabling check substring...\n")); |
33b8afdf JH |
776 | SvREFCNT_dec(do_utf8 ? prog->check_utf8 : prog->check_substr); |
777 | if (do_utf8 ? prog->check_substr : prog->check_utf8) | |
778 | SvREFCNT_dec(do_utf8 ? prog->check_substr : prog->check_utf8); | |
a0714e2c SS |
779 | prog->check_substr = prog->check_utf8 = NULL; /* disable */ |
780 | prog->float_substr = prog->float_utf8 = NULL; /* clear */ | |
781 | check = NULL; /* abort */ | |
cad2e5aa | 782 | s = strpos; |
3cf5c195 IZ |
783 | /* XXXX This is a remnant of the old implementation. It |
784 | looks wasteful, since now INTUIT can use many | |
6eb5f6b9 | 785 | other heuristics. */ |
cad2e5aa JH |
786 | prog->reganch &= ~RE_USE_INTUIT; |
787 | } | |
788 | else | |
789 | s = strpos; | |
790 | } | |
791 | ||
6eb5f6b9 JH |
792 | /* Last resort... */ |
793 | /* XXXX BmUSEFUL already changed, maybe multiple change is meaningful... */ | |
07be1b83 | 794 | if (prog->regstclass && OP(prog->regstclass)!=TRIE) { |
6eb5f6b9 JH |
795 | /* minlen == 0 is possible if regstclass is \b or \B, |
796 | and the fixed substr is ''$. | |
797 | Since minlen is already taken into account, s+1 is before strend; | |
798 | accidentally, minlen >= 1 guaranties no false positives at s + 1 | |
799 | even for \b or \B. But (minlen? 1 : 0) below assumes that | |
800 | regstclass does not come from lookahead... */ | |
801 | /* If regstclass takes bytelength more than 1: If charlength==1, OK. | |
802 | This leaves EXACTF only, which is dealt with in find_byclass(). */ | |
890ce7af | 803 | const U8* const str = (U8*)STRING(prog->regstclass); |
3dab1dad | 804 | const int cl_l = (PL_regkind[OP(prog->regstclass)] == EXACT |
1aa99e6b | 805 | ? CHR_DIST(str+STR_LEN(prog->regstclass), str) |
66e933ab | 806 | : 1); |
07be1b83 | 807 | const char * endpos = (prog->anchored_substr || prog->anchored_utf8 || ml_anch) |
1aa99e6b | 808 | ? HOP3c(s, (prog->minlen ? cl_l : 0), strend) |
33b8afdf | 809 | : (prog->float_substr || prog->float_utf8 |
1aa99e6b IH |
810 | ? HOP3c(HOP3c(check_at, -start_shift, strbeg), |
811 | cl_l, strend) | |
812 | : strend); | |
07be1b83 YO |
813 | /*if (OP(prog->regstclass) == TRIE) |
814 | endpos++;*/ | |
6eb5f6b9 | 815 | t = s; |
3b0527fe | 816 | s = find_byclass(prog, prog->regstclass, s, endpos, NULL); |
6eb5f6b9 JH |
817 | if (!s) { |
818 | #ifdef DEBUGGING | |
cbbf8932 | 819 | const char *what = NULL; |
6eb5f6b9 JH |
820 | #endif |
821 | if (endpos == strend) { | |
a3621e74 | 822 | DEBUG_EXECUTE_r( PerlIO_printf(Perl_debug_log, |
6eb5f6b9 JH |
823 | "Could not match STCLASS...\n") ); |
824 | goto fail; | |
825 | } | |
a3621e74 | 826 | DEBUG_EXECUTE_r( PerlIO_printf(Perl_debug_log, |
66e933ab | 827 | "This position contradicts STCLASS...\n") ); |
653099ff GS |
828 | if ((prog->reganch & ROPT_ANCH) && !ml_anch) |
829 | goto fail; | |
6eb5f6b9 | 830 | /* Contradict one of substrings */ |
33b8afdf JH |
831 | if (prog->anchored_substr || prog->anchored_utf8) { |
832 | if ((do_utf8 ? prog->anchored_utf8 : prog->anchored_substr) == check) { | |
a3621e74 | 833 | DEBUG_EXECUTE_r( what = "anchored" ); |
6eb5f6b9 | 834 | hop_and_restart: |
1aa99e6b | 835 | s = HOP3c(t, 1, strend); |
66e933ab GS |
836 | if (s + start_shift + end_shift > strend) { |
837 | /* XXXX Should be taken into account earlier? */ | |
a3621e74 | 838 | DEBUG_EXECUTE_r( PerlIO_printf(Perl_debug_log, |
66e933ab GS |
839 | "Could not match STCLASS...\n") ); |
840 | goto fail; | |
841 | } | |
5e39e1e5 HS |
842 | if (!check) |
843 | goto giveup; | |
a3621e74 | 844 | DEBUG_EXECUTE_r( PerlIO_printf(Perl_debug_log, |
f5952150 | 845 | "Looking for %s substr starting at offset %ld...\n", |
6eb5f6b9 JH |
846 | what, (long)(s + start_shift - i_strpos)) ); |
847 | goto restart; | |
848 | } | |
66e933ab | 849 | /* Have both, check_string is floating */ |
6eb5f6b9 JH |
850 | if (t + start_shift >= check_at) /* Contradicts floating=check */ |
851 | goto retry_floating_check; | |
852 | /* Recheck anchored substring, but not floating... */ | |
9041c2e3 | 853 | s = check_at; |
5e39e1e5 HS |
854 | if (!check) |
855 | goto giveup; | |
a3621e74 | 856 | DEBUG_EXECUTE_r( PerlIO_printf(Perl_debug_log, |
f5952150 | 857 | "Looking for anchored substr starting at offset %ld...\n", |
6eb5f6b9 JH |
858 | (long)(other_last - i_strpos)) ); |
859 | goto do_other_anchored; | |
860 | } | |
60e71179 GS |
861 | /* Another way we could have checked stclass at the |
862 | current position only: */ | |
863 | if (ml_anch) { | |
864 | s = t = t + 1; | |
5e39e1e5 HS |
865 | if (!check) |
866 | goto giveup; | |
a3621e74 | 867 | DEBUG_EXECUTE_r( PerlIO_printf(Perl_debug_log, |
f5952150 | 868 | "Looking for /%s^%s/m starting at offset %ld...\n", |
e4584336 | 869 | PL_colors[0], PL_colors[1], (long)(t - i_strpos)) ); |
60e71179 | 870 | goto try_at_offset; |
66e933ab | 871 | } |
33b8afdf | 872 | if (!(do_utf8 ? prog->float_utf8 : prog->float_substr)) /* Could have been deleted */ |
60e71179 | 873 | goto fail; |
6eb5f6b9 JH |
874 | /* Check is floating subtring. */ |
875 | retry_floating_check: | |
876 | t = check_at - start_shift; | |
a3621e74 | 877 | DEBUG_EXECUTE_r( what = "floating" ); |
6eb5f6b9 JH |
878 | goto hop_and_restart; |
879 | } | |
b7953727 | 880 | if (t != s) { |
a3621e74 | 881 | DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, |
6eb5f6b9 | 882 | "By STCLASS: moving %ld --> %ld\n", |
b7953727 JH |
883 | (long)(t - i_strpos), (long)(s - i_strpos)) |
884 | ); | |
885 | } | |
886 | else { | |
a3621e74 | 887 | DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, |
b7953727 JH |
888 | "Does not contradict STCLASS...\n"); |
889 | ); | |
890 | } | |
6eb5f6b9 | 891 | } |
5e39e1e5 | 892 | giveup: |
a3621e74 | 893 | DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "%s%s:%s match at offset %ld\n", |
5e39e1e5 HS |
894 | PL_colors[4], (check ? "Guessed" : "Giving up"), |
895 | PL_colors[5], (long)(s - i_strpos)) ); | |
cad2e5aa | 896 | return s; |
2c2d71f5 JH |
897 | |
898 | fail_finish: /* Substring not found */ | |
33b8afdf JH |
899 | if (prog->check_substr || prog->check_utf8) /* could be removed already */ |
900 | BmUSEFUL(do_utf8 ? prog->check_utf8 : prog->check_substr) += 5; /* hooray */ | |
cad2e5aa | 901 | fail: |
a3621e74 | 902 | DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "%sMatch rejected by optimizer%s\n", |
e4584336 | 903 | PL_colors[4], PL_colors[5])); |
bd61b366 | 904 | return NULL; |
cad2e5aa | 905 | } |
9661b544 | 906 | |
6eb5f6b9 | 907 | /* We know what class REx starts with. Try to find this position... */ |
3b0527fe | 908 | /* if reginfo is NULL, its a dryrun */ |
07be1b83 YO |
909 | /* annoyingly all the vars in this routine have different names from their counterparts |
910 | in regmatch. /grrr */ | |
3b0527fe | 911 | |
4cadc6a9 YO |
912 | #define REXEC_TRIE_READ_CHAR(trie_type, trie, uc, uscan, len, uvc, charid, \ |
913 | foldlen, foldbuf, uniflags) STMT_START { \ | |
914 | switch (trie_type) { \ | |
915 | case trie_utf8_fold: \ | |
916 | if ( foldlen>0 ) { \ | |
917 | uvc = utf8n_to_uvuni( uscan, UTF8_MAXLEN, &len, uniflags ); \ | |
918 | foldlen -= len; \ | |
919 | uscan += len; \ | |
920 | len=0; \ | |
921 | } else { \ | |
922 | uvc = utf8n_to_uvuni( (U8*)uc, UTF8_MAXLEN, &len, uniflags ); \ | |
923 | uvc = to_uni_fold( uvc, foldbuf, &foldlen ); \ | |
924 | foldlen -= UNISKIP( uvc ); \ | |
925 | uscan = foldbuf + UNISKIP( uvc ); \ | |
926 | } \ | |
927 | break; \ | |
928 | case trie_utf8: \ | |
929 | uvc = utf8n_to_uvuni( (U8*)uc, UTF8_MAXLEN, &len, uniflags ); \ | |
930 | break; \ | |
931 | case trie_plain: \ | |
932 | uvc = (UV)*uc; \ | |
933 | len = 1; \ | |
934 | } \ | |
935 | \ | |
936 | if (uvc < 256) { \ | |
937 | charid = trie->charmap[ uvc ]; \ | |
938 | } \ | |
939 | else { \ | |
940 | charid = 0; \ | |
941 | if (trie->widecharmap) { \ | |
942 | SV** const svpp = hv_fetch(trie->widecharmap, \ | |
943 | (char*)&uvc, sizeof(UV), 0); \ | |
944 | if (svpp) \ | |
945 | charid = (U16)SvIV(*svpp); \ | |
946 | } \ | |
947 | } \ | |
948 | } STMT_END | |
949 | ||
950 | #define REXEC_FBC_EXACTISH_CHECK(CoNd) \ | |
951 | if ( (CoNd) \ | |
952 | && (ln == len || \ | |
953 | ibcmp_utf8(s, NULL, 0, do_utf8, \ | |
954 | m, NULL, ln, (bool)UTF)) \ | |
955 | && (!reginfo || regtry(reginfo, s)) ) \ | |
956 | goto got_it; \ | |
957 | else { \ | |
958 | U8 foldbuf[UTF8_MAXBYTES_CASE+1]; \ | |
959 | uvchr_to_utf8(tmpbuf, c); \ | |
960 | f = to_utf8_fold(tmpbuf, foldbuf, &foldlen); \ | |
961 | if ( f != c \ | |
962 | && (f == c1 || f == c2) \ | |
963 | && (ln == foldlen || \ | |
964 | !ibcmp_utf8((char *) foldbuf, \ | |
965 | NULL, foldlen, do_utf8, \ | |
966 | m, \ | |
967 | NULL, ln, (bool)UTF)) \ | |
968 | && (!reginfo || regtry(reginfo, s)) ) \ | |
969 | goto got_it; \ | |
970 | } \ | |
971 | s += len | |
972 | ||
973 | #define REXEC_FBC_EXACTISH_SCAN(CoNd) \ | |
974 | STMT_START { \ | |
975 | while (s <= e) { \ | |
976 | if ( (CoNd) \ | |
977 | && (ln == 1 || !(OP(c) == EXACTF \ | |
978 | ? ibcmp(s, m, ln) \ | |
979 | : ibcmp_locale(s, m, ln))) \ | |
980 | && (!reginfo || regtry(reginfo, s)) ) \ | |
981 | goto got_it; \ | |
982 | s++; \ | |
983 | } \ | |
984 | } STMT_END | |
985 | ||
986 | #define REXEC_FBC_UTF8_SCAN(CoDe) \ | |
987 | STMT_START { \ | |
988 | while (s + (uskip = UTF8SKIP(s)) <= strend) { \ | |
989 | CoDe \ | |
990 | s += uskip; \ | |
991 | } \ | |
992 | } STMT_END | |
993 | ||
994 | #define REXEC_FBC_SCAN(CoDe) \ | |
995 | STMT_START { \ | |
996 | while (s < strend) { \ | |
997 | CoDe \ | |
998 | s++; \ | |
999 | } \ | |
1000 | } STMT_END | |
1001 | ||
1002 | #define REXEC_FBC_UTF8_CLASS_SCAN(CoNd) \ | |
1003 | REXEC_FBC_UTF8_SCAN( \ | |
1004 | if (CoNd) { \ | |
1005 | if (tmp && (!reginfo || regtry(reginfo, s))) \ | |
1006 | goto got_it; \ | |
1007 | else \ | |
1008 | tmp = doevery; \ | |
1009 | } \ | |
1010 | else \ | |
1011 | tmp = 1; \ | |
1012 | ) | |
1013 | ||
1014 | #define REXEC_FBC_CLASS_SCAN(CoNd) \ | |
1015 | REXEC_FBC_SCAN( \ | |
1016 | if (CoNd) { \ | |
1017 | if (tmp && (!reginfo || regtry(reginfo, s))) \ | |
1018 | goto got_it; \ | |
1019 | else \ | |
1020 | tmp = doevery; \ | |
1021 | } \ | |
1022 | else \ | |
1023 | tmp = 1; \ | |
1024 | ) | |
1025 | ||
1026 | #define REXEC_FBC_TRYIT \ | |
1027 | if ((!reginfo || regtry(reginfo, s))) \ | |
1028 | goto got_it | |
1029 | ||
1030 | #define REXEC_FBC_CSCAN_PRELOAD(UtFpReLoAd,CoNdUtF8,CoNd) \ | |
1031 | if (do_utf8) { \ | |
1032 | UtFpReLoAd; \ | |
1033 | REXEC_FBC_UTF8_CLASS_SCAN(CoNdUtF8); \ | |
1034 | } \ | |
1035 | else { \ | |
1036 | REXEC_FBC_CLASS_SCAN(CoNd); \ | |
1037 | } \ | |
1038 | break | |
1039 | ||
1040 | #define REXEC_FBC_CSCAN_TAINT(CoNdUtF8,CoNd) \ | |
1041 | PL_reg_flags |= RF_tainted; \ | |
1042 | if (do_utf8) { \ | |
1043 | REXEC_FBC_UTF8_CLASS_SCAN(CoNdUtF8); \ | |
1044 | } \ | |
1045 | else { \ | |
1046 | REXEC_FBC_CLASS_SCAN(CoNd); \ | |
1047 | } \ | |
1048 | break | |
1049 | ||
3c3eec57 | 1050 | STATIC char * |
07be1b83 YO |
1051 | S_find_byclass(pTHX_ regexp * prog, const regnode *c, char *s, |
1052 | const char *strend, const regmatch_info *reginfo) | |
a687059c | 1053 | { |
27da23d5 | 1054 | dVAR; |
1df70142 | 1055 | const I32 doevery = (prog->reganch & ROPT_SKIP) == 0; |
6eb5f6b9 | 1056 | char *m; |
d8093b23 | 1057 | STRLEN ln; |
5dab1207 | 1058 | STRLEN lnc; |
078c425b | 1059 | register STRLEN uskip; |
d8093b23 G |
1060 | unsigned int c1; |
1061 | unsigned int c2; | |
6eb5f6b9 JH |
1062 | char *e; |
1063 | register I32 tmp = 1; /* Scratch variable? */ | |
a3b680e6 | 1064 | register const bool do_utf8 = PL_reg_match_utf8; |
cad2e5aa | 1065 | |
6eb5f6b9 JH |
1066 | /* We know what class it must start with. */ |
1067 | switch (OP(c)) { | |
6eb5f6b9 | 1068 | case ANYOF: |
388cc4de | 1069 | if (do_utf8) { |
4cadc6a9 | 1070 | REXEC_FBC_UTF8_CLASS_SCAN((ANYOF_FLAGS(c) & ANYOF_UNICODE) || |
388cc4de | 1071 | !UTF8_IS_INVARIANT((U8)s[0]) ? |
32fc9b6a | 1072 | reginclass(prog, c, (U8*)s, 0, do_utf8) : |
4cadc6a9 | 1073 | REGINCLASS(prog, c, (U8*)s)); |
388cc4de HS |
1074 | } |
1075 | else { | |
1076 | while (s < strend) { | |
1077 | STRLEN skip = 1; | |
1078 | ||
32fc9b6a | 1079 | if (REGINCLASS(prog, c, (U8*)s) || |
388cc4de HS |
1080 | (ANYOF_FOLD_SHARP_S(c, s, strend) && |
1081 | /* The assignment of 2 is intentional: | |
1082 | * for the folded sharp s, the skip is 2. */ | |
1083 | (skip = SHARP_S_SKIP))) { | |
3b0527fe | 1084 | if (tmp && (!reginfo || regtry(reginfo, s))) |
388cc4de HS |
1085 | goto got_it; |
1086 | else | |
1087 | tmp = doevery; | |
1088 | } | |
1089 | else | |
1090 | tmp = 1; | |
1091 | s += skip; | |
1092 | } | |
a0d0e21e | 1093 | } |
6eb5f6b9 | 1094 | break; |
f33976b4 | 1095 | case CANY: |
4cadc6a9 | 1096 | REXEC_FBC_SCAN( |
3b0527fe | 1097 | if (tmp && (!reginfo || regtry(reginfo, s))) |
f33976b4 DB |
1098 | goto got_it; |
1099 | else | |
1100 | tmp = doevery; | |
4cadc6a9 | 1101 | ); |
f33976b4 | 1102 | break; |
6eb5f6b9 | 1103 | case EXACTF: |
5dab1207 NIS |
1104 | m = STRING(c); |
1105 | ln = STR_LEN(c); /* length to match in octets/bytes */ | |
1106 | lnc = (I32) ln; /* length to match in characters */ | |
1aa99e6b | 1107 | if (UTF) { |
a2a2844f | 1108 | STRLEN ulen1, ulen2; |
5dab1207 | 1109 | U8 *sm = (U8 *) m; |
89ebb4a3 JH |
1110 | U8 tmpbuf1[UTF8_MAXBYTES_CASE+1]; |
1111 | U8 tmpbuf2[UTF8_MAXBYTES_CASE+1]; | |
