X-Git-Url: http://git.shadowcat.co.uk/gitweb/gitweb.cgi?a=blobdiff_plain;f=regexec.c;h=9ded51187d4f47bfa13ae7b5fd1a731c95397cd3;hb=38cbaf553edb0c7111e9e4365a9be9fe8ffa2c69;hp=c283b2eec351fa81cec3e13036fac8d25d0bc00e;hpb=6bda09f9db748451f9bb2b0d8c798ce595a6609c;p=p5sagit%2Fp5-mst-13.2.git diff --git a/regexec.c b/regexec.c index c283b2e..9ded511 100644 --- a/regexec.c +++ b/regexec.c @@ -150,6 +150,7 @@ } \ } STMT_END + static void restore_pos(pTHX_ void *arg); STATIC CHECKPOINT @@ -167,6 +168,7 @@ S_regcppush(pTHX_ I32 parenfloor) #define REGCP_OTHER_ELEMS 8 SSGROW(paren_elems_to_push + REGCP_OTHER_ELEMS); + for (p = PL_regsize; p > parenfloor; p--) { /* REGCP_PARENS_ELEMS are pushed per pairs of parentheses. */ SSPUSHINT(PL_regendp[p]); @@ -216,7 +218,7 @@ STATIC char * S_regcppop(pTHX_ const regexp *rex) { dVAR; - I32 i; + U32 i; char *input; GET_RE_DEBUG_FLAGS_DECL; @@ -270,7 +272,7 @@ S_regcppop(pTHX_ const regexp *rex) * building DynaLoader will fail: * "Error: '*' not in typemap in DynaLoader.xs, line 164" * --jhi */ - for (i = *PL_reglastparen + 1; (U32)i <= rex->nparens; i++) { + for (i = *PL_reglastparen + 1; i <= rex->nparens; i++) { if (i > PL_regsize) PL_regstartp[i] = -1; PL_regendp[i] = -1; @@ -329,7 +331,11 @@ Perl_pregexec(pTHX_ register regexp *prog, char *stringarg, register char *stren RExen without fixed substrings. Similarly, it is assumed that lengths of all the strings are no more than minlen, thus they cannot come from lookahead. - (Or minlen should take into account lookahead.) */ + (Or minlen should take into account lookahead.) + NOTE: Some of this comment is not correct. minlen does now take account + of lookahead/behind. Further research is required. -- demerphq + +*/ /* A failure to find a constant substring means that there is no need to make an expensive call to REx engine, thus we celebrate a failure. Similarly, @@ -365,7 +371,8 @@ Perl_re_intuit_start(pTHX_ regexp *prog, SV *sv, char *strpos, I32 ml_anch; register char *other_last = NULL; /* other substr checked before this */ char *check_at = NULL; /* check substr found at this pos */ - const I32 multiline = prog->reganch & PMf_MULTILINE; + const I32 multiline = prog->extflags & RXf_PMf_MULTILINE; + RXi_GET_DECL(prog,progi); #ifdef DEBUGGING const char * const i_strpos = strpos; #endif @@ -374,7 +381,7 @@ Perl_re_intuit_start(pTHX_ regexp *prog, SV *sv, char *strpos, RX_MATCH_UTF8_set(prog,do_utf8); - if (prog->reganch & ROPT_UTF8) { + if (prog->extflags & RXf_UTF8) { PL_reg_flags |= RF_utf8; } DEBUG_EXECUTE_r( @@ -406,14 +413,14 @@ Perl_re_intuit_start(pTHX_ regexp *prog, SV *sv, char *strpos, "Non-utf8 string cannot match utf8 check string\n")); goto fail; } - if (prog->reganch & ROPT_ANCH) { /* Match at beg-of-str or after \n */ - ml_anch = !( (prog->reganch & ROPT_ANCH_SINGLE) - || ( (prog->reganch & ROPT_ANCH_BOL) + if (prog->extflags & RXf_ANCH) { /* Match at beg-of-str or after \n */ + ml_anch = !