4ad0818d | 1112 | const U32 uniflags = UTF8_ALLOW_DEFAULT; |
a2a2844f JH |
1113 | |
1114 | to_utf8_lower((U8*)m, tmpbuf1, &ulen1); | |
1115 | to_utf8_upper((U8*)m, tmpbuf2, &ulen2); | |
1116 | ||
89ebb4a3 | 1117 | c1 = utf8n_to_uvchr(tmpbuf1, UTF8_MAXBYTES_CASE, |
041457d9 | 1118 | 0, uniflags); |
89ebb4a3 | 1119 | c2 = utf8n_to_uvchr(tmpbuf2, UTF8_MAXBYTES_CASE, |
041457d9 | 1120 | 0, uniflags); |
5dab1207 NIS |
1121 | lnc = 0; |
1122 | while (sm < ((U8 *) m + ln)) { | |
1123 | lnc++; | |
1124 | sm += UTF8SKIP(sm); | |
1125 | } | |
1aa99e6b IH |
1126 | } |
1127 | else { | |
1128 | c1 = *(U8*)m; | |
1129 | c2 = PL_fold[c1]; | |
1130 | } | |
6eb5f6b9 JH |
1131 | goto do_exactf; |
1132 | case EXACTFL: | |
5dab1207 NIS |
1133 | m = STRING(c); |
1134 | ln = STR_LEN(c); | |
1135 | lnc = (I32) ln; | |
d8093b23 | 1136 | c1 = *(U8*)m; |
6eb5f6b9 JH |
1137 | c2 = PL_fold_locale[c1]; |
1138 | do_exactf: | |
db12adc6 | 1139 | e = HOP3c(strend, -((I32)lnc), s); |
b3c9acc1 | 1140 | |
3b0527fe | 1141 | if (!reginfo && e < s) |
6eb5f6b9 | 1142 | e = s; /* Due to minlen logic of intuit() */ |
1aa99e6b | 1143 | |
60a8b682 JH |
1144 | /* The idea in the EXACTF* cases is to first find the |
1145 | * first character of the EXACTF* node and then, if | |
1146 | * necessary, case-insensitively compare the full | |
1147 | * text of the node. The c1 and c2 are the first | |
1148 | * characters (though in Unicode it gets a bit | |
1149 | * more complicated because there are more cases | |
7f16dd3d JH |
1150 | * than just upper and lower: one needs to use |
1151 | * the so-called folding case for case-insensitive | |
1152 | * matching (called "loose matching" in Unicode). | |
1153 | * ibcmp_utf8() will do just that. */ | |
60a8b682 | 1154 | |
1aa99e6b | 1155 | if (do_utf8) { |
575cac57 | 1156 | UV c, f; |
89ebb4a3 | 1157 | U8 tmpbuf [UTF8_MAXBYTES+1]; |
575cac57 | 1158 | STRLEN len, foldlen; |
4ad0818d | 1159 | const U32 uniflags = UTF8_ALLOW_DEFAULT; |
09091399 | 1160 | if (c1 == c2) { |
5dab1207 NIS |
1161 | /* Upper and lower of 1st char are equal - |
1162 | * probably not a "letter". */ | |
1aa99e6b | 1163 | while (s <= e) { |
89ebb4a3 | 1164 | c = utf8n_to_uvchr((U8*)s, UTF8_MAXBYTES, &len, |
041457d9 | 1165 | uniflags); |
4cadc6a9 | 1166 | REXEC_FBC_EXACTISH_CHECK(c == c1); |
1aa99e6b | 1167 | } |
09091399 JH |
1168 | } |
1169 | else { | |
1aa99e6b | 1170 | while (s <= e) { |
89ebb4a3 | 1171 | c = utf8n_to_uvchr((U8*)s, UTF8_MAXBYTES, &len, |
041457d9 | 1172 | uniflags); |
80aecb99 | 1173 | |
60a8b682 | 1174 | /* Handle some of the three Greek sigmas cases. |
8c01da3c JH |
1175 | * Note that not all the possible combinations |
1176 | * are handled here: some of them are handled | |
1177 | * by the standard folding rules, and some of | |
1178 | * them (the character class or ANYOF cases) | |
1179 | * are handled during compiletime in | |
1180 | * regexec.c:S_regclass(). */ | |
880bd946 JH |
1181 | if (c == (UV)UNICODE_GREEK_CAPITAL_LETTER_SIGMA || |
1182 | c == (UV)UNICODE_GREEK_SMALL_LETTER_FINAL_SIGMA) | |
1183 | c = (UV)UNICODE_GREEK_SMALL_LETTER_SIGMA; | |
80aecb99 | 1184 | |
4cadc6a9 | 1185 | REXEC_FBC_EXACTISH_CHECK(c == c1 || c == c2); |
1aa99e6b | 1186 | } |
09091399 | 1187 | } |
1aa99e6b IH |
1188 | } |
1189 | else { | |
1190 | if (c1 == c2) | |
4cadc6a9 | 1191 | REXEC_FBC_EXACTISH_SCAN(*(U8*)s == c1); |
1aa99e6b | 1192 | else |
4cadc6a9 | 1193 | REXEC_FBC_EXACTISH_SCAN(*(U8*)s == c1 || *(U8*)s == c2); |
b3c9acc1 IZ |
1194 | } |
1195 | break; | |
bbce6d69 | 1196 | case BOUNDL: |
3280af22 | 1197 | PL_reg_flags |= RF_tainted; |
bbce6d69 | 1198 | /* FALL THROUGH */ |
a0d0e21e | 1199 | case BOUND: |
ffc61ed2 | 1200 | if (do_utf8) { |
12d33761 | 1201 | if (s == PL_bostr) |
ffc61ed2 JH |
1202 | tmp = '\n'; |
1203 | else { | |
6136c704 | 1204 | U8 * const r = reghop3((U8*)s, -1, (U8*)PL_bostr); |
4ad0818d | 1205 | tmp = utf8n_to_uvchr(r, UTF8SKIP(r), 0, UTF8_ALLOW_DEFAULT); |
ffc61ed2 JH |
1206 | } |
1207 | tmp = ((OP(c) == BOUND ? | |
9041c2e3 | 1208 | isALNUM_uni(tmp) : isALNUM_LC_uvchr(UNI_TO_NATIVE(tmp))) != 0); |
1a4fad37 | 1209 | LOAD_UTF8_CHARCLASS_ALNUM(); |
4cadc6a9 | 1210 | REXEC_FBC_UTF8_SCAN( |
ffc61ed2 | 1211 | if (tmp == !(OP(c) == BOUND ? |
bb7a0f54 | 1212 | (bool)swash_fetch(PL_utf8_alnum, (U8*)s, do_utf8) : |
ffc61ed2 JH |
1213 | isALNUM_LC_utf8((U8*)s))) |
1214 | { | |
1215 | tmp = !tmp; | |
4cadc6a9 | 1216 | REXEC_FBC_TRYIT; |
a687059c | 1217 | } |
4cadc6a9 | 1218 | ); |
a0d0e21e | 1219 | } |
667bb95a | 1220 | else { |
12d33761 | 1221 | tmp = (s != PL_bostr) ? UCHARAT(s - 1) : '\n'; |
ffc61ed2 | 1222 | tmp = ((OP(c) == BOUND ? isALNUM(tmp) : isALNUM_LC(tmp)) != 0); |
4cadc6a9 | 1223 | REXEC_FBC_SCAN( |
ffc61ed2 JH |
1224 | if (tmp == |
1225 | !(OP(c) == BOUND ? isALNUM(*s) : isALNUM_LC(*s))) { | |
1226 | tmp = !tmp; | |
4cadc6a9 | 1227 | REXEC_FBC_TRYIT; |
a0ed51b3 | 1228 | } |
4cadc6a9 | 1229 | ); |
a0ed51b3 | 1230 | } |
3b0527fe | 1231 | if ((!prog->minlen && tmp) && (!reginfo || regtry(reginfo, s))) |
a0ed51b3 LW |
1232 | goto got_it; |
1233 | break; | |
bbce6d69 | 1234 | case NBOUNDL: |
3280af22 | 1235 | PL_reg_flags |= RF_tainted; |
bbce6d69 | 1236 | /* FALL THROUGH */ |
a0d0e21e | 1237 | case NBOUND: |
ffc61ed2 | 1238 | if (do_utf8) { |
12d33761 | 1239 | if (s == PL_bostr) |
ffc61ed2 JH |
1240 | tmp = '\n'; |
1241 | else { | |
6136c704 | 1242 | U8 * const r = reghop3((U8*)s, -1, (U8*)PL_bostr); |
4ad0818d | 1243 | tmp = utf8n_to_uvchr(r, UTF8SKIP(r), 0, UTF8_ALLOW_DEFAULT); |
ffc61ed2 JH |
1244 | } |
1245 | tmp = ((OP(c) == NBOUND ? | |
9041c2e3 | 1246 | isALNUM_uni(tmp) : isALNUM_LC_uvchr(UNI_TO_NATIVE(tmp))) != 0); |
1a4fad37 | 1247 | LOAD_UTF8_CHARCLASS_ALNUM(); |
4cadc6a9 | 1248 | REXEC_FBC_UTF8_SCAN( |
ffc61ed2 | 1249 | if (tmp == !(OP(c) == NBOUND ? |
bb7a0f54 | 1250 | (bool)swash_fetch(PL_utf8_alnum, (U8*)s, do_utf8) : |
ffc61ed2 JH |
1251 | isALNUM_LC_utf8((U8*)s))) |
1252 | tmp = !tmp; | |
4cadc6a9 YO |
1253 | else REXEC_FBC_TRYIT; |
1254 | ); | |
a0d0e21e | 1255 | } |
667bb95a | 1256 | else { |
12d33761 | 1257 | tmp = (s != PL_bostr) ? UCHARAT(s - 1) : '\n'; |
ffc61ed2 JH |
1258 | tmp = ((OP(c) == NBOUND ? |
1259 | isALNUM(tmp) : isALNUM_LC(tmp)) != 0); | |
4cadc6a9 | 1260 | REXEC_FBC_SCAN( |
ffc61ed2 JH |
1261 | if (tmp == |
1262 | !(OP(c) == NBOUND ? isALNUM(*s) : isALNUM_LC(*s))) | |
1263 | tmp = !tmp; | |
4cadc6a9 YO |
1264 | else REXEC_FBC_TRYIT; |
1265 | ); | |
a0ed51b3 | 1266 | } |
3b0527fe | 1267 | if ((!prog->minlen && !tmp) && (!reginfo || regtry(reginfo, s))) |
a0ed51b3 LW |
1268 | goto got_it; |
1269 | break; | |
a0d0e21e | 1270 | case ALNUM: |
4cadc6a9 YO |
1271 | REXEC_FBC_CSCAN_PRELOAD( |
1272 | LOAD_UTF8_CHARCLASS_ALNUM(), | |
1273 | swash_fetch(PL_utf8_alnum, (U8*)s, do_utf8), | |
1274 | isALNUM(*s) | |
1275 | ); | |
bbce6d69 | 1276 | case ALNUML: |
4cadc6a9 YO |
1277 | REXEC_FBC_CSCAN_TAINT( |
1278 | isALNUM_LC_utf8((U8*)s), | |
1279 | isALNUM_LC(*s) | |
1280 | ); | |
a0d0e21e | 1281 | case NALNUM: |
4cadc6a9 YO |
1282 | REXEC_FBC_CSCAN_PRELOAD( |
1283 | LOAD_UTF8_CHARCLASS_ALNUM(), | |
1284 | !swash_fetch(PL_utf8_alnum, (U8*)s, do_utf8), | |
1285 | !isALNUM(*s) | |
1286 | ); | |
bbce6d69 | 1287 | case NALNUML: |
4cadc6a9 YO |
1288 | REXEC_FBC_CSCAN_TAINT( |
1289 | !isALNUM_LC_utf8((U8*)s), | |
1290 | !isALNUM_LC(*s) | |
1291 | ); | |
a0d0e21e | 1292 | case SPACE: |
4cadc6a9 YO |
1293 | REXEC_FBC_CSCAN_PRELOAD( |
1294 | LOAD_UTF8_CHARCLASS_SPACE(), | |
1295 | *s == ' ' || swash_fetch(PL_utf8_space,(U8*)s, do_utf8), | |
1296 | isSPACE(*s) | |
1297 | ); | |
bbce6d69 | 1298 | case SPACEL: |
4cadc6a9 YO |
1299 | REXEC_FBC_CSCAN_TAINT( |
1300 | *s == ' ' || isSPACE_LC_utf8((U8*)s), | |
1301 | isSPACE_LC(*s) | |
1302 | ); | |
a0d0e21e | 1303 | case NSPACE: |
4cadc6a9 YO |
1304 | REXEC_FBC_CSCAN_PRELOAD( |
1305 | LOAD_UTF8_CHARCLASS_SPACE(), | |
1306 | !(*s == ' ' || swash_fetch(PL_utf8_space,(U8*)s, do_utf8)), | |
1307 | !isSPACE(*s) | |
1308 | ); | |
bbce6d69 | 1309 | case NSPACEL: |
4cadc6a9 YO |
1310 | REXEC_FBC_CSCAN_TAINT( |
1311 | !(*s == ' ' || isSPACE_LC_utf8((U8*)s)), | |
1312 | !isSPACE_LC(*s) | |
1313 | ); | |
a0d0e21e | 1314 | case DIGIT: |
4cadc6a9 YO |
1315 | REXEC_FBC_CSCAN_PRELOAD( |
1316 | LOAD_UTF8_CHARCLASS_DIGIT(), | |
1317 | swash_fetch(PL_utf8_digit,(U8*)s, do_utf8), | |
1318 | isDIGIT(*s) | |
1319 | ); | |
b8c5462f | 1320 | case DIGITL: |
4cadc6a9 YO |
1321 | REXEC_FBC_CSCAN_TAINT( |
1322 | isDIGIT_LC_utf8((U8*)s), | |
1323 | isDIGIT_LC(*s) | |
1324 | ); | |
a0d0e21e | 1325 | case NDIGIT: |
4cadc6a9 YO |
1326 | REXEC_FBC_CSCAN_PRELOAD( |
1327 | LOAD_UTF8_CHARCLASS_DIGIT(), | |
1328 | !swash_fetch(PL_utf8_digit,(U8*)s, do_utf8), | |
1329 | !isDIGIT(*s) | |
1330 | ); | |
b8c5462f | 1331 | case NDIGITL: |
4cadc6a9 YO |
1332 | REXEC_FBC_CSCAN_TAINT( |
1333 | !isDIGIT_LC_utf8((U8*)s), | |
1334 | !isDIGIT_LC(*s) | |
1335 | ); | |
07be1b83 YO |
1336 | case TRIE: |
1337 | /*Perl_croak(aTHX_ "panic: unknown regstclass TRIE");*/ | |
1338 | { | |
1339 | const enum { trie_plain, trie_utf8, trie_utf8_fold } | |
1340 | trie_type = do_utf8 ? | |
1341 | (c->flags == EXACT ? trie_utf8 : trie_utf8_fold) | |
1342 | : trie_plain; | |
1343 | /* what trie are we using right now */ | |
1344 | reg_ac_data *aho | |
1345 | = (reg_ac_data*)prog->data->data[ ARG( c ) ]; | |
1346 | reg_trie_data *trie=aho->trie; | |
1347 | ||
1348 | const char *last_start = strend - trie->minlen; | |
6148ee25 | 1349 | #ifdef DEBUGGING |
07be1b83 | 1350 | const char *real_start = s; |
6148ee25 | 1351 | #endif |
07be1b83 | 1352 | STRLEN maxlen = trie->maxlen; |
be8e71aa YO |
1353 | SV *sv_points; |
1354 | U8 **points; /* map of where we were in the input string | |
1355 | when reading a given string. For ASCII this | |
1356 | is unnecessary overhead as the relationship | |
1357 | is always 1:1, but for unicode, especially | |
1358 | case folded unicode this is not true. */ | |
f9e705e8 | 1359 | U8 foldbuf[ UTF8_MAXBYTES_CASE + 1 ]; |
07be1b83 YO |
1360 | |
1361 | GET_RE_DEBUG_FLAGS_DECL; | |
1362 | ||
be8e71aa YO |
1363 | /* We can't just allocate points here. We need to wrap it in |
1364 | * an SV so it gets freed properly if there is a croak while | |
1365 | * running the match */ | |
1366 | ENTER; | |
1367 | SAVETMPS; | |
1368 | sv_points=newSV(maxlen * sizeof(U8 *)); | |
1369 | SvCUR_set(sv_points, | |
1370 | maxlen * sizeof(U8 *)); | |
1371 | SvPOK_on(sv_points); | |
1372 | sv_2mortal(sv_points); | |
1373 | points=(U8**)SvPV_nolen(sv_points ); | |
07be1b83 YO |
1374 | |
1375 | if (trie->bitmap && trie_type != trie_utf8_fold) { | |
13a24bad | 1376 | while (s <= last_start && !TRIE_BITMAP_TEST(trie,*s) ) { |
07be1b83 YO |
1377 | s++; |
1378 | } | |
1379 | } | |
1380 | ||
1381 | while (s <= last_start) { | |
1382 | const U32 uniflags = UTF8_ALLOW_DEFAULT; | |
1383 | U8 *uc = (U8*)s; | |
1384 | U16 charid = 0; | |
1385 | U32 base = 1; | |
1386 | U32 state = 1; | |
1387 | UV uvc = 0; | |
1388 | STRLEN len = 0; | |
1389 | STRLEN foldlen = 0; | |
1390 | U8 *uscan = (U8*)NULL; | |
1391 | U8 *leftmost = NULL; | |
1392 | ||
1393 | U32 pointpos = 0; | |
1394 | ||
1395 | while ( state && uc <= (U8*)strend ) { | |
1396 | int failed=0; | |
1397 | if (aho->states[ state ].wordnum) { | |
1398 | U8 *lpos= points[ (pointpos - trie->wordlen[aho->states[ state ].wordnum-1] ) % maxlen ]; | |
1399 | if (!leftmost || lpos < leftmost) | |
1400 | leftmost= lpos; | |
1401 | if (base==0) break; | |
1402 | } | |
1403 | points[pointpos++ % maxlen]= uc; | |
4cadc6a9 YO |
1404 | REXEC_TRIE_READ_CHAR(trie_type, trie, uc, uscan, len, |
1405 | uvc, charid, foldlen, foldbuf, uniflags); | |
07be1b83 YO |
1406 | DEBUG_TRIE_EXECUTE_r( |
1407 | PerlIO_printf(Perl_debug_log, | |
1408 | "Pos: %d Charid:%3x CV:%4"UVxf" ", | |
1409 | (int)((const char*)uc - real_start), charid, uvc) | |
1410 | ); | |
1411 | uc += len; | |
1412 | ||
1413 | do { | |
6148ee25 | 1414 | #ifdef DEBUGGING |
07be1b83 | 1415 | U32 word = aho->states[ state ].wordnum; |
6148ee25 | 1416 | #endif |
07be1b83 YO |
1417 | base = aho->states[ state ].trans.base; |
1418 | ||
1419 | DEBUG_TRIE_EXECUTE_r( | |
1420 | PerlIO_printf( Perl_debug_log, | |
1421 | "%sState: %4"UVxf", Base: 0x%-4"UVxf" uvc=%"UVxf" word=%"UVxf"\n", | |
1422 | failed ? "Fail transition to " : "", | |
1423 | state, base, uvc, word) | |
1424 | ); | |
1425 | if ( base ) { | |
1426 | U32 tmp; | |
1427 | if (charid && | |
1428 | (base + charid > trie->uniquecharcount ) | |
1429 | && (base + charid - 1 - trie->uniquecharcount | |
1430 | < trie->lasttrans) | |
1431 | && trie->trans[base + charid - 1 - | |
1432 | trie->uniquecharcount].check == state | |
1433 | && (tmp=trie->trans[base + charid - 1 - | |
1434 | trie->uniquecharcount ].next)) | |
1435 | { | |
1436 | state = tmp; | |
1437 | break; | |
1438 | } | |
1439 | else { | |
1440 | failed++; | |
1441 | if ( state == 1 ) | |
1442 | break; | |
1443 | else | |
1444 | state = aho->fail[state]; | |
1445 | } | |
1446 | } | |
1447 | else { | |
1448 | /* we must be accepting here */ | |
1449 | failed++; | |
1450 | break; | |
1451 | } | |
1452 | } while(state); | |
1453 | if (failed) { | |
1454 | if (leftmost) | |
1455 | break; | |
1456 | else if (!charid && trie->bitmap && trie_type != trie_utf8_fold) { | |
13a24bad | 1457 | while ( uc <= (U8*)last_start && !TRIE_BITMAP_TEST(trie,*uc) ) { |
07be1b83 YO |
1458 | uc++; |
1459 | } | |
1460 | } | |
1461 | } | |
1462 | } | |
1463 | if ( aho->states[ state ].wordnum ) { | |
1464 | U8 *lpos = points[ (pointpos - trie->wordlen[aho->states[ state ].wordnum-1]) % maxlen ]; | |
1465 | if (!leftmost || lpos < leftmost) | |
1466 | leftmost = lpos; | |
1467 | } | |
1468 | DEBUG_TRIE_EXECUTE_r( | |
1469 | PerlIO_printf( Perl_debug_log, | |
1470 | "%sState: %4"UVxf", Base: 0x%-4"UVxf" uvc=%"UVxf"\n", | |
1471 | "All done: ", | |
1472 | state, base, uvc) | |
1473 | ); | |
1474 | if (leftmost) { | |
1475 | s = (char*)leftmost; | |
be8e71aa YO |
1476 | if (!reginfo || regtry(reginfo, s)) { |
1477 | FREETMPS; | |
1478 | LEAVE; | |
07be1b83 | 1479 | goto got_it; |
be8e71aa | 1480 | } |
07be1b83 YO |
1481 | s = HOPc(s,1); |
1482 | } else { | |
1483 | break; | |
1484 | } | |
1485 | } | |
be8e71aa YO |
1486 | FREETMPS; |
1487 | LEAVE; | |
07be1b83 YO |
1488 | } |
1489 | break; | |
b3c9acc1 | 1490 | default: |
3c3eec57 GS |
1491 | Perl_croak(aTHX_ "panic: unknown regstclass %d", (int)OP(c)); |
1492 | break; | |
d6a28714 | 1493 | } |
6eb5f6b9 JH |
1494 | return 0; |
1495 | got_it: | |
1496 | return s; | |
1497 | } | |
1498 | ||
1499 | /* | |
1500 | - regexec_flags - match a regexp against a string | |
1501 | */ | |
1502 | I32 | |
1503 | Perl_regexec_flags(pTHX_ register regexp *prog, char *stringarg, register char *strend, | |
1504 | char *strbeg, I32 minend, SV *sv, void *data, U32 flags) | |
1505 | /* strend: pointer to null at end of string */ | |
1506 | /* strbeg: real beginning of string */ | |
1507 | /* minend: end of match must be >=minend after stringarg. */ | |
1508 | /* data: May be used for some additional optimizations. */ | |
1509 | /* nosave: For optimizations. */ | |
1510 | { | |
97aff369 | 1511 | dVAR; |
6eb5f6b9 JH |
1512 | register char *s; |
1513 | register regnode *c; | |
1514 | register char *startpos = stringarg; | |
6eb5f6b9 JH |
1515 | I32 minlen; /* must match at least this many chars */ |
1516 | I32 dontbother = 0; /* how many characters not to try at end */ | |
6eb5f6b9 JH |
1517 | I32 end_shift = 0; /* Same for the end. */ /* CC */ |
1518 | I32 scream_pos = -1; /* Internal iterator of scream. */ | |
ccac19ea | 1519 | char *scream_olds = NULL; |
3dab1dad | 1520 | SV* const oreplsv = GvSV(PL_replgv); |
1df70142 | 1521 | const bool do_utf8 = DO_UTF8(sv); |
2757e526 | 1522 | I32 multiline; |
0df25f3d | 1523 | |
3b0527fe | 1524 | regmatch_info reginfo; /* create some info to pass to regtry etc */ |
a3621e74 YO |
1525 | |
1526 | GET_RE_DEBUG_FLAGS_DECL; | |
1527 | ||
9d4ba2ae | 1528 | PERL_UNUSED_ARG(data); |
6eb5f6b9 JH |
1529 | |
1530 | /* Be paranoid... */ | |
1531 | if (prog == NULL || startpos == NULL) { | |
1532 | Perl_croak(aTHX_ "NULL regexp parameter"); | |
1533 | return 0; | |
1534 | } | |
1535 | ||
2757e526 | 1536 | multiline = prog->reganch & PMf_MULTILINE; |
3b0527fe | 1537 | reginfo.prog = prog; |
2757e526 | 1538 | |
bac06658 JH |
1539 | RX_MATCH_UTF8_set(prog, do_utf8); |
1540 | ||
6eb5f6b9 | 1541 | minlen = prog->minlen; |
61a36c01 | 1542 | if (strend - startpos < minlen) { |
a3621e74 | 1543 | DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, |
a72c7584 JH |
1544 | "String too short [regexec_flags]...\n")); |
1545 | goto phooey; | |
1aa99e6b | 1546 | } |
6eb5f6b9 | 1547 | |
6eb5f6b9 JH |
1548 | /* Check validity of program. */ |
1549 | if (UCHARAT(prog->program) != REG_MAGIC) { | |
1550 | Perl_croak(aTHX_ "corrupted regexp program"); | |
1551 | } | |
1552 | ||
1553 | PL_reg_flags = 0; | |
1554 | PL_reg_eval_set = 0; | |
1555 | PL_reg_maxiter = 0; | |
1556 | ||
1557 | if (prog->reganch & ROPT_UTF8) | |
1558 | PL_reg_flags |= RF_utf8; | |
1559 | ||
1560 | /* Mark beginning of line for ^ and lookbehind. */ | |
3b0527fe | 1561 | reginfo.bol = startpos; /* XXX not used ??? */ |
6eb5f6b9 | 1562 | PL_bostr = strbeg; |
3b0527fe | 1563 | reginfo.sv = sv; |
6eb5f6b9 JH |
1564 | |
1565 | /* Mark end of line for $ (and such) */ | |
1566 | PL_regeol = strend; | |
1567 | ||
1568 | /* see how far we have to get to not match where we matched before */ | |
3b0527fe | 1569 | reginfo.till = startpos+minend; |
6eb5f6b9 | 1570 | |
6eb5f6b9 JH |
1571 | /* If there is a "must appear" string, look for it. */ |
1572 | s = startpos; | |
1573 | ||
3b0527fe | 1574 | if (prog->reganch & ROPT_GPOS_SEEN) { /* Need to set reginfo->ganch */ |
6eb5f6b9 JH |
1575 | MAGIC *mg; |
1576 | ||
1577 | if (flags & REXEC_IGNOREPOS) /* Means: check only at start */ | |
3b0527fe | 1578 | reginfo.ganch = startpos; |
6eb5f6b9 JH |
1579 | else if (sv && SvTYPE(sv) >= SVt_PVMG |
1580 | && SvMAGIC(sv) | |
14befaf4 DM |
1581 | && (mg = mg_find(sv, PERL_MAGIC_regex_global)) |
1582 | && mg->mg_len >= 0) { | |
3b0527fe | 1583 | reginfo.ganch = strbeg + mg->mg_len; /* Defined pos() */ |
6eb5f6b9 | 1584 | if (prog->reganch & ROPT_ANCH_GPOS) { |
3b0527fe | 1585 | if (s > reginfo.ganch) |
6eb5f6b9 | 1586 | goto phooey; |
3b0527fe | 1587 | s = reginfo.ganch; |
6eb5f6b9 JH |
1588 | } |
1589 | } | |
1590 | else /* pos() not defined */ | |
3b0527fe | 1591 | reginfo.ganch = strbeg; |
6eb5f6b9 JH |
1592 | } |
1593 | ||
a0714e2c | 1594 | if (!(flags & REXEC_CHECKED) && (prog->check_substr != NULL || prog->check_utf8 != NULL)) { |
6eb5f6b9 JH |
1595 | re_scream_pos_data d; |
1596 | ||
1597 | d.scream_olds = &scream_olds; | |
1598 | d.scream_pos = &scream_pos; | |
1599 | s = re_intuit_start(prog, sv, s, strend, flags, &d); | |
3fa9c3d7 | 1600 | if (!s) { |
a3621e74 | 1601 | DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "Not present...\n")); |
6eb5f6b9 | 1602 | goto phooey; /* not present */ |
3fa9c3d7 | 1603 | } |
6eb5f6b9 JH |
1604 | } |
1605 | ||
a3621e74 | 1606 | DEBUG_EXECUTE_r({ |
0df25f3d YO |
1607 | RE_PV_DISPLAY_DECL(s0, len0, UTF, |
1608 | PERL_DEBUG_PAD_ZERO(0), prog->precomp, prog->prelen, 60); | |
1609 | RE_PV_DISPLAY_DECL(s1, len1, do_utf8, | |
1610 | PERL_DEBUG_PAD_ZERO(1), startpos, strend - startpos, 60); | |
1611 | ||
2a782b5b JH |
1612 | if (!PL_colorset) |
1613 | reginitcolors(); | |
1614 | PerlIO_printf(Perl_debug_log, | |
a0288114 | 1615 | "%sMatching REx%s \"%s%*.*s%s%s\" against \"%s%.*s%s%s\"\n", |
e4584336 | 1616 | PL_colors[4], PL_colors[5], PL_colors[0], |
9e55ce06 | 1617 | len0, len0, s0, |
2a782b5b | 1618 | PL_colors[1], |
9e55ce06 | 1619 | len0 > 60 ? "..." : "", |
2a782b5b | 1620 | PL_colors[0], |
9e55ce06 JH |
1621 | (int)(len1 > 60 ? 60 : len1), |
1622 | s1, PL_colors[1], | |
1623 | (len1 > 60 ? "..." : "") | |
2a782b5b JH |
1624 | ); |
1625 | }); | |
6eb5f6b9 JH |
1626 | |
1627 | /* Simplest case: anchored match need be tried only once. */ | |
1628 | /* [unless only anchor is BOL and multiline is set] */ | |
1629 | if (prog->reganch & (ROPT_ANCH & ~ROPT_ANCH_GPOS)) { | |
3b0527fe | 1630 | if (s == startpos && regtry(®info, startpos)) |
6eb5f6b9 | 1631 | goto got_it; |
7fba1cd6 | 1632 | else if (multiline || (prog->reganch & ROPT_IMPLICIT) |
6eb5f6b9 JH |
1633 | || (prog->reganch & ROPT_ANCH_MBOL)) /* XXXX SBOL? */ |
1634 | { | |
1635 | char *end; | |
1636 | ||
1637 | if (minlen) | |
1638 | dontbother = minlen - 1; | |
1aa99e6b | 1639 | end = HOP3c(strend, -dontbother, strbeg) - 1; |
6eb5f6b9 | 1640 | /* for multiline we only have to try after newlines */ |
33b8afdf | 1641 | if (prog->check_substr || prog->check_utf8) { |
6eb5f6b9 JH |
1642 | if (s == startpos) |
1643 | goto after_try; | |
1644 | while (1) { | |
3b0527fe | 1645 | if (regtry(®info, s)) |
6eb5f6b9 JH |
1646 | goto got_it; |
1647 | after_try: | |
1648 | if (s >= end) | |
1649 | goto phooey; | |
1650 | if (prog->reganch & RE_USE_INTUIT) { | |
1651 | s = re_intuit_start(prog, sv, s + 1, strend, flags, NULL); | |
1652 | if (!s) | |
1653 | goto phooey; | |
1654 | } | |
1655 | else | |
1656 | s++; | |
1657 | } | |
1658 | } else { | |
1659 | if (s > startpos) | |
1660 | s--; | |
1661 | while (s < end) { | |
1662 | if (*s++ == '\n') { /* don't need PL_utf8skip here */ | |
3b0527fe | 1663 | if (regtry(®info, s)) |
6eb5f6b9 JH |
1664 | goto got_it; |
1665 | } | |
1666 | } | |
1667 | } | |
1668 | } | |
1669 | goto phooey; | |
1670 | } else if (prog->reganch & ROPT_ANCH_GPOS) { | |
3b0527fe | 1671 | if (regtry(®info, reginfo.ganch)) |
6eb5f6b9 JH |
1672 | goto got_it; |
1673 | goto phooey; | |
1674 | } | |
1675 | ||
1676 | /* Messy cases: unanchored match. */ | |
33b8afdf | 1677 | if ((prog->anchored_substr || prog->anchored_utf8) && prog->reganch & ROPT_SKIP) { |
6eb5f6b9 JH |
1678 | /* we have /x+whatever/ */ |
1679 | /* it must be a one character string (XXXX Except UTF?) */ | |
33b8afdf | 1680 | char ch; |
bf93d4cc GS |
1681 | #ifdef DEBUGGING |
1682 | int did_match = 0; | |
1683 | #endif | |
33b8afdf JH |
1684 | if (!(do_utf8 ? prog->anchored_utf8 : prog->anchored_substr)) |
1685 | do_utf8 ? to_utf8_substr(prog) : to_byte_substr(prog); | |
3f7c398e | 1686 | ch = SvPVX_const(do_utf8 ? prog->anchored_utf8 : prog->anchored_substr)[0]; |
bf93d4cc | 1687 | |
1aa99e6b | 1688 | if (do_utf8) { |
4cadc6a9 | 1689 | REXEC_FBC_SCAN( |
6eb5f6b9 | 1690 | if (*s == ch) { |
a3621e74 | 1691 | DEBUG_EXECUTE_r( did_match = 1 ); |
3b0527fe | 1692 | if (regtry(®info, s)) goto got_it; |
6eb5f6b9 JH |
1693 | s += UTF8SKIP(s); |
1694 | while (s < strend && *s == ch) | |
1695 | s += UTF8SKIP(s); | |
1696 | } | |
4cadc6a9 | 1697 | ); |
6eb5f6b9 JH |
1698 | } |
1699 | else { | |
4cadc6a9 | 1700 | REXEC_FBC_SCAN( |
6eb5f6b9 | 1701 | if (*s == ch) { |
a3621e74 | 1702 | DEBUG_EXECUTE_r( did_match = 1 ); |
3b0527fe | 1703 | if (regtry(®info, s)) goto got_it; |
6eb5f6b9 JH |
1704 | s++; |
1705 | while (s < strend && *s == ch) | |
1706 | s++; | |
1707 | } | |
4cadc6a9 | 1708 | ); |
6eb5f6b9 | 1709 | } |
a3621e74 | 1710 | DEBUG_EXECUTE_r(if (!did_match) |
bf93d4cc | 1711 | PerlIO_printf(Perl_debug_log, |
b7953727 JH |
1712 | "Did not find anchored character...\n") |
1713 | ); | |
6eb5f6b9 | 1714 | } |
a0714e2c SS |
1715 | else if (prog->anchored_substr != NULL |
1716 | || prog->anchored_utf8 != NULL | |
1717 | || ((prog->float_substr != NULL || prog->float_utf8 != NULL) | |
33b8afdf JH |
1718 | && prog->float_max_offset < strend - s)) { |
1719 | SV *must; | |
1720 | I32 back_max; | |
1721 | I32 back_min; | |
1722 | char *last; | |
6eb5f6b9 | 1723 | char *last1; /* Last position checked before */ |
bf93d4cc GS |
1724 | #ifdef DEBUGGING |
1725 | int did_match = 0; | |
1726 | #endif | |
33b8afdf JH |
1727 | if (prog->anchored_substr || prog->anchored_utf8) { |
1728 | if (!(do_utf8 ? prog->anchored_utf8 : prog->anchored_substr)) | |
1729 | do_utf8 ? to_utf8_substr(prog) : to_byte_substr(prog); | |
1730 | must = do_utf8 ? prog->anchored_utf8 : prog->anchored_substr; | |
1731 | back_max = back_min = prog->anchored_offset; | |
1732 | } else { | |
1733 | if (!(do_utf8 ? prog->float_utf8 : prog->float_substr)) | |
1734 | do_utf8 ? to_utf8_substr(prog) : to_byte_substr(prog); | |
1735 | must = do_utf8 ? prog->float_utf8 : prog->float_substr; | |
1736 | back_max = prog->float_max_offset; | |
1737 | back_min = prog->float_min_offset; | |
1738 | } | |
1739 | if (must == &PL_sv_undef) | |
1740 | /* could not downgrade utf8 check substring, so must fail */ | |
1741 | goto phooey; | |
1742 | ||
1743 | last = HOP3c(strend, /* Cannot start after this */ | |
1744 | -(I32)(CHR_SVLEN(must) | |
1745 | - (SvTAIL(must) != 0) + back_min), strbeg); | |
6eb5f6b9 JH |
1746 | |
1747 | if (s > PL_bostr) | |
1748 | last1 = HOPc(s, -1); | |
1749 | else | |
1750 | last1 = s - 1; /* bogus */ | |
1751 | ||
a0288114 | 1752 | /* XXXX check_substr already used to find "s", can optimize if |
6eb5f6b9 JH |
1753 | check_substr==must. */ |
1754 | scream_pos = -1; | |
1755 | dontbother = end_shift; | |
1756 | strend = HOPc(strend, -dontbother); | |
1757 | while ( (s <= last) && | |
9041c2e3 | 1758 | ((flags & REXEC_SCREAM) |
1aa99e6b | 1759 | ? (s = screaminstr(sv, must, HOP3c(s, back_min, strend) - strbeg, |
6eb5f6b9 | 1760 | end_shift, &scream_pos, 0)) |
1aa99e6b | 1761 | : (s = fbm_instr((unsigned char*)HOP3(s, back_min, strend), |
9041c2e3 | 1762 | (unsigned char*)strend, must, |
7fba1cd6 | 1763 | multiline ? FBMrf_MULTILINE : 0))) ) { |
4addbd3b HS |
1764 | /* we may be pointing at the wrong string */ |
1765 | if ((flags & REXEC_SCREAM) && RX_MATCH_COPIED(prog)) | |
3f7c398e | 1766 | s = strbeg + (s - SvPVX_const(sv)); |
a3621e74 | 1767 | DEBUG_EXECUTE_r( did_match = 1 ); |
6eb5f6b9 JH |
1768 | if (HOPc(s, -back_max) > last1) { |
1769 | last1 = HOPc(s, -back_min); | |
1770 | s = HOPc(s, -back_max); | |
1771 | } | |
1772 | else { | |
52657f30 | 1773 | char * const t = (last1 >= PL_bostr) ? HOPc(last1, 1) : last1 + 1; |
6eb5f6b9 JH |
1774 | |
1775 | last1 = HOPc(s, -back_min); | |
52657f30 | 1776 | s = t; |
6eb5f6b9 | 1777 | } |
1aa99e6b | 1778 | if (do_utf8) { |
6eb5f6b9 | 1779 | while (s <= last1) { |
3b0527fe | 1780 | if (regtry(®info, s)) |
6eb5f6b9 JH |
1781 | goto got_it; |
1782 | s += UTF8SKIP(s); | |
1783 | } | |
1784 | } | |
1785 | else { | |
1786 | while (s <= last1) { | |
3b0527fe | 1787 | if (regtry(®info, s)) |
6eb5f6b9 JH |
1788 | goto got_it; |
1789 | s++; | |
1790 | } | |
1791 | } | |
1792 | } | |
a3621e74 | 1793 | DEBUG_EXECUTE_r(if (!did_match) |
b7953727 | 1794 | PerlIO_printf(Perl_debug_log, |
a0288114 | 1795 | "Did not find %s substr \"%s%.*s%s\"%s...\n", |
33b8afdf | 1796 | ((must == prog->anchored_substr || must == prog->anchored_utf8) |
bf93d4cc GS |