( (prog->extflags & RXf_ANCH_SINGLE) + || ( (prog->extflags & RXf_ANCH_BOL) && !multiline ) ); /* Check after \n? */ if (!ml_anch) { - if ( !(prog->reganch & (ROPT_ANCH_GPOS /* Checked by the caller */ - | ROPT_IMPLICIT)) /* not a real BOL */ + if ( !(prog->extflags & RXf_ANCH_GPOS) /* Checked by the caller */ + && !(prog->intflags & PREGf_IMPLICIT) /* not a real BOL */ /* SvCUR is not set on references: SvRV and SvPVX_const overlap */ && sv && !SvROK(sv) && (strpos != strbeg)) { @@ -421,7 +428,7 @@ Perl_re_intuit_start(pTHX_ regexp *prog, SV *sv, char *strpos, goto fail; } if (prog->check_offset_min == prog->check_offset_max && - !(prog->reganch & ROPT_CANY_SEEN)) { + !(prog->extflags & RXf_CANY_SEEN)) { /* Substring at constant offset from beg-of-str... */ I32 slen; @@ -507,7 +514,7 @@ Perl_re_intuit_start(pTHX_ regexp *prog, SV *sv, char *strpos, if (PL_screamfirst[BmRARE(check)] >= 0 || ( BmRARE(check) == '\n' - && (BmPREVIOUS(check) == SvCUR(check) - 1) + && (BmPREVIOUS(check) == (U8)SvCUR(check) - 1) && SvTAIL(check) )) s = screaminstr(sv, check, srch_start_shift + (s - strbeg), srch_end_shift, pp, 0); @@ -522,7 +529,7 @@ Perl_re_intuit_start(pTHX_ regexp *prog, SV *sv, char *strpos, else { U8* start_point; U8* end_point; - if (prog->reganch & ROPT_CANY_SEEN) { + if (prog->extflags & RXf_CANY_SEEN) { start_point= (U8*)(s + srch_start_shift); end_point= (U8*)(strend - srch_end_shift); } else { @@ -808,17 +815,17 @@ Perl_re_intuit_start(pTHX_ regexp *prog, SV *sv, char *strpos, if (ml_anch && sv && !SvROK(sv) /* See prev comment on SvROK */ && (strpos != strbeg) && strpos[-1] != '\n' /* May be due to an implicit anchor of m{.*foo} */ - && !(prog->reganch & ROPT_IMPLICIT)) + && !(prog->intflags & PREGf_IMPLICIT)) { t = strpos; goto find_anchor; } DEBUG_EXECUTE_r( if (ml_anch) PerlIO_printf(Perl_debug_log, "Position at offset %ld does not contradict /%s^%s/m...\n", - (long)(strpos - i_strpos), PL_colors[0], PL_colors[1]); + (long)(strpos - i_strpos), PL_colors[0], PL_colors[1]); ); success_at_start: - if (!(prog->reganch & ROPT_NAUGHTY) /* XXXX If strpos moved? */ + if (!(prog->intflags & PREGf_NAUGHTY) /* XXXX If strpos moved? */ && (do_utf8 ? ( prog->check_utf8 /* Could be deleted already */ && --BmUSEFUL(prog->check_utf8) < 0 @@ -841,7 +848,7 @@ Perl_re_intuit_start(pTHX_ regexp *prog, SV *sv, char *strpos, /* XXXX This is a remnant of the old implementation. It looks wasteful, since now INTUIT can use many other heuristics. */ - prog->reganch &= ~RE_USE_INTUIT; + prog->extflags &= ~RXf_USE_INTUIT; } else s = strpos; @@ -851,7 +858,7 @@ Perl_re_intuit_start(pTHX_ regexp *prog, SV *sv, char *strpos, /* XXXX BmUSEFUL already changed, maybe multiple change is meaningful... */ /* trie stclasses are too expensive to use here, we are better off to leave it to regmatch itself */ - if (prog->regstclass && PL_regkind[OP(prog->regstclass)]!=TRIE) { + if (progi->regstclass && PL_regkind[OP(progi->regstclass)]!