1797 | ? "anchored" : "floating"), |
1798 | PL_colors[0], | |
1799 | (int)(SvCUR(must) - (SvTAIL(must)!=0)), | |
3f7c398e | 1800 | SvPVX_const(must), |
b7953727 JH |
1801 | PL_colors[1], (SvTAIL(must) ? "$" : "")) |
1802 | ); | |
6eb5f6b9 JH |
1803 | goto phooey; |
1804 | } | |
155aba94 | 1805 | else if ((c = prog->regstclass)) { |
f14c76ed | 1806 | if (minlen) { |
be8e71aa | 1807 | const OPCODE op = OP(prog->regstclass); |
66e933ab | 1808 | /* don't bother with what can't match */ |
07be1b83 | 1809 | if (PL_regkind[op] != EXACT && op != CANY && op != TRIE) |
f14c76ed RGS |
1810 | strend = HOPc(strend, -(minlen - 1)); |
1811 | } | |
a3621e74 | 1812 | DEBUG_EXECUTE_r({ |
be8e71aa | 1813 | SV * const prop = sv_newmortal(); |
32fc9b6a | 1814 | regprop(prog, prop, c); |
0df25f3d YO |
1815 | { |
1816 | RE_PV_DISPLAY_DECL(s0,len0,UTF, | |
1817 | PERL_DEBUG_PAD_ZERO(0),SvPVX_const(prop),SvCUR(prop),60); | |
1818 | RE_PV_DISPLAY_DECL(s1,len1,UTF, | |
1819 | PERL_DEBUG_PAD_ZERO(1),s,strend-s,60); | |
1820 | PerlIO_printf(Perl_debug_log, | |
1821 | "Matching stclass \"%*.*s\" against \"%*.*s\" (%d chars)\n", | |
1822 | len0, len0, s0, | |
1823 | len1, len1, s1, (int)(strend - s)); | |
1824 | } | |
ffc61ed2 | 1825 | }); |
3b0527fe | 1826 | if (find_byclass(prog, c, s, strend, ®info)) |
6eb5f6b9 | 1827 | goto got_it; |
07be1b83 | 1828 | DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "Contradicts stclass... [regexec_flags]\n")); |
d6a28714 JH |
1829 | } |
1830 | else { | |
1831 | dontbother = 0; | |
a0714e2c | 1832 | if (prog->float_substr != NULL || prog->float_utf8 != NULL) { |
33b8afdf | 1833 | /* Trim the end. */ |
d6a28714 | 1834 | char *last; |
33b8afdf JH |
1835 | SV* float_real; |
1836 | ||
1837 | if (!(do_utf8 ? prog->float_utf8 : prog->float_substr)) | |
1838 | do_utf8 ? to_utf8_substr(prog) : to_byte_substr(prog); | |
1839 | float_real = do_utf8 ? prog->float_utf8 : prog->float_substr; | |
d6a28714 JH |
1840 | |
1841 | if (flags & REXEC_SCREAM) { | |
33b8afdf | 1842 | last = screaminstr(sv, float_real, s - strbeg, |
d6a28714 JH |
1843 | end_shift, &scream_pos, 1); /* last one */ |
1844 | if (!last) | |
ffc61ed2 | 1845 | last = scream_olds; /* Only one occurrence. */ |
4addbd3b HS |
1846 | /* we may be pointing at the wrong string */ |
1847 | else if (RX_MATCH_COPIED(prog)) | |
3f7c398e | 1848 | s = strbeg + (s - SvPVX_const(sv)); |
b8c5462f | 1849 | } |
d6a28714 JH |
1850 | else { |
1851 | STRLEN len; | |
cfd0369c | 1852 | const char * const little = SvPV_const(float_real, len); |
d6a28714 | 1853 | |
33b8afdf | 1854 | if (SvTAIL(float_real)) { |
d6a28714 JH |
1855 | if (memEQ(strend - len + 1, little, len - 1)) |
1856 | last = strend - len + 1; | |
7fba1cd6 | 1857 | else if (!multiline) |
9041c2e3 | 1858 | last = memEQ(strend - len, little, len) |
bd61b366 | 1859 | ? strend - len : NULL; |
b8c5462f | 1860 | else |
d6a28714 JH |
1861 | goto find_last; |
1862 | } else { | |
1863 | find_last: | |
9041c2e3 | 1864 | if (len) |
d6a28714 | 1865 | last = rninstr(s, strend, little, little + len); |
b8c5462f | 1866 | else |
a0288114 | 1867 | last = strend; /* matching "$" */ |
b8c5462f | 1868 | } |
b8c5462f | 1869 | } |
bf93d4cc | 1870 | if (last == NULL) { |
a3621e74 | 1871 | DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, |
bf93d4cc | 1872 | "%sCan't trim the tail, match fails (should not happen)%s\n", |
e4584336 | 1873 | PL_colors[4], PL_colors[5])); |
bf93d4cc GS |
1874 | goto phooey; /* Should not happen! */ |
1875 | } | |
d6a28714 JH |
1876 | dontbother = strend - last + prog->float_min_offset; |
1877 | } | |
1878 | if (minlen && (dontbother < minlen)) | |
1879 | dontbother = minlen - 1; | |
1880 | strend -= dontbother; /* this one's always in bytes! */ | |
1881 | /* We don't know much -- general case. */ | |
1aa99e6b | 1882 | if (do_utf8) { |
d6a28714 | 1883 | for (;;) { |
3b0527fe | 1884 | if (regtry(®info, s)) |
d6a28714 JH |
1885 | goto got_it; |
1886 | if (s >= strend) | |
1887 | break; | |
b8c5462f | 1888 | s += UTF8SKIP(s); |
d6a28714 JH |
1889 | }; |
1890 | } | |
1891 | else { | |
1892 | do { | |
3b0527fe | 1893 | if (regtry(®info, s)) |
d6a28714 JH |
1894 | goto got_it; |
1895 | } while (s++ < strend); | |
1896 | } | |
1897 | } | |
1898 | ||
1899 | /* Failure. */ | |
1900 | goto phooey; | |
1901 | ||
1902 | got_it: | |
1903 | RX_MATCH_TAINTED_set(prog, PL_reg_flags & RF_tainted); | |
1904 | ||
1905 | if (PL_reg_eval_set) { | |
1906 | /* Preserve the current value of $^R */ | |
1907 | if (oreplsv != GvSV(PL_replgv)) | |
1908 | sv_setsv(oreplsv, GvSV(PL_replgv));/* So that when GvSV(replgv) is | |
1909 | restored, the value remains | |
1910 | the same. */ | |
4f639d21 | 1911 | restore_pos(aTHX_ prog); |
d6a28714 JH |
1912 | } |
1913 | ||
1914 | /* make sure $`, $&, $', and $digit will work later */ | |
1915 | if ( !(flags & REXEC_NOT_FIRST) ) { | |
ed252734 | 1916 | RX_MATCH_COPY_FREE(prog); |
d6a28714 | 1917 | if (flags & REXEC_COPY_STR) { |
be8e71aa | 1918 | const I32 i = PL_regeol - startpos + (stringarg - strbeg); |
f8c7b90f | 1919 | #ifdef PERL_OLD_COPY_ON_WRITE |
ed252734 NC |
1920 | if ((SvIsCOW(sv) |
1921 | || (SvFLAGS(sv) & CAN_COW_MASK) == CAN_COW_FLAGS)) { | |
1922 | if (DEBUG_C_TEST) { | |
1923 | PerlIO_printf(Perl_debug_log, | |
1924 | "Copy on write: regexp capture, type %d\n", | |
1925 | (int) SvTYPE(sv)); | |
1926 | } | |
1927 | prog->saved_copy = sv_setsv_cow(prog->saved_copy, sv); | |
d5263905 | 1928 | prog->subbeg = (char *)SvPVX_const(prog->saved_copy); |
ed252734 NC |
1929 | assert (SvPOKp(prog->saved_copy)); |
1930 | } else | |
1931 | #endif | |
1932 | { | |
1933 | RX_MATCH_COPIED_on(prog); | |
1934 | s = savepvn(strbeg, i); | |
1935 | prog->subbeg = s; | |
1936 | } | |
d6a28714 | 1937 | prog->sublen = i; |
d6a28714 JH |
1938 | } |
1939 | else { | |
1940 | prog->subbeg = strbeg; | |
1941 | prog->sublen = PL_regeol - strbeg; /* strend may have been modified */ | |
1942 | } | |
1943 | } | |
9041c2e3 | 1944 | |
d6a28714 JH |
1945 | return 1; |
1946 | ||
1947 | phooey: | |
a3621e74 | 1948 | DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "%sMatch failed%s\n", |
e4584336 | 1949 | PL_colors[4], PL_colors[5])); |
d6a28714 | 1950 | if (PL_reg_eval_set) |
4f639d21 | 1951 | restore_pos(aTHX_ prog); |
d6a28714 JH |
1952 | return 0; |
1953 | } | |
1954 | ||
1955 | /* | |
1956 | - regtry - try match at specific point | |
1957 | */ | |
1958 | STATIC I32 /* 0 failure, 1 success */ | |
3b0527fe | 1959 | S_regtry(pTHX_ const regmatch_info *reginfo, char *startpos) |
d6a28714 | 1960 | { |
97aff369 | 1961 | dVAR; |
d6a28714 JH |
1962 | register I32 *sp; |
1963 | register I32 *ep; | |
1964 | CHECKPOINT lastcp; | |
3b0527fe | 1965 | regexp *prog = reginfo->prog; |
a3621e74 | 1966 | GET_RE_DEBUG_FLAGS_DECL; |
d6a28714 | 1967 | |
02db2b7b IZ |
1968 | #ifdef DEBUGGING |
1969 | PL_regindent = 0; /* XXXX Not good when matches are reenterable... */ | |
1970 | #endif | |
d6a28714 JH |
1971 | if ((prog->reganch & ROPT_EVAL_SEEN) && !PL_reg_eval_set) { |
1972 | MAGIC *mg; | |
1973 | ||
1974 | PL_reg_eval_set = RS_init; | |
a3621e74 | 1975 | DEBUG_EXECUTE_r(DEBUG_s( |
b900a521 JH |
1976 | PerlIO_printf(Perl_debug_log, " setting stack tmpbase at %"IVdf"\n", |
1977 | (IV)(PL_stack_sp - PL_stack_base)); | |
d6a28714 | 1978 | )); |
e8347627 | 1979 | SAVEI32(cxstack[cxstack_ix].blk_oldsp); |
d6a28714 JH |
1980 | cxstack[cxstack_ix].blk_oldsp = PL_stack_sp - PL_stack_base; |
1981 | /* Otherwise OP_NEXTSTATE will free whatever on stack now. */ | |
1982 | SAVETMPS; | |
1983 | /* Apparently this is not needed, judging by wantarray. */ | |
e8347627 | 1984 | /* SAVEI8(cxstack[cxstack_ix].blk_gimme); |
d6a28714 JH |
1985 | cxstack[cxstack_ix].blk_gimme = G_SCALAR; */ |
1986 | ||
3b0527fe | 1987 | if (reginfo->sv) { |
d6a28714 | 1988 | /* Make $_ available to executed code. */ |
3b0527fe | 1989 | if (reginfo->sv != DEFSV) { |
59f00321 | 1990 | SAVE_DEFSV; |
3b0527fe | 1991 | DEFSV = reginfo->sv; |
b8c5462f | 1992 | } |
d6a28714 | 1993 | |
3b0527fe DM |
1994 | if (!(SvTYPE(reginfo->sv) >= SVt_PVMG && SvMAGIC(reginfo->sv) |
1995 | && (mg = mg_find(reginfo->sv, PERL_MAGIC_regex_global)))) { | |
d6a28714 | 1996 | /* prepare for quick setting of pos */ |
d300d9fa NC |
1997 | #ifdef PERL_OLD_COPY_ON_WRITE |
1998 | if (SvIsCOW(sv)) | |
1999 | sv_force_normal_flags(sv, 0); | |
2000 | #endif | |
3dab1dad | 2001 | mg = sv_magicext(reginfo->sv, NULL, PERL_MAGIC_regex_global, |
d300d9fa | 2002 | &PL_vtbl_mglob, NULL, 0); |
d6a28714 | 2003 | mg->mg_len = -1; |
b8c5462f | 2004 | } |
d6a28714 JH |
2005 | PL_reg_magic = mg; |
2006 | PL_reg_oldpos = mg->mg_len; | |
4f639d21 | 2007 | SAVEDESTRUCTOR_X(restore_pos, prog); |
d6a28714 | 2008 | } |
09687e5a | 2009 | if (!PL_reg_curpm) { |
a02a5408 | 2010 | Newxz(PL_reg_curpm, 1, PMOP); |
09687e5a AB |
2011 | #ifdef USE_ITHREADS |
2012 | { | |
be8e71aa | 2013 | SV* const repointer = newSViv(0); |
577e12cc | 2014 | /* so we know which PL_regex_padav element is PL_reg_curpm */ |
35061a7e | 2015 | SvFLAGS(repointer) |= SVf_BREAK; |
09687e5a AB |
2016 | av_push(PL_regex_padav,repointer); |
2017 | PL_reg_curpm->op_pmoffset = av_len(PL_regex_padav); | |
2018 | PL_regex_pad = AvARRAY(PL_regex_padav); | |
2019 | } | |
2020 | #endif | |
2021 | } | |
aaa362c4 | 2022 | PM_SETRE(PL_reg_curpm, prog); |
d6a28714 JH |
2023 | PL_reg_oldcurpm = PL_curpm; |
2024 | PL_curpm = PL_reg_curpm; | |
2025 | if (RX_MATCH_COPIED(prog)) { | |
2026 | /* Here is a serious problem: we cannot rewrite subbeg, | |
2027 | since it may be needed if this match fails. Thus | |
2028 | $` inside (?{}) could fail... */ | |
2029 | PL_reg_oldsaved = prog->subbeg; | |
2030 | PL_reg_oldsavedlen = prog->sublen; | |
f8c7b90f | 2031 | #ifdef PERL_OLD_COPY_ON_WRITE |
ed252734 NC |
2032 | PL_nrs = prog->saved_copy; |
2033 | #endif | |
d6a28714 JH |
2034 | RX_MATCH_COPIED_off(prog); |
2035 | } | |
2036 | else | |
bd61b366 | 2037 | PL_reg_oldsaved = NULL; |
d6a28714 JH |
2038 | prog->subbeg = PL_bostr; |
2039 | prog->sublen = PL_regeol - PL_bostr; /* strend may have been modified */ | |
2040 | } | |
973dddac | 2041 | prog->startp[0] = startpos - PL_bostr; |
d6a28714 JH |
2042 | PL_reginput = startpos; |
2043 | PL_regstartp = prog->startp; | |
2044 | PL_regendp = prog->endp; | |
2045 | PL_reglastparen = &prog->lastparen; | |
a01268b5 | 2046 | PL_reglastcloseparen = &prog->lastcloseparen; |
d6a28714 | 2047 | prog->lastparen = 0; |
03994de8 | 2048 | prog->lastcloseparen = 0; |
d6a28714 | 2049 | PL_regsize = 0; |
a3621e74 | 2050 | DEBUG_EXECUTE_r(PL_reg_starttry = startpos); |
d6a28714 JH |
2051 | if (PL_reg_start_tmpl <= prog->nparens) { |
2052 | PL_reg_start_tmpl = prog->nparens*3/2 + 3; | |
2053 | if(PL_reg_start_tmp) | |
2054 | Renew(PL_reg_start_tmp, PL_reg_start_tmpl, char*); | |
2055 | else | |
a02a5408 | 2056 | Newx(PL_reg_start_tmp, PL_reg_start_tmpl, char*); |
d6a28714 JH |
2057 | } |
2058 | ||
2059 | /* XXXX What this code is doing here?!!! There should be no need | |
2060 | to do this again and again, PL_reglastparen should take care of | |
3dd2943c | 2061 | this! --ilya*/ |
dafc8851 JH |
2062 | |
2063 | /* Tests pat.t#187 and split.t#{13,14} seem to depend on this code. | |
2064 | * Actually, the code in regcppop() (which Ilya may be meaning by | |
daf18116 JH |
2065 | * PL_reglastparen), is not needed at all by the test suite |
2066 | * (op/regexp, op/pat, op/split), but that code is needed, oddly | |
2067 | * enough, for building DynaLoader, or otherwise this | |
2068 | * "Error: '*' not in typemap in DynaLoader.xs, line 164" | |
2069 | * will happen. Meanwhile, this code *is* needed for the | |
2070 | * above-mentioned test suite tests to succeed. The common theme | |
2071 | * on those tests seems to be returning null fields from matches. | |
2072 | * --jhi */ | |
dafc8851 | 2073 | #if 1 |
d6a28714 JH |
2074 | sp = prog->startp; |
2075 | ep = prog->endp; | |
2076 | if (prog->nparens) { | |
097eb12c | 2077 | register I32 i; |
eb160463 | 2078 | for (i = prog->nparens; i > (I32)*PL_reglastparen; i--) { |
d6a28714 JH |
2079 | *++sp = -1; |
2080 | *++ep = -1; | |
2081 | } | |
2082 | } | |
dafc8851 | 2083 | #endif |
02db2b7b | 2084 | REGCP_SET(lastcp); |
3b0527fe | 2085 | if (regmatch(reginfo, prog->program + 1)) { |
d6a28714 JH |
2086 | prog->endp[0] = PL_reginput - PL_bostr; |
2087 | return 1; | |
2088 | } | |
02db2b7b | 2089 | REGCP_UNWIND(lastcp); |
d6a28714 JH |
2090 | return 0; |
2091 | } | |
2092 | ||
02db2b7b | 2093 | |
8ba1375e MJD |
2094 | #define sayYES goto yes |
2095 | #define sayNO goto no | |
e0f9d4a8 | 2096 | #define sayNO_ANYOF goto no_anyof |
8ba1375e | 2097 | #define sayYES_FINAL goto yes_final |
8ba1375e MJD |
2098 | #define sayNO_FINAL goto no_final |
2099 | #define sayNO_SILENT goto do_no | |
2100 | #define saySAME(x) if (x) goto yes; else goto no | |
2101 | ||
3ab3c9b4 | 2102 | #define CACHEsayNO STMT_START { \ |
3298f257 DM |
2103 | if (st->u.whilem.cache_offset | st->u.whilem.cache_bit) \ |
2104 | PL_reg_poscache[st->u.whilem.cache_offset] |= \ | |
2105 | (1<<st->u.whilem.cache_bit); \ | |
3ab3c9b4 HS |
2106 | sayNO; \ |
2107 | } STMT_END | |
2108 | ||
3298f257 | 2109 | |
a3621e74 YO |
2110 | /* this is used to determine how far from the left messages like |
2111 | 'failed...' are printed. Currently 29 makes these messages line | |