=TRIE) { /* minlen == 0 is possible if regstclass is \b or \B, and the fixed substr is ''$. Since minlen is already taken into account, s+1 is before strend; @@ -860,9 +867,9 @@ Perl_re_intuit_start(pTHX_ regexp *prog, SV *sv, char *strpos, regstclass does not come from lookahead... */ /* If regstclass takes bytelength more than 1: If charlength==1, OK. This leaves EXACTF only, which is dealt with in find_byclass(). */ - const U8* const str = (U8*)STRING(prog->regstclass); - const int cl_l = (PL_regkind[OP(prog->regstclass)] == EXACT - ? CHR_DIST(str+STR_LEN(prog->regstclass), str) + const U8* const str = (U8*)STRING(progi->regstclass); + const int cl_l = (PL_regkind[OP(progi->regstclass)] == EXACT + ? CHR_DIST(str+STR_LEN(progi->regstclass), str) : 1); char * endpos; if (prog->anchored_substr || prog->anchored_utf8 || ml_anch) @@ -872,11 +879,11 @@ Perl_re_intuit_start(pTHX_ regexp *prog, SV *sv, char *strpos, else endpos= strend; - DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "start_shift: %"IVdf" check_at: %d s: %d endpos: %d\n", - (IV)start_shift, check_at - strbeg, s - strbeg, endpos - strbeg)); + DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "start_shift: %"IVdf" check_at: %"IVdf" s: %"IVdf" endpos: %"IVdf"\n", + (IV)start_shift, (IV)(check_at - strbeg), (IV)(s - strbeg), (IV)(endpos - strbeg))); t = s; - s = find_byclass(prog, prog->regstclass, s, endpos, NULL); + s = find_byclass(prog, progi->regstclass, s, endpos, NULL); if (!s) { #ifdef DEBUGGING const char *what = NULL; @@ -888,7 +895,7 @@ Perl_re_intuit_start(pTHX_ regexp *prog, SV *sv, char *strpos, } DEBUG_EXECUTE_r( PerlIO_printf(Perl_debug_log, "This position contradicts STCLASS...\n") ); - if ((prog->reganch & ROPT_ANCH) && !ml_anch) + if ((prog->extflags & RXf_ANCH) && !ml_anch) goto fail; /* Contradict one of substrings */ if (prog->anchored_substr || prog->anchored_utf8) { @@ -969,8 +976,8 @@ Perl_re_intuit_start(pTHX_ regexp *prog, SV *sv, char *strpos, -#define REXEC_TRIE_READ_CHAR(trie_type, trie, uc, uscan, len, uvc, charid, \ -foldlen, foldbuf, uniflags) STMT_START { \ +#define REXEC_TRIE_READ_CHAR(trie_type, trie, widecharmap, uc, uscan, len, \ +uvc, charid, foldlen, foldbuf, uniflags) STMT_START { \ switch (trie_type) { \ case trie_utf8_fold: \ if ( foldlen>0 ) { \ @@ -998,8 +1005,8 @@ foldlen, foldbuf, uniflags) STMT_START { \ } \ else { \ charid = 0; \ - if (trie->widecharmap) { \ - SV** const svpp = hv_fetch(trie->widecharmap, \ + if (widecharmap) { \ + SV** const svpp = hv_fetch(widecharmap, \ (char*)&uvc, sizeof(UV), 0); \ if (svpp) \ charid = (U16)SvIV(*svpp); \ @@ -1012,7 +1019,7 @@ foldlen, foldbuf, uniflags) STMT_START { \ && (ln == len || \ ibcmp_utf8(s, NULL, 0, do_utf8, \ m, NULL, ln, (bool)UTF)) \ - && (!reginfo || regtry(reginfo, s)) ) \ + && (!reginfo || regtry(reginfo, &s)) ) \ goto got_it; \ else { \ U8 foldbuf[UTF8_MAXBYTES_CASE+1]; \ @@ -1025,7 +1032,7 @@ foldlen, foldbuf, uniflags) STMT_START { \ NULL, foldlen, do_utf8, \ m, \ NULL, ln, (bool)UTF)) \ - && (!reginfo || regtry(reginfo, s)) ) \ + && (!reginfo || regtry(reginfo, &s)) ) \ goto got_it; \ } \ s += len @@ -1037,7 +1044,7 @@ STMT_START { \ && (ln == 1 || !