2112 | up with the opcode they refer to. Earlier perls used 25 which | |
2113 | left these messages outdented making reviewing a debug output | |
2114 | quite difficult. | |
2115 | */ | |
2116 | #define REPORT_CODE_OFF 29 | |
2117 | ||
2118 | ||
2119 | /* Make sure there is a test for this +1 options in re_tests */ | |
2120 | #define TRIE_INITAL_ACCEPT_BUFFLEN 4; | |
2121 | ||
40a82448 DM |
2122 | #define CHRTEST_UNINIT -1001 /* c1/c2 haven't been calculated yet */ |
2123 | #define CHRTEST_VOID -1000 /* the c1/c2 "next char" test should be skipped */ | |
9e137952 | 2124 | |
86545054 DM |
2125 | #define SLAB_FIRST(s) (&(s)->states[0]) |
2126 | #define SLAB_LAST(s) (&(s)->states[PERL_REGMATCH_SLAB_SLOTS-1]) | |
2127 | ||
5d9a96ca DM |
2128 | /* grab a new slab and return the first slot in it */ |
2129 | ||
2130 | STATIC regmatch_state * | |
2131 | S_push_slab(pTHX) | |
2132 | { | |
54df2634 NC |
2133 | #if PERL_VERSION < 9 |
2134 | dMY_CXT; | |
2135 | #endif | |
5d9a96ca DM |
2136 | regmatch_slab *s = PL_regmatch_slab->next; |
2137 | if (!s) { | |
2138 | Newx(s, 1, regmatch_slab); | |
2139 | s->prev = PL_regmatch_slab; | |
2140 | s->next = NULL; | |
2141 | PL_regmatch_slab->next = s; | |
2142 | } | |
2143 | PL_regmatch_slab = s; | |
86545054 | 2144 | return SLAB_FIRST(s); |
5d9a96ca | 2145 | } |
5b47454d | 2146 | |
95b24440 DM |
2147 | /* simulate a recursive call to regmatch */ |
2148 | ||
2149 | #define REGMATCH(ns, where) \ | |
5d9a96ca DM |
2150 | st->scan = scan; \ |
2151 | scan = (ns); \ | |
2152 | st->resume_state = resume_##where; \ | |
95b24440 DM |
2153 | goto start_recurse; \ |
2154 | resume_point_##where: | |
2155 | ||
40a82448 DM |
2156 | /* push a new state then goto it */ |
2157 | ||
2158 | #define PUSH_STATE_GOTO(state, node) \ | |
2159 | scan = node; \ | |
2160 | st->resume_state = state; \ | |
2161 | goto push_state; | |
2162 | ||
2163 | /* push a new state with success backtracking, then goto it */ | |
2164 | ||
2165 | #define PUSH_YES_STATE_GOTO(state, node) \ | |
2166 | scan = node; \ | |
2167 | st->resume_state = state; \ | |
2168 | goto push_yes_state; | |
2169 | ||
aa283a38 | 2170 | |
aa283a38 | 2171 | |
d6a28714 JH |
2172 | /* |
2173 | - regmatch - main matching routine | |
2174 | * | |
2175 | * Conceptually the strategy is simple: check to see whether the current | |
2176 | * node matches, call self recursively to see whether the rest matches, | |
2177 | * and then act accordingly. In practice we make some effort to avoid | |
2178 | * recursion, in particular by going through "ordinary" nodes (that don't | |
2179 | * need to know whether the rest of the match failed) by a loop instead of | |
2180 | * by recursion. | |
2181 | */ | |
2182 | /* [lwall] I've hoisted the register declarations to the outer block in order to | |
2183 | * maybe save a little bit of pushing and popping on the stack. It also takes | |
2184 | * advantage of machines that use a register save mask on subroutine entry. | |
95b24440 DM |
2185 | * |
2186 | * This function used to be heavily recursive, but since this had the | |
2187 | * effect of blowing the CPU stack on complex regexes, it has been | |
2188 | * restructured to be iterative, and to save state onto the heap rather | |
2189 | * than the stack. Essentially whereever regmatch() used to be called, it | |
2190 | * pushes the current state, notes where to return, then jumps back into | |
2191 | * the main loop. | |
2192 | * | |
2193 | * Originally the structure of this function used to look something like | |
2194 | ||
2195 | S_regmatch() { | |
2196 | int a = 1, b = 2; | |
2197 | ... | |
2198 | while (scan != NULL) { | |
5d9a96ca | 2199 | a++; // do stuff with a and b |
95b24440 DM |
2200 | ... |
2201 | switch (OP(scan)) { | |
2202 | case FOO: { | |
2203 | int local = 3; | |
2204 | ... | |
2205 | if (regmatch(...)) // recurse | |
2206 | goto yes; | |
2207 | } | |
2208 | ... | |
2209 | } | |
2210 | } | |
2211 | yes: | |
2212 | return 1; | |
2213 | } | |
2214 | ||
2215 | * Now it looks something like this: | |
2216 | ||
5d9a96ca | 2217 | typedef struct { |
95b24440 DM |
2218 | int a, b, local; |
2219 | int resume_state; | |
5d9a96ca | 2220 | } regmatch_state; |
95b24440 DM |
2221 | |
2222 | S_regmatch() { | |
5d9a96ca DM |
2223 | regmatch_state *st = new(); |
2224 | int depth=0; | |
2225 | st->a++; // do stuff with a and b | |
95b24440 DM |
2226 | ... |
2227 | while (scan != NULL) { | |
2228 | ... | |
2229 | switch (OP(scan)) { | |
2230 | case FOO: { | |
5d9a96ca | 2231 | st->local = 3; |
95b24440 | 2232 | ... |
5d9a96ca DM |
2233 | st->scan = scan; |
2234 | scan = ...; | |
2235 | st->resume_state = resume_FOO; | |
2236 | goto start_recurse; // recurse | |
95b24440 | 2237 | |
5d9a96ca DM |
2238 | resume_point_FOO: |
2239 | if (result) | |
95b24440 DM |
2240 | goto yes; |
2241 | } | |
2242 | ... | |
2243 | } | |
5d9a96ca DM |
2244 | start_recurse: |
2245 | st = new(); push a new state | |
2246 | st->a = 1; st->b = 2; | |
2247 | depth++; | |
95b24440 | 2248 | } |
5d9a96ca | 2249 | yes: |
95b24440 | 2250 | result = 1; |
5d9a96ca DM |
2251 | if (depth--) { |
2252 | st = pop(); | |
95b24440 DM |
2253 | switch (resume_state) { |
2254 | case resume_FOO: | |
2255 | goto resume_point_FOO; | |
2256 | ... | |
2257 | } | |
2258 | } | |
2259 | return result | |
2260 | } | |
2261 | ||
2262 | * WARNING: this means that any line in this function that contains a | |
2263 | * REGMATCH() or TRYPAREN() is actually simulating a recursive call to | |
2264 | * regmatch() using gotos instead. Thus the values of any local variables | |
2265 | * not saved in the regmatch_state structure will have been lost when | |
2266 | * execution resumes on the next line . | |
5d9a96ca DM |
2267 | * |
2268 | * States (ie the st pointer) are allocated in slabs of about 4K in size. | |
2269 | * PL_regmatch_state always points to the currently active state, and | |
2270 | * PL_regmatch_slab points to the slab currently containing PL_regmatch_state. | |
2271 | * The first time regmatch is called, the first slab is allocated, and is | |
2272 | * never freed until interpreter desctruction. When the slab is full, | |
2273 | * a new one is allocated chained to the end. At exit from regmatch, slabs | |
2274 | * allocated since entry are freed. | |
d6a28714 | 2275 | */ |
95b24440 | 2276 | |
40a82448 | 2277 | /* *** every FOO_fail should = FOO+1 */ |
c255a977 DM |
2278 | #define TRIE_next (REGNODE_MAX+1) |
2279 | #define TRIE_next_fail (REGNODE_MAX+2) | |
2280 | #define EVAL_A (REGNODE_MAX+3) | |
2281 | #define EVAL_A_fail (REGNODE_MAX+4) | |
2282 | #define resume_CURLYX (REGNODE_MAX+5) | |
2283 | #define resume_WHILEM1 (REGNODE_MAX+6) | |
2284 | #define resume_WHILEM2 (REGNODE_MAX+7) | |
2285 | #define resume_WHILEM3 (REGNODE_MAX+8) | |
2286 | #define resume_WHILEM4 (REGNODE_MAX+9) | |
2287 | #define resume_WHILEM5 (REGNODE_MAX+10) | |
2288 | #define resume_WHILEM6 (REGNODE_MAX+11) | |
2289 | #define BRANCH_next (REGNODE_MAX+12) | |
2290 | #define BRANCH_next_fail (REGNODE_MAX+13) | |
2291 | #define CURLYM_A (REGNODE_MAX+14) | |
2292 | #define CURLYM_A_fail (REGNODE_MAX+15) | |
2293 | #define CURLYM_B (REGNODE_MAX+16) | |
2294 | #define CURLYM_B_fail (REGNODE_MAX+17) | |
2295 | #define IFMATCH_A (REGNODE_MAX+18) | |
2296 | #define IFMATCH_A_fail (REGNODE_MAX+19) | |
2297 | #define CURLY_B_min_known (REGNODE_MAX+20) | |
2298 | #define CURLY_B_min_known_fail (REGNODE_MAX+21) | |
2299 | #define CURLY_B_min (REGNODE_MAX+22) | |
2300 | #define CURLY_B_min_fail (REGNODE_MAX+23) | |
2301 | #define CURLY_B_max (REGNODE_MAX+24) | |
2302 | #define CURLY_B_max_fail (REGNODE_MAX+25) | |
40a82448 DM |
2303 | |
2304 | ||
3dab1dad | 2305 | #define REG_NODE_NUM(x) ((x) ? (int)((x)-prog) : -1) |
95b24440 | 2306 | |
3df15adc YO |
2307 | #ifdef DEBUGGING |
2308 | ||
2309 | STATIC void | |
07be1b83 YO |
2310 | S_dump_exec_pos(pTHX_ const char *locinput, const regnode *scan, const bool do_utf8) |
2311 | { | |
2312 | const int docolor = *PL_colors[0]; | |
2313 | const int taill = (docolor ? 10 : 7); /* 3 chars for "> <" */ | |
2314 | int l = (PL_regeol - locinput) > taill ? taill : (PL_regeol - locinput); | |
2315 | /* The part of the string before starttry has one color | |
2316 | (pref0_len chars), between starttry and current | |
2317 | position another one (pref_len - pref0_len chars), | |
2318 | after the current position the third one. | |
2319 | We assume that pref0_len <= pref_len, otherwise we | |
2320 | decrease pref0_len. */ | |
2321 | int pref_len = (locinput - PL_bostr) > (5 + taill) - l | |
2322 | ? (5 + taill) - l : locinput - PL_bostr; | |
2323 | int pref0_len; | |
2324 | ||
2325 | while (do_utf8 && UTF8_IS_CONTINUATION(*(U8*)(locinput - pref_len))) | |
2326 | pref_len++; | |
2327 | pref0_len = pref_len - (locinput - PL_reg_starttry); | |
2328 | if (l + pref_len < (5 + taill) && l < PL_regeol - locinput) | |
2329 | l = ( PL_regeol - locinput > (5 + taill) - pref_len | |
2330 | ? (5 + taill) - pref_len : PL_regeol - locinput); | |
2331 | while (do_utf8 && UTF8_IS_CONTINUATION(*(U8*)(locinput + l))) | |
2332 | l--; | |
2333 | if (pref0_len < 0) | |
2334 | pref0_len = 0; | |
2335 | if (pref0_len > pref_len) | |
2336 | pref0_len = pref_len; | |
2337 | { | |
3df15adc | 2338 | const int is_uni = (do_utf8 && OP(scan) != CANY) ? 1 : 0; |
0df25f3d YO |
2339 | |
2340 | RE_PV_DISPLAY_DECL(s0,len0,is_uni,PERL_DEBUG_PAD(0), | |
2341 | (locinput - pref_len),pref0_len, 60); | |
2342 | ||
2343 | RE_PV_DISPLAY_DECL(s1,len1,is_uni,PERL_DEBUG_PAD(1), | |
3df15adc | 2344 | (locinput - pref_len + pref0_len), |
0df25f3d YO |
2345 | pref_len - pref0_len, 60); |
2346 | ||
2347 | RE_PV_DISPLAY_DECL(s2,len2,is_uni,PERL_DEBUG_PAD(2), | |
2348 | locinput, PL_regeol - locinput, 60); | |
2349 | ||
3df15adc | 2350 | PerlIO_printf(Perl_debug_log, |
07be1b83 YO |
2351 | "%4"IVdf" <%s%.*s%s%s%.*s%s%s%s%.*s%s>%*s|", |
2352 | (IV)(locinput - PL_bostr), | |
2353 | PL_colors[4], | |
2354 | len0, s0, | |
2355 | PL_colors[5], | |
2356 | PL_colors[2], | |
2357 | len1, s1, | |
2358 | PL_colors[3], | |
2359 | (docolor ? "" : "> <"), | |
2360 | PL_colors[0], | |
2361 | len2, s2, | |
2362 | PL_colors[1], | |
2363 | 15 - l - pref_len + 1, | |
2364 | ""); | |
2365 | } | |
2366 | } | |
3df15adc | 2367 | |
07be1b83 YO |
2368 | #endif |
2369 | ||
d6a28714 | 2370 | STATIC I32 /* 0 failure, 1 success */ |
3b0527fe | 2371 | S_regmatch(pTHX_ const regmatch_info *reginfo, regnode *prog) |
d6a28714 | 2372 | { |
54df2634 NC |
2373 | #if PERL_VERSION < 9 |
2374 | dMY_CXT; | |
2375 | #endif | |
27da23d5 | 2376 | dVAR; |
95b24440 | 2377 | register const bool do_utf8 = PL_reg_match_utf8; |
4ad0818d | 2378 | const U32 uniflags = UTF8_ALLOW_DEFAULT; |
95b24440 | 2379 | |
3b0527fe DM |
2380 | regexp *rex = reginfo->prog; |
2381 | ||
5d9a96ca DM |
2382 | regmatch_slab *orig_slab; |
2383 | regmatch_state *orig_state; | |
a3621e74 | 2384 | |
5d9a96ca DM |
2385 | /* the current state. This is a cached copy of PL_regmatch_state */ |
2386 | register regmatch_state *st; | |
95b24440 | 2387 | |
5d9a96ca DM |
2388 | /* cache heavy used fields of st in registers */ |
2389 | register regnode *scan; | |
2390 | register regnode *next; | |
2391 | register I32 n = 0; /* initialize to shut up compiler warning */ | |
2392 | register char *locinput = PL_reginput; | |
95b24440 | 2393 | |
5d9a96ca DM |
2394 | /* these variables are NOT saved during a recusive RFEGMATCH: */ |
2395 | register I32 nextchr; /* is always set to UCHARAT(locinput) */ | |
2396 | bool result; /* return value of S_regmatch */ | |
5d9a96ca | 2397 | int depth = 0; /* depth of recursion */ |
77cb431f DM |
2398 | regmatch_state *yes_state = NULL; /* state to pop to on success of |
2399 | subpattern */ | |
40a82448 | 2400 | U32 state_num; |
95b24440 DM |
2401 | |
2402 | #ifdef DEBUGGING | |
e68ec53f | 2403 | GET_RE_DEBUG_FLAGS_DECL; |
d6a28714 JH |
2404 | PL_regindent++; |
2405 | #endif | |
2406 | ||
5d9a96ca DM |
2407 | /* on first ever call to regmatch, allocate first slab */ |
2408 | if (!PL_regmatch_slab) { | |
2409 | Newx(PL_regmatch_slab, 1, regmatch_slab); | |
2410 | PL_regmatch_slab->prev = NULL; | |
2411 | PL_regmatch_slab->next = NULL; | |
86545054 | 2412 | PL_regmatch_state = SLAB_FIRST(PL_regmatch_slab); |
5d9a96ca DM |
2413 | } |
2414 | ||
2415 | /* remember current high-water mark for exit */ | |
2416 | /* XXX this should be done with SAVE* instead */ | |
2417 | orig_slab = PL_regmatch_slab; | |
2418 | orig_state = PL_regmatch_state; | |
2419 | ||
2420 | /* grab next free state slot */ | |
2421 | st = ++PL_regmatch_state; | |
86545054 | 2422 | if (st > SLAB_LAST(PL_regmatch_slab)) |
5d9a96ca DM |
2423 | st = PL_regmatch_state = S_push_slab(aTHX); |
2424 | ||
2425 | st->minmod = 0; | |
2426 | st->sw = 0; | |
2427 | st->logical = 0; | |
5d9a96ca | 2428 | st->cc = NULL; |
d6a28714 JH |
2429 | /* Note that nextchr is a byte even in UTF */ |
2430 | nextchr = UCHARAT(locinput); | |
2431 | scan = prog; | |
2432 | while (scan != NULL) { | |
8ba1375e | 2433 | |
a3621e74 | 2434 | DEBUG_EXECUTE_r( { |
6136c704 | 2435 | SV * const prop = sv_newmortal(); |
07be1b83 | 2436 | dump_exec_pos( locinput, scan, do_utf8 ); |
32fc9b6a | 2437 | regprop(rex, prop, scan); |
07be1b83 YO |
2438 | |
2439 | PerlIO_printf(Perl_debug_log, | |
2440 | "%3"IVdf":%*s%s(%"IVdf")\n", | |
2441 | (IV)(scan - rex->program), PL_regindent*2, "", | |
2442 | SvPVX_const(prop), | |
2443 | PL_regkind[OP(scan)] == END ? 0 : (IV)(regnext(scan) - rex->program)); | |
2a782b5b | 2444 | }); |
d6a28714 JH |
2445 | |
2446 | next = scan + NEXT_OFF(scan); | |
2447 | if (next == scan) | |
2448 | next = NULL; | |
40a82448 | 2449 | state_num = OP(scan); |
d6a28714 | 2450 | |
40a82448 DM |
2451 | reenter_switch: |
2452 | switch (state_num) { | |
d6a28714 | 2453 | case BOL: |
7fba1cd6 | 2454 | if (locinput == PL_bostr) |
d6a28714 | 2455 | { |
3b0527fe | 2456 | /* reginfo->till = reginfo->bol; */ |
b8c5462f JH |
2457 | break; |
2458 | } | |
d6a28714 JH |
2459 | sayNO; |
2460 | case MBOL: | |
12d33761 HS |
2461 | if (locinput == PL_bostr || |
2462 | ((nextchr || locinput < PL_regeol) && locinput[-1] == '\n')) | |
d6a28714 | 2463 | { |
b8c5462f JH |
2464 | break; |
2465 | } | |
d6a28714 JH |
2466 | sayNO; |
2467 | case SBOL: | |
c2a73568 | 2468 | if (locinput == PL_bostr) |
b8c5462f | 2469 | break; |
d6a28714 JH |
2470 | sayNO; |
2471 | case GPOS: | |
3b0527fe | 2472 | if (locinput == reginfo->ganch) |
d6a28714 JH |
2473 | break; |
2474 | sayNO; | |
2475 | case EOL: | |
d6a28714 JH |
2476 | goto seol; |
2477 | case MEOL: | |
d6a28714 | 2478 | if ((nextchr || locinput < PL_regeol) && nextchr != '\n') |
b8c5462f | 2479 | sayNO; |
b8c5462f | 2480 | break; |
d6a28714 JH |
2481 | case SEOL: |
2482 | seol: | |
2483 | if ((nextchr || locinput < PL_regeol) && nextchr != '\n') | |
b8c5462f | 2484 | sayNO; |
d6a28714 | 2485 | if (PL_regeol - locinput > 1) |
b8c5462f | 2486 | sayNO; |
b8c5462f | 2487 | break; |
d6a28714 JH |
2488 | case EOS: |
2489 | if (PL_regeol != locinput) | |
b8c5462f | 2490 | sayNO; |
d6a28714 | 2491 | break; |
ffc61ed2 | 2492 | case SANY: |
d6a28714 | 2493 | if (!nextchr && locinput >= PL_regeol) |
4633a7c4 | 2494 | sayNO; |
f33976b4 DB |
2495 | if (do_utf8) { |
2496 | locinput += PL_utf8skip[nextchr]; | |
2497 | if (locinput > PL_regeol) | |
2498 | sayNO; | |
2499 | nextchr = UCHARAT(locinput); | |
2500 | } | |
2501 | else | |
2502 | nextchr = UCHARAT(++locinput); | |
2503 | break; | |
2504 | case CANY: | |
2505 | if (!nextchr && locinput >= PL_regeol) | |
2506 | sayNO; | |
b8c5462f | 2507 | nextchr = UCHARAT(++locinput); |
a0d0e21e | 2508 | break; |
ffc61ed2 | 2509 | case REG_ANY: |
1aa99e6b IH |
2510 | if ((!nextchr && locinput >= PL_regeol) || nextchr == '\n') |
2511 | sayNO; | |
2512 | if (do_utf8) { | |
b8c5462f | 2513 | locinput += PL_utf8skip[nextchr]; |
d6a28714 JH |
2514 | if (locinput > PL_regeol) |
2515 | sayNO; | |
a0ed51b3 | 2516 | nextchr = UCHARAT(locinput); |
a0ed51b3 | 2517 | } |
1aa99e6b IH |
2518 | else |
2519 | nextchr = UCHARAT(++locinput); | |
a0ed51b3 | 2520 | break; |
166ba7cd DM |
2521 | |
2522 | #undef ST | |
2523 | #define ST st->u.trie | |
2524 | ||
5b47454d | 2525 | case TRIE: |
3dab1dad | 2526 | { |
07be1b83 | 2527 | /* what type of TRIE am I? (utf8 makes this contextual) */ |
3dab1dad YO |
2528 | const enum { trie_plain, trie_utf8, trie_utf8_fold } |
2529 | trie_type = do_utf8 ? | |
2530 | (scan->flags == EXACT ? trie_utf8 : trie_utf8_fold) | |
2531 | : trie_plain; | |
2532 | ||
2533 | /* what trie are we using right now */ | |
be8e71aa | 2534 | reg_trie_data * const trie |
3dab1dad YO |
2535 | = (reg_trie_data*)rex->data->data[ ARG( scan ) ]; |
2536 | U32 state = trie->startstate; | |
166ba7cd DM |
2537 | |
2538 | U8 *uc = ( U8* )locinput; | |
2539 | U16 charid = 0; | |
2540 | U32 base = 0; | |
2541 | UV uvc = 0; | |
2542 | STRLEN len = 0; | |
2543 | STRLEN foldlen = 0; | |
2544 | U8 *uscan = (U8*)NULL; | |
2545 | STRLEN bufflen=0; | |
2546 | SV *sv_accept_buff = NULL; | |
2547 | U8 foldbuf[ UTF8_MAXBYTES_CASE + 1 ]; | |
2548 | ||
2549 | ST.accepted = 0; /* how many accepting states we have seen */ | |
2550 | ST.B = next; | |
2551 | #ifdef DEBUGGING | |
2552 | ST.me = scan; | |
2553 | #endif | |
3dab1dad YO |
2554 | |
2555 | if (trie->bitmap && trie_type != trie_utf8_fold && | |
2556 | !TRIE_BITMAP_TEST(trie,*locinput) | |
2557 | ) { | |
2558 | if (trie->states[ state ].wordnum) { | |
2559 | DEBUG_EXECUTE_r( | |
2560 | PerlIO_printf(Perl_debug_log, | |
2561 | "%*s %smatched empty string...%s\n", | |
2562 | REPORT_CODE_OFF+PL_regindent*2, "", PL_colors[4], PL_colors[5]) | |
2563 | ); | |
2564 | break; | |
2565 | } else { | |
2566 | DEBUG_EXECUTE_r( | |
2567 | PerlIO_printf(Perl_debug_log, | |
2568 | "%*s %sfailed to match start class...%s\n", | |
2569 | REPORT_CODE_OFF+PL_regindent*2, "", PL_colors[4], PL_colors[5]) | |
2570 | ); | |
2571 | sayNO_SILENT; | |
2572 | } | |
2573 | } | |
166ba7cd | 2574 | |
07be1b83 YO |
2575 | /* |
2576 | traverse the TRIE keeping track of all accepting states | |
2577 | we transition through until we get to a failing node. | |
2578 | */ | |
2579 | ||
a3621e74 YO |
2580 | while ( state && uc <= (U8*)PL_regeol ) { |
2581 | ||
5b47454d | 2582 | if (trie->states[ state ].wordnum) { |
166ba7cd | 2583 | if (!ST.accepted ) { |
5b47454d DM |
2584 | ENTER; |
2585 | SAVETMPS; | |
2586 | bufflen = TRIE_INITAL_ACCEPT_BUFFLEN; | |
2587 | sv_accept_buff=newSV(bufflen * | |
2588 | sizeof(reg_trie_accepted) - 1); | |
2589 | SvCUR_set(sv_accept_buff, | |
2590 | sizeof(reg_trie_accepted)); | |
2591 | SvPOK_on(sv_accept_buff); | |
2592 | sv_2mortal(sv_accept_buff); | |
166ba7cd DM |
2593 | SAVETMPS; |
2594 | ST.accept_buff = | |
5b47454d DM |
2595 | (reg_trie_accepted*)SvPV_nolen(sv_accept_buff ); |
2596 | } | |
2597 | else { | |
166ba7cd | 2598 | if (ST.accepted >= bufflen) { |
5b47454d | 2599 | bufflen *= 2; |
166ba7cd | 2600 | ST.accept_buff =(reg_trie_accepted*) |
5b47454d DM |
2601 | SvGROW(sv_accept_buff, |
2602 | bufflen * sizeof(reg_trie_accepted)); | |
2603 | } | |
2604 | SvCUR_set(sv_accept_buff,SvCUR(sv_accept_buff) | |
2605 | + sizeof(reg_trie_accepted)); | |
2606 | } | |
166ba7cd DM |
2607 | ST.accept_buff[ST.accepted].wordnum = trie->states[state].wordnum; |
2608 | ST.accept_buff[ST.accepted].endpos = uc; | |
2609 | ++ST.accepted; | |
5b47454d | 2610 | } |
a3621e74 YO |
2611 | |
2612 | base = trie->states[ state ].trans.base; | |
2613 | ||
07be1b83 YO |
2614 | DEBUG_TRIE_EXECUTE_r({ |
2615 | dump_exec_pos( (char *)uc, scan, do_utf8 ); | |
a3621e74 | 2616 | PerlIO_printf( Perl_debug_log, |
e4584336 | 2617 | "%*s %sState: %4"UVxf", Base: %4"UVxf", Accepted: %4"UVxf" ", |
07be1b83 | 2618 | 2+PL_regindent * 2, "", PL_colors[4], |
166ba7cd | 2619 | (UV)state, (UV)base, (UV)ST.accepted ); |
07be1b83 | 2620 | }); |
a3621e74 YO |
2621 | |
2622 | if ( base ) { | |
4cadc6a9 YO |
2623 | REXEC_TRIE_READ_CHAR(trie_type, trie, uc, uscan, len, |
2624 | uvc, charid, foldlen, foldbuf, uniflags); | |
a3621e74 | 2625 | |
5b47454d DM |
2626 | if (charid && |
2627 | (base + charid > trie->uniquecharcount ) | |
2628 | && (base + charid - 1 - trie->uniquecharcount | |
2629 | < trie->lasttrans) | |
2630 | && trie->trans[base + charid - 1 - | |
2631 | trie->uniquecharcount].check == state) | |
2632 | { | |
2633 | state = trie->trans[base + charid - 1 - | |
2634 | trie->uniquecharcount ].next; | |
2635 | } | |
2636 | else { | |
2637 | state = 0; | |
2638 | } | |
2639 | uc += len; | |
2640 | ||
2641 | } | |
2642 | else { | |
a3621e74 YO |
2643 | state = 0; |
2644 | } | |
2645 | DEBUG_TRIE_EXECUTE_r( | |
e4584336 RB |
2646 | PerlIO_printf( Perl_debug_log, |
2647 | "Charid:%3x CV:%4"UVxf" After State: %4"UVxf"%s\n", | |
2648 | charid, uvc, (UV)state, PL_colors[5] ); | |
a3621e74 YO |
2649 | ); |
2650 | } | |
166ba7cd | 2651 | if (!ST.accepted ) |
a3621e74 | 2652 | sayNO; |
a3621e74 | 2653 | |
166ba7cd DM |
2654 | DEBUG_EXECUTE_r( |
2655 | PerlIO_printf( Perl_debug_log, | |
2656 | "%*s %sgot %"IVdf" possible matches%s\n", | |
2657 | REPORT_CODE_OFF + PL_regindent * 2, "", | |
2658 | PL_colors[4], (IV)ST.accepted, PL_colors[5] ); | |
2659 | ); | |
2660 | } | |
2661 | ||
2662 | /* FALL THROUGH */ | |
2663 | ||
2664 | case TRIE_next_fail: /* we failed - try next alterative */ | |
2665 | ||
2666 | if ( ST.accepted == 1 ) { | |
2667 | /* only one choice left - just continue */ | |
2668 | DEBUG_EXECUTE_r({ | |
2669 | reg_trie_data * const trie | |
2670 | = (reg_trie_data*)rex->data->data[ ARG(ST.me) ]; | |
2671 | SV ** const tmp = RX_DEBUG(reginfo->prog) | |
2672 | ? av_fetch( trie->words, ST.accept_buff[ 0 ].wordnum-1, 0 ) | |
2673 | : NULL; | |
2674 | PerlIO_printf( Perl_debug_log, | |
2675 | "%*s %sonly one match left: #%d <%s>%s\n", | |
2676 | REPORT_CODE_OFF+PL_regindent*2, "", PL_colors[4], | |
2677 | ST.accept_buff[ 0 ].wordnum, | |
2678 | tmp ? SvPV_nolen_const( *tmp ) : "not compiled under -Dr", | |
2679 | PL_colors[5] ); | |
2680 | }); | |
2681 | PL_reginput = (char *)ST.accept_buff[ 0 ].endpos; | |
2682 | /* in this case we free tmps/leave before we call regmatch | |
2683 | as we wont be using accept_buff again. */ | |
2684 | FREETMPS; | |
2685 | LEAVE; | |
2686 | locinput = PL_reginput; | |
2687 | nextchr = UCHARAT(locinput); | |
2688 | scan = ST.B; | |
2689 | continue; /* execute rest of RE */ | |
2690 | } | |
2691 | ||
2692 | if (!ST.accepted-- ) { | |
2693 | FREETMPS; | |
2694 | LEAVE; | |
2695 | sayNO; | |
2696 | } | |
2697 | ||
a3621e74 | 2698 | /* |
166ba7cd DM |
2699 | There are at least two accepting states left. Presumably |
2700 | the number of accepting states is going to be low, | |
2701 | typically two. So we simply scan through to find the one | |
2702 | with lowest wordnum. Once we find it, we swap the last | |
2703 | state into its place and decrement the size. We then try to | |
2704 | match the rest of the pattern at the point where the word | |
2705 | ends. If we succeed, control just continues along the | |
2706 | regex; if we fail we return here to try the next accepting | |
2707 | state | |
2708 | */ | |
a3621e74 | 2709 | |
166ba7cd DM |
2710 | { |
2711 | U32 best = 0; | |
2712 | U32 cur; | |
2713 | for( cur = 1 ; cur <= ST.accepted ; cur++ ) { | |
2714 | DEBUG_TRIE_EXECUTE_r( | |
f2278c82 | 2715 | PerlIO_printf( Perl_debug_log, |
166ba7cd DM |
2716 | "%*s %sgot %"IVdf" (%d) as best, looking at %"IVdf" (%d)%s\n", |
2717 | REPORT_CODE_OFF + PL_regindent * 2, "", PL_colors[4], | |
2718 | (IV)best, ST.accept_buff[ best ].wordnum, (IV)cur, | |
2719 | ST.accept_buff[ cur ].wordnum, PL_colors[5] ); | |
2720 | ); | |
2721 | ||
2722 | if (ST.accept_buff[cur].wordnum < | |
2723 | ST.accept_buff[best].wordnum) | |
2724 | best = cur; | |
a3621e74 | 2725 | } |
166ba7cd DM |
2726 | |
2727 | DEBUG_EXECUTE_r({ | |
2728 | reg_trie_data * const trie | |
2729 | = (reg_trie_data*)rex->data->data[ ARG(ST.me) ]; | |
2730 | SV ** const tmp = RX_DEBUG(reginfo->prog) | |
2731 | ? av_fetch( trie->words, ST.accept_buff[ best ].wordnum - 1, 0 ) | |
2732 | : NULL; | |
2733 | PerlIO_printf( Perl_debug_log, "%*s %strying alternation #%d <%s> at node #%d %s\n", | |
2734 | REPORT_CODE_OFF+PL_regindent*2, "", PL_colors[4], | |
2735 | ST.accept_buff[best].wordnum, | |
2736 | tmp ? SvPV_nolen_const( *tmp ) : "not compiled under -Dr", REG_NODE_NUM(scan), | |
2737 | PL_colors[5] ); | |
2738 | }); | |
2739 | ||
2740 | if ( best<ST.accepted ) { | |
2741 | reg_trie_accepted tmp = ST.accept_buff[ best ]; | |
2742 | ST.accept_buff[ best ] = ST.accept_buff[ ST.accepted ]; | |
2743 | ST.accept_buff[ ST.accepted ] = tmp; | |
2744 | best = ST.accepted; | |
a3621e74 | 2745 | } |
166ba7cd DM |
2746 | PL_reginput = (char *)ST.accept_buff[ best ].endpos; |
2747 | } | |
2748 | PUSH_STATE_GOTO(TRIE_next, ST.B); | |
2749 | /* NOTREACHED */ | |
2750 | ||
2751 | #undef ST | |
2752 | ||
95b24440 DM |
2753 | case EXACT: { |
2754 | char *s = STRING(scan); | |
5d9a96ca | 2755 | st->ln = STR_LEN(scan); |
eb160463 | 2756 | if (do_utf8 != UTF) { |
bc517b45 | 2757 | /* The target and the pattern have differing utf8ness. */ |
1aa99e6b | 2758 | char *l = locinput; |
be8e71aa | 2759 | const char * const e = s + st->ln; |
a72c7584 | 2760 | |
5ff6fc6d JH |
2761 | if (do_utf8) { |
2762 | /* The target is utf8, the pattern is not utf8. */ | |
1aa99e6b | 2763 | while (s < e) { |
a3b680e6 | 2764 | STRLEN ulen; |
1aa99e6b | 2765 | if (l >= PL_regeol) |
5ff6fc6d JH |
2766 | sayNO; |
2767 | if (NATIVE_TO_UNI(*(U8*)s) != | |
89ebb4a3 | 2768 | utf8n_to_uvuni((U8*)l, UTF8_MAXBYTES, &ulen, |
041457d9 | 2769 | uniflags)) |
5ff6fc6d | 2770 | sayNO; |
bc517b45 | 2771 | l += ulen; |
5ff6fc6d | 2772 | s ++; |
1aa99e6b | 2773 | } |
5ff6fc6d JH |
2774 | } |
2775 | else { | |
2776 | /* The target is not utf8, the pattern is utf8. */ | |
1aa99e6b | 2777 | while (s < e) { |
a3b680e6 | 2778 | STRLEN ulen; |
1aa99e6b IH |
2779 | if (l >= PL_regeol) |
2780 | sayNO; | |
5ff6fc6d | 2781 | if (NATIVE_TO_UNI(*((U8*)l)) != |
89ebb4a3 | 2782 | utf8n_to_uvuni((U8*)s, UTF8_MAXBYTES, &ulen, |
041457d9 | 2783 | uniflags)) |
1aa99e6b | 2784 | sayNO; |
bc517b45 | 2785 | s += ulen; |
a72c7584 | 2786 | l ++; |
1aa99e6b | 2787 | } |
5ff6fc6d | 2788 | } |
1aa99e6b IH |
2789 | locinput = l; |
2790 | nextchr = UCHARAT(locinput); | |
2791 | break; | |
2792 | } | |
bc517b45 | 2793 | /* The target and the pattern have the same utf8ness. */ |
d6a28714 JH |
2794 | /* Inline the first character, for speed. */ |
2795 | if (UCHARAT(s) != nextchr) | |
2796 | sayNO; | |
5d9a96ca | 2797 | if (PL_regeol - locinput < st->ln) |
d6a28714 | 2798 | sayNO; |
5d9a96ca | 2799 | if (st->ln > 1 && memNE(s, locinput, st->ln)) |
d6a28714 | 2800 | sayNO; |
5d9a96ca | 2801 | locinput += st->ln; |