(OP(c) == EXACTF \ ? ibcmp(s, m, ln) \ : ibcmp_locale(s, m, ln))) \ - && (!reginfo || regtry(reginfo, s)) ) \ + && (!reginfo || regtry(reginfo, &s)) ) \ goto got_it; \ s++; \ } \ @@ -1062,7 +1069,7 @@ STMT_START { \ #define REXEC_FBC_UTF8_CLASS_SCAN(CoNd) \ REXEC_FBC_UTF8_SCAN( \ if (CoNd) { \ - if (tmp && (!reginfo || regtry(reginfo, s))) \ + if (tmp && (!reginfo || regtry(reginfo, &s))) \ goto got_it; \ else \ tmp = doevery; \ @@ -1074,7 +1081,7 @@ REXEC_FBC_UTF8_SCAN( \ #define REXEC_FBC_CLASS_SCAN(CoNd) \ REXEC_FBC_SCAN( \ if (CoNd) { \ - if (tmp && (!reginfo || regtry(reginfo, s))) \ + if (tmp && (!reginfo || regtry(reginfo, &s))) \ goto got_it; \ else \ tmp = doevery; \ @@ -1084,7 +1091,7 @@ REXEC_FBC_SCAN( \ ) #define REXEC_FBC_TRYIT \ -if ((!reginfo || regtry(reginfo, s))) \ +if ((!reginfo || regtry(reginfo, &s))) \ goto got_it #define REXEC_FBC_CSCAN_PRELOAD(UtFpReLoAd,CoNdUtF8,CoNd) \ @@ -1117,10 +1124,10 @@ if ((!reginfo || regtry(reginfo, s))) \ STATIC char * S_find_byclass(pTHX_ regexp * prog, const regnode *c, char *s, - const char *strend, const regmatch_info *reginfo) + const char *strend, regmatch_info *reginfo) { dVAR; - const I32 doevery = (prog->reganch & ROPT_SKIP) == 0; + const I32 doevery = (prog->intflags & PREGf_SKIP) == 0; char *m; STRLEN ln; STRLEN lnc; @@ -1130,7 +1137,8 @@ S_find_byclass(pTHX_ regexp * prog, const regnode *c, char *s, char *e; register I32 tmp = 1; /* Scratch variable? */ register const bool do_utf8 = PL_reg_match_utf8; - + RXi_GET_DECL(prog,progi); + /* We know what class it must start with. */ switch (OP(c)) { case ANYOF: @@ -1149,7 +1157,7 @@ S_find_byclass(pTHX_ regexp * prog, const regnode *c, char *s, /* The assignment of 2 is intentional: * for the folded sharp s, the skip is 2. */ (skip = SHARP_S_SKIP))) { - if (tmp && (!reginfo || regtry(reginfo, s))) + if (tmp && (!reginfo || regtry(reginfo, &s))) goto got_it; else tmp = doevery; @@ -1162,7 +1170,7 @@ S_find_byclass(pTHX_ regexp * prog, const regnode *c, char *s, break; case CANY: REXEC_FBC_SCAN( - if (tmp && (!reginfo || regtry(reginfo, s))) + if (tmp && (!reginfo || regtry(reginfo, &s))) goto got_it; else tmp = doevery; @@ -1296,7 +1304,7 @@ S_find_byclass(pTHX_ regexp * prog, const regnode *c, char *s, } ); } - if ((!prog->minlen && tmp) && (!reginfo || regtry(reginfo, s))) + if ((!prog->minlen && tmp) && (!reginfo || regtry(reginfo, &s))) goto got_it; break; case NBOUNDL: @@ -1332,7 +1340,7 @@ S_find_byclass(pTHX_ regexp * prog, const regnode *c, char *s, else REXEC_FBC_TRYIT; ); } - if ((!prog->minlen && !tmp) && (!reginfo || regtry(reginfo, s))) + if ((!prog->minlen && !tmp) && (!reginfo || regtry(reginfo, &s))) goto got_it; break; case ALNUM: @@ -1410,8 +1418,10 @@ S_find_byclass(pTHX_ regexp * prog, const