d6a28714 JH |
2802 | nextchr = UCHARAT(locinput); |
2803 | break; | |
95b24440 | 2804 | } |
d6a28714 | 2805 | case EXACTFL: |
b8c5462f JH |
2806 | PL_reg_flags |= RF_tainted; |
2807 | /* FALL THROUGH */ | |
95b24440 | 2808 | case EXACTF: { |
be8e71aa | 2809 | char * const s = STRING(scan); |
5d9a96ca | 2810 | st->ln = STR_LEN(scan); |
d6a28714 | 2811 | |
d07ddd77 JH |
2812 | if (do_utf8 || UTF) { |
2813 | /* Either target or the pattern are utf8. */ | |
be8e71aa | 2814 | const char * const l = locinput; |
d07ddd77 | 2815 | char *e = PL_regeol; |
bc517b45 | 2816 | |
5d9a96ca | 2817 | if (ibcmp_utf8(s, 0, st->ln, (bool)UTF, |
1feea2c7 | 2818 | l, &e, 0, do_utf8)) { |
5486206c JH |
2819 | /* One more case for the sharp s: |
2820 | * pack("U0U*", 0xDF) =~ /ss/i, | |
2821 | * the 0xC3 0x9F are the UTF-8 | |
2822 | * byte sequence for the U+00DF. */ | |
2823 | if (!(do_utf8 && | |
2824 | toLOWER(s[0]) == 's' && | |
5d9a96ca | 2825 | st->ln >= 2 && |
5486206c JH |
2826 | toLOWER(s[1]) == 's' && |
2827 | (U8)l[0] == 0xC3 && | |
2828 | e - l >= 2 && | |
2829 | (U8)l[1] == 0x9F)) | |
2830 | sayNO; | |
2831 | } | |
d07ddd77 JH |
2832 | locinput = e; |
2833 | nextchr = UCHARAT(locinput); | |
2834 | break; | |
a0ed51b3 | 2835 | } |
d6a28714 | 2836 | |
bc517b45 JH |
2837 | /* Neither the target and the pattern are utf8. */ |
2838 | ||
d6a28714 JH |
2839 | /* Inline the first character, for speed. */ |
2840 | if (UCHARAT(s) != nextchr && | |
2841 | UCHARAT(s) != ((OP(scan) == EXACTF) | |
2842 | ? PL_fold : PL_fold_locale)[nextchr]) | |
a0ed51b3 | 2843 | sayNO; |
5d9a96ca | 2844 | if (PL_regeol - locinput < st->ln) |
b8c5462f | 2845 | sayNO; |
5d9a96ca DM |
2846 | if (st->ln > 1 && (OP(scan) == EXACTF |
2847 | ? ibcmp(s, locinput, st->ln) | |
2848 | : ibcmp_locale(s, locinput, st->ln))) | |
4633a7c4 | 2849 | sayNO; |
5d9a96ca | 2850 | locinput += st->ln; |
d6a28714 | 2851 | nextchr = UCHARAT(locinput); |
a0d0e21e | 2852 | break; |
95b24440 | 2853 | } |
d6a28714 | 2854 | case ANYOF: |
ffc61ed2 | 2855 | if (do_utf8) { |
9e55ce06 JH |
2856 | STRLEN inclasslen = PL_regeol - locinput; |
2857 | ||
32fc9b6a | 2858 | if (!reginclass(rex, scan, (U8*)locinput, &inclasslen, do_utf8)) |
e0f9d4a8 | 2859 | sayNO_ANYOF; |
ffc61ed2 JH |
2860 | if (locinput >= PL_regeol) |
2861 | sayNO; | |
0f0076b4 | 2862 | locinput += inclasslen ? inclasslen : UTF8SKIP(locinput); |
b8c5462f | 2863 | nextchr = UCHARAT(locinput); |
e0f9d4a8 | 2864 | break; |
ffc61ed2 JH |
2865 | } |
2866 | else { | |
2867 | if (nextchr < 0) | |
2868 | nextchr = UCHARAT(locinput); | |
32fc9b6a | 2869 | if (!REGINCLASS(rex, scan, (U8*)locinput)) |
e0f9d4a8 | 2870 | sayNO_ANYOF; |
ffc61ed2 JH |
2871 | if (!nextchr && locinput >= PL_regeol) |
2872 | sayNO; | |
2873 | nextchr = UCHARAT(++locinput); | |
e0f9d4a8 JH |
2874 | break; |
2875 | } | |
2876 | no_anyof: | |
2877 | /* If we might have the case of the German sharp s | |
2878 | * in a casefolding Unicode character class. */ | |
2879 | ||
ebc501f0 JH |
2880 | if (ANYOF_FOLD_SHARP_S(scan, locinput, PL_regeol)) { |
2881 | locinput += SHARP_S_SKIP; | |
e0f9d4a8 | 2882 | nextchr = UCHARAT(locinput); |
ffc61ed2 | 2883 | } |
e0f9d4a8 JH |
2884 | else |
2885 | sayNO; | |
b8c5462f | 2886 | break; |
d6a28714 | 2887 | case ALNUML: |
b8c5462f JH |
2888 | PL_reg_flags |= RF_tainted; |
2889 | /* FALL THROUGH */ | |
d6a28714 | 2890 | case ALNUM: |
b8c5462f | 2891 | if (!nextchr) |
4633a7c4 | 2892 | sayNO; |
ffc61ed2 | 2893 | if (do_utf8) { |
1a4fad37 | 2894 | LOAD_UTF8_CHARCLASS_ALNUM(); |
ffc61ed2 | 2895 | if (!(OP(scan) == ALNUM |
bb7a0f54 | 2896 | ? (bool)swash_fetch(PL_utf8_alnum, (U8*)locinput, do_utf8) |
d6a28714 | 2897 | : isALNUM_LC_utf8((U8*)locinput))) |
b8c5462f JH |
2898 | { |
2899 | sayNO; | |
a0ed51b3 | 2900 | } |
b8c5462f | 2901 | locinput += PL_utf8skip[nextchr]; |
a0ed51b3 LW |
2902 | nextchr = UCHARAT(locinput); |
2903 | break; | |
2904 | } | |
ffc61ed2 | 2905 | if (!(OP(scan) == ALNUM |
d6a28714 | 2906 | ? isALNUM(nextchr) : isALNUM_LC(nextchr))) |
bbce6d69 | 2907 | sayNO; |
b8c5462f | 2908 | nextchr = UCHARAT(++locinput); |
a0d0e21e | 2909 | break; |
d6a28714 | 2910 | case NALNUML: |
b8c5462f JH |
2911 | PL_reg_flags |= RF_tainted; |
2912 | /* FALL THROUGH */ | |
d6a28714 JH |
2913 | case NALNUM: |
2914 | if (!nextchr && locinput >= PL_regeol) | |
a0ed51b3 | 2915 | sayNO; |
ffc61ed2 | 2916 | if (do_utf8) { |
1a4fad37 | 2917 | LOAD_UTF8_CHARCLASS_ALNUM(); |
ffc61ed2 | 2918 | if (OP(scan) == NALNUM |
bb7a0f54 | 2919 | ? (bool)swash_fetch(PL_utf8_alnum, (U8*)locinput, do_utf8) |
d6a28714 JH |
2920 | : isALNUM_LC_utf8((U8*)locinput)) |
2921 | { | |
b8c5462f | 2922 | sayNO; |
d6a28714 | 2923 | } |
b8c5462f JH |
2924 | locinput += PL_utf8skip[nextchr]; |
2925 | nextchr = UCHARAT(locinput); | |
2926 | break; | |
2927 | } | |
ffc61ed2 | 2928 | if (OP(scan) == NALNUM |
d6a28714 | 2929 | ? isALNUM(nextchr) : isALNUM_LC(nextchr)) |
b8c5462f | 2930 | sayNO; |
76e3520e | 2931 | nextchr = UCHARAT(++locinput); |
a0d0e21e | 2932 | break; |
d6a28714 JH |
2933 | case BOUNDL: |
2934 | case NBOUNDL: | |
3280af22 | 2935 | PL_reg_flags |= RF_tainted; |
bbce6d69 | 2936 | /* FALL THROUGH */ |
d6a28714 JH |
2937 | case BOUND: |
2938 | case NBOUND: | |
2939 | /* was last char in word? */ | |
ffc61ed2 | 2940 | if (do_utf8) { |
12d33761 | 2941 | if (locinput == PL_bostr) |
5d9a96ca | 2942 | st->ln = '\n'; |
ffc61ed2 | 2943 | else { |
a3b680e6 | 2944 | const U8 * const r = reghop3((U8*)locinput, -1, (U8*)PL_bostr); |
9041c2e3 | 2945 | |
4ad0818d | 2946 | st->ln = utf8n_to_uvchr(r, UTF8SKIP(r), 0, uniflags); |
ffc61ed2 JH |
2947 | } |
2948 | if (OP(scan) == BOUND || OP(scan) == NBOUND) { | |
5d9a96ca | 2949 | st->ln = isALNUM_uni(st->ln); |
1a4fad37 | 2950 | LOAD_UTF8_CHARCLASS_ALNUM(); |
3568d838 | 2951 | n = swash_fetch(PL_utf8_alnum, (U8*)locinput, do_utf8); |
ffc61ed2 JH |
2952 | } |
2953 | else { | |
5d9a96ca | 2954 | st->ln = isALNUM_LC_uvchr(UNI_TO_NATIVE(st->ln)); |
ffc61ed2 JH |
2955 | n = isALNUM_LC_utf8((U8*)locinput); |
2956 | } | |
a0ed51b3 | 2957 | } |
d6a28714 | 2958 | else { |
5d9a96ca | 2959 | st->ln = (locinput != PL_bostr) ? |
12d33761 | 2960 | UCHARAT(locinput - 1) : '\n'; |
ffc61ed2 | 2961 | if (OP(scan) == BOUND || OP(scan) == NBOUND) { |
5d9a96ca | 2962 | st->ln = isALNUM(st->ln); |
ffc61ed2 JH |
2963 | n = isALNUM(nextchr); |
2964 | } | |
2965 | else { | |
5d9a96ca | 2966 | st->ln = isALNUM_LC(st->ln); |
ffc61ed2 JH |
2967 | n = isALNUM_LC(nextchr); |
2968 | } | |
d6a28714 | 2969 | } |
5d9a96ca | 2970 | if (((!st->ln) == (!n)) == (OP(scan) == BOUND || |
ffc61ed2 JH |
2971 | OP(scan) == BOUNDL)) |
2972 | sayNO; | |
a0ed51b3 | 2973 | break; |
d6a28714 | 2974 | case SPACEL: |
3280af22 | 2975 | PL_reg_flags |= RF_tainted; |
bbce6d69 | 2976 | /* FALL THROUGH */ |
d6a28714 | 2977 | case SPACE: |
9442cb0e | 2978 | if (!nextchr) |
4633a7c4 | 2979 | sayNO; |
1aa99e6b | 2980 | if (do_utf8) { |
fd400ab9 | 2981 | if (UTF8_IS_CONTINUED(nextchr)) { |
1a4fad37 | 2982 | LOAD_UTF8_CHARCLASS_SPACE(); |
ffc61ed2 | 2983 | if (!(OP(scan) == SPACE |
bb7a0f54 | 2984 | ? (bool)swash_fetch(PL_utf8_space, (U8*)locinput, do_utf8) |
ffc61ed2 JH |
2985 | : isSPACE_LC_utf8((U8*)locinput))) |
2986 | { | |
2987 | sayNO; | |
2988 | } | |
2989 | locinput += PL_utf8skip[nextchr]; | |
2990 | nextchr = UCHARAT(locinput); | |
2991 | break; | |
d6a28714 | 2992 | } |
ffc61ed2 JH |
2993 | if (!(OP(scan) == SPACE |
2994 | ? isSPACE(nextchr) : isSPACE_LC(nextchr))) | |
2995 | sayNO; | |
2996 | nextchr = UCHARAT(++locinput); | |
2997 | } | |
2998 | else { | |
2999 | if (!(OP(scan) == SPACE | |
3000 | ? isSPACE(nextchr) : isSPACE_LC(nextchr))) | |
3001 | sayNO; | |
3002 | nextchr = UCHARAT(++locinput); | |
a0ed51b3 | 3003 | } |
a0ed51b3 | 3004 | break; |
d6a28714 | 3005 | case NSPACEL: |
3280af22 | 3006 | PL_reg_flags |= RF_tainted; |
bbce6d69 | 3007 | /* FALL THROUGH */ |
d6a28714 | 3008 | case NSPACE: |
9442cb0e | 3009 | if (!nextchr && locinput >= PL_regeol) |
b8c5462f | 3010 | sayNO; |
1aa99e6b | 3011 | if (do_utf8) { |
1a4fad37 | 3012 | LOAD_UTF8_CHARCLASS_SPACE(); |
ffc61ed2 | 3013 | if (OP(scan) == NSPACE |
bb7a0f54 | 3014 | ? (bool)swash_fetch(PL_utf8_space, (U8*)locinput, do_utf8) |
d6a28714 | 3015 | : isSPACE_LC_utf8((U8*)locinput)) |
b8c5462f JH |
3016 | { |
3017 | sayNO; | |
3018 | } | |
3019 | locinput += PL_utf8skip[nextchr]; | |
3020 | nextchr = UCHARAT(locinput); | |
3021 | break; | |
a0ed51b3 | 3022 | } |
ffc61ed2 | 3023 | if (OP(scan) == NSPACE |
d6a28714 | 3024 | ? isSPACE(nextchr) : isSPACE_LC(nextchr)) |
4633a7c4 | 3025 | sayNO; |
76e3520e | 3026 | nextchr = UCHARAT(++locinput); |
a0d0e21e | 3027 | break; |
d6a28714 | 3028 | case DIGITL: |
a0ed51b3 LW |
3029 | PL_reg_flags |= RF_tainted; |
3030 | /* FALL THROUGH */ | |
d6a28714 | 3031 | case DIGIT: |
9442cb0e | 3032 | if (!nextchr) |
a0ed51b3 | 3033 | sayNO; |
1aa99e6b | 3034 | if (do_utf8) { |
1a4fad37 | 3035 | LOAD_UTF8_CHARCLASS_DIGIT(); |
ffc61ed2 | 3036 | if (!(OP(scan) == DIGIT |
bb7a0f54 | 3037 | ? (bool)swash_fetch(PL_utf8_digit, (U8*)locinput, do_utf8) |
9442cb0e | 3038 | : isDIGIT_LC_utf8((U8*)locinput))) |
dfe13c55 | 3039 | { |
a0ed51b3 | 3040 | sayNO; |
dfe13c55 | 3041 | } |
6f06b55f | 3042 | locinput += PL_utf8skip[nextchr]; |
a0ed51b3 LW |
3043 | nextchr = UCHARAT(locinput); |
3044 | break; | |
3045 | } | |
ffc61ed2 | 3046 | if (!(OP(scan) == DIGIT |
9442cb0e | 3047 | ? isDIGIT(nextchr) : isDIGIT_LC(nextchr))) |
a0ed51b3 LW |
3048 | sayNO; |
3049 | nextchr = UCHARAT(++locinput); | |
3050 | break; | |
d6a28714 | 3051 | case NDIGITL: |
b8c5462f JH |
3052 | PL_reg_flags |= RF_tainted; |
3053 | /* FALL THROUGH */ | |
d6a28714 | 3054 | case NDIGIT: |
9442cb0e | 3055 | if (!nextchr && locinput >= PL_regeol) |
b8c5462f | 3056 | sayNO; |
1aa99e6b | 3057 | if (do_utf8) { |
1a4fad37 | 3058 | LOAD_UTF8_CHARCLASS_DIGIT(); |
ffc61ed2 | 3059 | if (OP(scan) == NDIGIT |
bb7a0f54 | 3060 | ? (bool)swash_fetch(PL_utf8_digit, (U8*)locinput, do_utf8) |
9442cb0e GS |
3061 | : isDIGIT_LC_utf8((U8*)locinput)) |
3062 | { | |
a0ed51b3 | 3063 | sayNO; |
9442cb0e | 3064 | } |
6f06b55f | 3065 | locinput += PL_utf8skip[nextchr]; |
a0ed51b3 LW |
3066 | nextchr = UCHARAT(locinput); |
3067 | break; | |
3068 | } | |
ffc61ed2 | 3069 | if (OP(scan) == NDIGIT |
9442cb0e | 3070 | ? isDIGIT(nextchr) : isDIGIT_LC(nextchr)) |
a0ed51b3 LW |
3071 | sayNO; |
3072 | nextchr = UCHARAT(++locinput); | |
3073 | break; | |
3074 | case CLUMP: | |
b7c83a7e | 3075 | if (locinput >= PL_regeol) |
a0ed51b3 | 3076 | sayNO; |
b7c83a7e | 3077 | if (do_utf8) { |
1a4fad37 | 3078 | LOAD_UTF8_CHARCLASS_MARK(); |
b7c83a7e JH |
3079 | if (swash_fetch(PL_utf8_mark,(U8*)locinput, do_utf8)) |
3080 | sayNO; | |
3081 | locinput += PL_utf8skip[nextchr]; | |
3082 | while (locinput < PL_regeol && | |
3083 | swash_fetch(PL_utf8_mark,(U8*)locinput, do_utf8)) | |
3084 | locinput += UTF8SKIP(locinput); | |
3085 | if (locinput > PL_regeol) | |
3086 | sayNO; | |
eb08e2da JH |
3087 | } |
3088 | else | |
3089 | locinput++; | |
a0ed51b3 LW |
3090 | nextchr = UCHARAT(locinput); |
3091 | break; | |
c8756f30 | 3092 | case REFFL: |
3280af22 | 3093 | PL_reg_flags |= RF_tainted; |
c8756f30 | 3094 | /* FALL THROUGH */ |
c277df42 | 3095 | case REF: |
95b24440 DM |
3096 | case REFF: { |
3097 | char *s; | |
c277df42 | 3098 | n = ARG(scan); /* which paren pair */ |
5d9a96ca | 3099 | st->ln = PL_regstartp[n]; |
2c2d71f5 | 3100 | PL_reg_leftiter = PL_reg_maxiter; /* Void cache */ |
5d9a96ca | 3101 | if ((I32)*PL_reglastparen < n || st->ln == -1) |
af3f8c16 | 3102 | sayNO; /* Do not match unless seen CLOSEn. */ |
5d9a96ca | 3103 | if (st->ln == PL_regendp[n]) |
a0d0e21e | 3104 | break; |
a0ed51b3 | 3105 | |
5d9a96ca | 3106 | s = PL_bostr + st->ln; |
1aa99e6b | 3107 | if (do_utf8 && OP(scan) != REF) { /* REF can do byte comparison */ |
a0ed51b3 | 3108 | char *l = locinput; |
a3b680e6 | 3109 | const char *e = PL_bostr + PL_regendp[n]; |
a0ed51b3 LW |
3110 | /* |
3111 | * Note that we can't do the "other character" lookup trick as | |
3112 | * in the 8-bit case (no pun intended) because in Unicode we | |
3113 | * have to map both upper and title case to lower case. | |
3114 | */ | |
3115 | if (OP(scan) == REFF) { | |
3116 | while (s < e) { | |
a3b680e6 AL |
3117 | STRLEN ulen1, ulen2; |
3118 | U8 tmpbuf1[UTF8_MAXBYTES_CASE+1]; | |
3119 | U8 tmpbuf2[UTF8_MAXBYTES_CASE+1]; | |
3120 | ||
a0ed51b3 LW |
3121 | if (l >= PL_regeol) |
3122 | sayNO; | |
a2a2844f JH |
3123 | toLOWER_utf8((U8*)s, tmpbuf1, &ulen1); |
3124 | toLOWER_utf8((U8*)l, tmpbuf2, &ulen2); | |
7114a2d2 | 3125 | if (ulen1 != ulen2 || memNE((char *)tmpbuf1, (char *)tmpbuf2, ulen1)) |
a0ed51b3 | 3126 | sayNO; |
a2a2844f JH |
3127 | s += ulen1; |
3128 | l += ulen2; | |
a0ed51b3 LW |
3129 | } |
3130 | } | |
3131 | locinput = l; | |
3132 | nextchr = UCHARAT(locinput); | |
3133 | break; | |
3134 | } | |
3135 | ||
a0d0e21e | 3136 | /* Inline the first character, for speed. */ |
76e3520e | 3137 | if (UCHARAT(s) != nextchr && |
c8756f30 AK |
3138 | (OP(scan) == REF || |
3139 | (UCHARAT(s) != ((OP(scan) == REFF | |
22c35a8c | 3140 | ? PL_fold : PL_fold_locale)[nextchr])))) |
4633a7c4 | 3141 | sayNO; |
5d9a96ca DM |
3142 | st->ln = PL_regendp[n] - st->ln; |
3143 | if (locinput + st->ln > PL_regeol) | |
4633a7c4 | 3144 | sayNO; |
5d9a96ca DM |
3145 | if (st->ln > 1 && (OP(scan) == REF |
3146 | ? memNE(s, locinput, st->ln) | |
c8756f30 | 3147 | : (OP(scan) == REFF |
5d9a96ca DM |
3148 | ? ibcmp(s, locinput, st->ln) |
3149 | : ibcmp_locale(s, locinput, st->ln)))) | |
4633a7c4 | 3150 | sayNO; |
5d9a96ca | 3151 | locinput += st->ln; |
76e3520e | 3152 | nextchr = UCHARAT(locinput); |
a0d0e21e | 3153 | break; |
95b24440 | 3154 | } |
a0d0e21e LW |
3155 | |
3156 | case NOTHING: | |
c277df42 | 3157 | case TAIL: |
a0d0e21e LW |
3158 | break; |
3159 | case BACK: | |
3160 | break; | |
40a82448 DM |
3161 | |
3162 | #undef ST | |
3163 | #define ST st->u.eval | |
3164 | ||
3165 | case EVAL: /* /(?{A})B/ /(??{A})B/ and /(?(?{A})X|Y)B/ */ | |
c277df42 | 3166 | { |
c277df42 | 3167 | SV *ret; |
8e5e9ebe | 3168 | { |
4aabdb9b DM |
3169 | /* execute the code in the {...} */ |
3170 | dSP; | |
6136c704 | 3171 | SV ** const before = SP; |
4aabdb9b DM |
3172 | OP_4tree * const oop = PL_op; |
3173 | COP * const ocurcop = PL_curcop; | |
3174 | PAD *old_comppad; | |
4aabdb9b DM |
3175 | |
3176 | n = ARG(scan); | |
32fc9b6a | 3177 | PL_op = (OP_4tree*)rex->data->data[n]; |
4aabdb9b | 3178 | DEBUG_EXECUTE_r( PerlIO_printf(Perl_debug_log, " re_eval 0x%"UVxf"\n", PTR2UV(PL_op)) ); |
32fc9b6a | 3179 | PAD_SAVE_LOCAL(old_comppad, (PAD*)rex->data->data[n + 2]); |
4aabdb9b DM |
3180 | PL_regendp[0] = PL_reg_magic->mg_len = locinput - PL_bostr; |
3181 | ||
8e5e9ebe RGS |
3182 | CALLRUNOPS(aTHX); /* Scalar context. */ |
3183 | SPAGAIN; | |
3184 | if (SP == before) | |
075aa684 | 3185 | ret = &PL_sv_undef; /* protect against empty (?{}) blocks. */ |
8e5e9ebe RGS |
3186 | else { |
3187 | ret = POPs; | |
3188 | PUTBACK; | |
3189 | } | |
4aabdb9b DM |
3190 | |
3191 | PL_op = oop; | |
3192 | PAD_RESTORE_LOCAL(old_comppad); | |
3193 | PL_curcop = ocurcop; | |
3194 | if (!st->logical) { | |
3195 | /* /(?{...})/ */ | |
3196 | sv_setsv(save_scalar(PL_replgv), ret); | |
4aabdb9b DM |
3197 | break; |
3198 | } | |
8e5e9ebe | 3199 | } |
4aabdb9b DM |
3200 | if (st->logical == 2) { /* Postponed subexpression: /(??{...})/ */ |
3201 | regexp *re; | |
4aabdb9b | 3202 | { |
4f639d21 DM |
3203 | /* extract RE object from returned value; compiling if |
3204 | * necessary */ | |
3205 | ||
6136c704 | 3206 | MAGIC *mg = NULL; |
be8e71aa | 3207 | const SV *sv; |
faf82a0b AE |
3208 | if(SvROK(ret) && SvSMAGICAL(sv = SvRV(ret))) |
3209 | mg = mg_find(sv, PERL_MAGIC_qr); | |
3210 | else if (SvSMAGICAL(ret)) { | |
3211 | if (SvGMAGICAL(ret)) | |
3212 | sv_unmagic(ret, PERL_MAGIC_qr); | |
3213 | else | |
3214 | mg = mg_find(ret, PERL_MAGIC_qr); | |
0f5d15d6 | 3215 | } |
faf82a0b | 3216 | |
0f5d15d6 IZ |
3217 | if (mg) { |
3218 | re = (regexp *)mg->mg_obj; | |
df0003d4 | 3219 | (void)ReREFCNT_inc(re); |
0f5d15d6 IZ |
3220 | } |
3221 | else { | |
3222 | STRLEN len; | |
6136c704 | 3223 | const char * const t = SvPV_const(ret, len); |
0f5d15d6 | 3224 | PMOP pm; |
a3b680e6 | 3225 | const I32 osize = PL_regsize; |
0f5d15d6 | 3226 | |
5fcd1c1b | 3227 | Zero(&pm, 1, PMOP); |
4aabdb9b | 3228 | if (DO_UTF8(ret)) pm.op_pmdynflags |= PMdf_DYN_UTF8; |
83003860 | 3229 | re = CALLREGCOMP(aTHX_ (char*)t, (char*)t + len, &pm); |
9041c2e3 | 3230 | if (!(SvFLAGS(ret) |
faf82a0b AE |
3231 | & (SVs_TEMP | SVs_PADTMP | SVf_READONLY |
3232 | | SVs_GMG))) | |
14befaf4 DM |
3233 | sv_magic(ret,(SV*)ReREFCNT_inc(re), |
3234 | PERL_MAGIC_qr,0,0); | |
0f5d15d6 | 3235 | PL_regsize = osize; |
0f5d15d6 | 3236 | } |
4aabdb9b | 3237 | } |
aa283a38 DM |
3238 | |
3239 | /* run the pattern returned from (??{...}) */ | |
3240 | ||
4aabdb9b DM |
3241 | DEBUG_EXECUTE_r( |
3242 | PerlIO_printf(Perl_debug_log, | |
3243 | "Entering embedded \"%s%.60s%s%s\"\n", | |
3244 | PL_colors[0], | |
3245 | re->precomp, | |
3246 | PL_colors[1], | |
3247 | (strlen(re->precomp) > 60 ? "..." : "")) | |
3248 | ); | |
2c2d71f5 | 3249 | |
40a82448 DM |
3250 | ST.cp = regcppush(0); /* Save *all* the positions. */ |
3251 | REGCP_SET(ST.lastcp); | |
4aabdb9b DM |
3252 | *PL_reglastparen = 0; |
3253 | *PL_reglastcloseparen = 0; | |
4aabdb9b | 3254 | PL_reginput = locinput; |
4aabdb9b DM |
3255 | |
3256 | /* XXXX This is too dramatic a measure... */ | |
3257 | PL_reg_maxiter = 0; | |
3258 | ||
5d9a96ca | 3259 | st->logical = 0; |
40a82448 | 3260 | ST.toggleutf = ((PL_reg_flags & RF_utf8) != 0) ^ |
aa283a38 | 3261 | ((re->reganch & ROPT_UTF8) != 0); |
40a82448 DM |
3262 | if (ST.toggleutf) PL_reg_flags ^= RF_utf8; |
3263 | ST.prev_rex = rex; | |
aa283a38 | 3264 | rex = re; |
aa283a38 | 3265 | |
40a82448 | 3266 | ST.B = next; |
aa283a38 | 3267 | /* now continue from first node in postoned RE */ |
40a82448 | 3268 | PUSH_YES_STATE_GOTO(EVAL_A, re->program + 1); |
4aabdb9b | 3269 | /* NOTREACHED */ |
a0ed51b3 | 3270 | } |
4aabdb9b DM |
3271 | /* /(?(?{...})X|Y)/ */ |
3272 | st->sw = SvTRUE(ret); | |
3273 | st->logical = 0; | |
c277df42 IZ |
3274 | break; |
3275 | } | |
40a82448 DM |
3276 | |
3277 | case EVAL_A: /* successfully ran inner rex (??{rex}) */ | |
3278 | if (ST.toggleutf) | |
3279 | PL_reg_flags ^= RF_utf8; | |
3280 | ReREFCNT_dec(rex); | |
3281 | rex = ST.prev_rex; | |
3282 | /* XXXX This is too dramatic a measure... */ | |
3283 | PL_reg_maxiter = 0; | |
3284 | /* Restore parens of the caller without popping the | |
3285 | * savestack */ | |
3286 | { | |
3287 | const I32 tmp = PL_savestack_ix; | |
3288 | PL_savestack_ix = ST.lastcp; | |
3289 | regcppop(rex); | |
3290 | PL_savestack_ix = tmp; | |
3291 | } | |
3292 | PL_reginput = locinput; | |
3293 | /* continue at the node following the (??{...}) */ | |
3294 | scan = ST.B; | |
3295 | continue; | |
3296 | ||
3297 | case EVAL_A_fail: /* unsuccessfully ran inner rex (??{rex}) */ | |
3298 | /* Restore state to the outer re then re-throw the failure */ | |
3299 | if (ST.toggleutf) | |
3300 | PL_reg_flags ^= RF_utf8; | |
3301 | ReREFCNT_dec(rex); | |
3302 | rex = ST.prev_rex; | |
3303 | ||
3304 | /* XXXX This is too dramatic a measure... */ | |
3305 | PL_reg_maxiter = 0; | |
3306 | ||
3307 | PL_reginput = locinput; | |
3308 | REGCP_UNWIND(ST.lastcp); | |
3309 | regcppop(rex); | |
3310 | sayNO_SILENT; | |
3311 | ||
3312 | #undef ST | |
3313 | ||
a0d0e21e | 3314 | case OPEN: |
c277df42 | 3315 | n = ARG(scan); /* which paren pair */ |
3280af22 NIS |
3316 | PL_reg_start_tmp[n] = locinput; |
3317 | if (n > PL_regsize) | |
3318 | PL_regsize = n; | |
a0d0e21e LW |
3319 | break; |
3320 | case CLOSE: | |
c277df42 | 3321 | n = ARG(scan); /* which paren pair */ |
cf93c79d IZ |
3322 | PL_regstartp[n] = PL_reg_start_tmp[n] - PL_bostr; |
3323 | PL_regendp[n] = locinput - PL_bostr; | |
eb160463 | 3324 | if (n > (I32)*PL_reglastparen) |
3280af22 | 3325 | *PL_reglastparen = n; |
a01268b5 | 3326 | *PL_reglastcloseparen = n; |
a0d0e21e | 3327 | break; |
c277df42 IZ |
3328 | case GROUPP: |
3329 | n = ARG(scan); /* which paren pair */ | |
5d9a96ca | 3330 | st->sw = ((I32)*PL_reglastparen >= n && PL_regendp[n] != -1); |
c277df42 IZ |
3331 | break; |
3332 | case IFTHEN: | |
2c2d71f5 | 3333 | PL_reg_leftiter = PL_reg_maxiter; /* Void cache */ |
5d9a96ca | 3334 | if (st->sw) |
c277df42 IZ |
3335 | next = NEXTOPER(NEXTOPER(scan)); |
3336 | else { | |
3337 | next = scan + ARG(scan); | |
3338 | if (OP(next) == IFTHEN) /* Fake one. */ | |
3339 | next = NEXTOPER(NEXTOPER(next)); | |
3340 | } | |
3341 | break; | |
3342 | case LOGICAL: | |
5d9a96ca | 3343 | st->logical = scan->flags; |
c277df42 | 3344 | break; |
2ab05381 | 3345 | /******************************************************************* |
a0374537 DM |
3346 | cc points to the regmatch_state associated with the most recent CURLYX. |
3347 | This struct contains info about the innermost (...)* loop (an | |
3348 | "infoblock"), and a pointer to the next outer cc. | |
2ab05381 IZ |
3349 | |
3350 | Here is how Y(A)*Z is processed (if it is compiled into CURLYX/WHILEM): | |
3351 | ||
95b24440 | 3352 | 1) After matching Y, regnode for CURLYX is processed; |
2ab05381 | 3353 | |
a0374537 | 3354 | 2) This regnode populates cc, and calls regmatch() recursively |
95b24440 | 3355 | with the starting point at WHILEM node; |
2ab05381 IZ |
3356 | |
3357 | 3) Each hit of WHILEM node tries to match A and Z (in the order | |
3358 | depending on the current iteration, min/max of {min,max} and | |
3359 | greediness). The information about where are nodes for "A" | |
a0374537 | 3360 | and "Z" is read from cc, as is info on how many times "A" |
2ab05381 IZ |
3361 | was already matched, and greediness. |
3362 | ||
3363 | 4) After A matches, the same WHILEM node is hit again. | |
3364 | ||
95b24440 | 3365 | 5) Each time WHILEM is hit, cc is the infoblock created by CURLYX |
2ab05381 | 3366 | of the same pair. Thus when WHILEM tries to match Z, it temporarily |
95b24440 | 3367 | resets cc, since this Y(A)*Z can be a part of some other loop: |
2ab05381 IZ |
3368 | as in (Y(A)*Z)*. If Z matches, the automaton will hit the WHILEM node |
3369 | of the external loop. | |
3370 | ||
a0374537 | 3371 | Currently present infoblocks form a tree with a stem formed by st->cc |
2ab05381 IZ |
3372 | and whatever it mentions via ->next, and additional attached trees |
3373 | corresponding to temporarily unset infoblocks as in "5" above. | |
3374 | ||
95b24440 | 3375 | In the following picture, infoblocks for outer loop of |
2ab05381 IZ |
3376 | (Y(A)*?Z)*?T are denoted O, for inner I. NULL starting block |
3377 | is denoted by x. The matched string is YAAZYAZT. Temporarily postponed | |
3378 | infoblocks are drawn below the "reset" infoblock. | |
3379 | ||
3380 | In fact in the picture below we do not show failed matches for Z and T | |
3381 | by WHILEM blocks. [We illustrate minimal matches, since for them it is | |
3382 | more obvious *why* one needs to *temporary* unset infoblocks.] | |
3383 | ||
3384 | Matched REx position InfoBlocks Comment | |
3385 | (Y(A)*?Z)*?T x | |
3386 | Y(A)*?Z)*?T x <- O | |
3387 | Y (A)*?Z)*?T x <- O | |
3388 | Y A)*?Z)*?T x <- O <- I | |
3389 | YA )*?Z)*?T x <- O <- I | |
3390 | YA A)*?Z)*?T x <- O <- I | |
3391 | YAA )*?Z)*?T x <- O <- I | |
3392 | YAA Z)*?T x <- O # Temporary unset I | |
3393 | I | |
3394 | ||
3395 | YAAZ Y(A)*?Z)*?T x <- O | |
3396 | I | |
3397 | ||
3398 | YAAZY (A)*?Z)*?T x <- O | |
3399 | I | |
3400 | ||
3401 | YAAZY A)*?Z)*?T x <- O <- I | |
3402 | I | |
3403 | ||
3404 | YAAZYA )*?Z)*?T x <- O <- I | |
3405 | I | |
3406 | ||
3407 | YAAZYA Z)*?T x <- O # Temporary unset I | |
3408 | I,I | |
3409 | ||
3410 | YAAZYAZ )*?T x <- O | |
3411 | I,I | |
3412 | ||
3413 | YAAZYAZ T x # Temporary unset O | |
3414 | O | |
3415 | I,I | |
3416 | ||
3417 | YAAZYAZT x | |
3418 | O | |
3419 | I,I | |
3420 | *******************************************************************/ | |
95b24440 | 3421 | |
a0d0e21e | 3422 | case CURLYX: { |
cb434fcc IZ |
3423 | /* No need to save/restore up to this paren */ |
3424 | I32 parenfloor = scan->flags; | |
c277df42 | 3425 | |
c2b7afd3 NC |
3426 | /* Dave says: |
3427 | ||
3428 | CURLYX and WHILEM are always paired: they're the moral | |
3429 | equivalent of pp_enteriter anbd pp_iter. | |
3430 | ||
3431 | The only time next could be null is if the node tree is | |
3432 | corrupt. This was mentioned on p5p a few days ago. | |
3433 | ||
3434 | See http://www.xray.mpe.mpg.de/mailing-lists/perl5-porters/2006-04/msg00556.html | |
3435 | So we'll assert that this is true: | |
3436 | */ | |
3437 | assert(next); | |
30b2893d | 3438 | if (OP(PREVOPER(next)) == NOTHING) /* LONGJMP */ |
c277df42 | 3439 | next += ARG(next); |
cb434fcc IZ |
3440 | /* XXXX Probably it is better to teach regpush to support |
3441 | parenfloor > PL_regsize... */ | |
eb160463 | 3442 | if (parenfloor > (I32)*PL_reglastparen) |
cb434fcc | 3443 | parenfloor = *PL_reglastparen; /* Pessimization... */ |
a0374537 | 3444 | |
d8319b27 DM |
3445 | st->u.curlyx.cp = PL_savestack_ix; |
3446 | st->u.curlyx.outercc = st->cc; | |
a0374537 DM |
3447 | st->cc = st; |
3448 | /* these fields contain the state of the current curly. | |
3449 | * they are accessed by subsequent WHILEMs; | |
3450 | * cur and lastloc are also updated by WHILEM */ | |
d8319b27 DM |
3451 | st->u.curlyx.parenfloor = parenfloor; |
3452 | st->u.curlyx.cur = -1; /* this will be updated by WHILEM */ | |
3453 | st->u.curlyx.min = ARG1(scan); | |
3454 | st->u.curlyx.max = ARG2(scan); | |
3455 | st->u.curlyx.scan = NEXTOPER(scan) + EXTRA_STEP_2ARGS; | |
3456 | st->u.curlyx.lastloc = 0; | |
a0374537 DM |
3457 | /* st->next and st->minmod are also read by WHILEM */ |
3458 | ||
3280af22 | 3459 | PL_reginput = locinput; |
95b24440 DM |
3460 | REGMATCH(PREVOPER(next), CURLYX); /* start on the WHILEM */ |
3461 | /*** all unsaved local vars undefined at this point */ | |
d8319b27 DM |
3462 | regcpblow(st->u.curlyx.cp); |
3463 | st->cc = st->u.curlyx.outercc; | |
95b24440 | 3464 | saySAME(result); |
a0d0e21e | 3465 | } |
5f66b61c | 3466 | /* NOTREACHED */ |
a0d0e21e LW |
3467 | case WHILEM: { |
3468 | /* | |
3469 | * This is really hard to understand, because after we match | |
3470 | * what we're trying to match, we must make sure the rest of | |
2c2d71f5 | 3471 | * the REx is going to match for sure, and to do that we have |
a0d0e21e LW |
3472 | * to go back UP the parse tree by recursing ever deeper. And |
3473 | * if it fails, we have to reset our parent's current state | |
3474 | * that we can try again after backing off. | |
3475 | */ |