regnode *c, char *s, : trie_plain; /* what trie are we using right now */ reg_ac_data *aho - = (reg_ac_data*)prog->data->data[ ARG( c ) ]; - reg_trie_data *trie=aho->trie; + = (reg_ac_data*)progi->data->data[ ARG( c ) ]; + reg_trie_data *trie + = (reg_trie_data*)progi->data->data[ aho->trie ]; + HV *widecharmap = (HV*) progi->data->data[ aho->trie + 1 ]; const char *last_start = strend - trie->minlen; #ifdef DEBUGGING @@ -1514,8 +1524,9 @@ S_find_byclass(pTHX_ regexp * prog, const regnode *c, char *s, } points[pointpos++ % maxlen]= uc; - REXEC_TRIE_READ_CHAR(trie_type, trie, uc, uscan, len, - uvc, charid, foldlen, foldbuf, uniflags); + REXEC_TRIE_READ_CHAR(trie_type, trie, widecharmap, uc, + uscan, len, uvc, charid, foldlen, + foldbuf, uniflags); DEBUG_TRIE_EXECUTE_r({ dump_exec_pos( (char *)uc, c, strend, real_start, s, do_utf8 ); @@ -1588,11 +1599,11 @@ S_find_byclass(pTHX_ regexp * prog, const regnode *c, char *s, s = (char*)leftmost; DEBUG_TRIE_EXECUTE_r({ PerlIO_printf( - Perl_debug_log,"Matches word #%"UVxf" at position %d. Trying full pattern...\n", - (UV)accepted_word, s - real_start + Perl_debug_log,"Matches word #%"UVxf" at position %"IVdf". Trying full pattern...\n", + (UV)accepted_word, (IV)(s - real_start) ); }); - if (!reginfo || regtry(reginfo, s)) { + if (!reginfo || regtry(reginfo, &s)) { FREETMPS; LEAVE; goto got_it; @@ -1629,13 +1640,15 @@ Perl_regexec_flags(pTHX_ register regexp *prog, char *stringarg, register char * /* strend: pointer to null at end of string */ /* strbeg: real beginning of string */ /* minend: end of match must be >=minend after stringarg. */ -/* data: May be used for some additional optimizations. */ +/* data: May be used for some additional optimizations. + Currently its only used, with a U32 cast, for transmitting + the ganch offset when doing a /g match. This will change */ /* nosave: For optimizations. */ { dVAR; - register char *s; + /*register*/ char *s; register regnode *c; - register char *startpos = stringarg; + /*register*/ char *startpos = stringarg; I32 minlen; /* must match at least this many chars */ I32 dontbother = 0; /* how many characters not to try at end */ I32 end_shift = 0; /* Same for the end. */ /* CC */ @@ -1644,7 +1657,7 @@ Perl_regexec_flags(pTHX_ register regexp *prog, char *stringarg, register char * SV* const oreplsv = GvSV(PL_replgv); const bool do_utf8 = (bool)DO_UTF8(sv); I32 multiline; - + RXi_GET_DECL(prog,progi); regmatch_info reginfo; /* create some info to pass to regtry etc */ GET_RE_DEBUG_FLAGS_DECL; @@ -1657,7 +1670,7 @@ Perl_regexec_flags(pTHX_ register regexp *prog, char *stringarg, register char * return 0; } - multiline = prog->reganch & PMf_MULTILINE; + multiline = prog->extflags & RXf_PMf_MULTILINE; reginfo.prog = prog; RX_MATCH_UTF8_set(prog, do_utf8); @@ -1676,7 +1689,7 @@ Perl_regexec_flags(pTHX_ register regexp *prog, char *stringarg, register char * /* Check validity of program. */ - if (UCHARAT(prog->program) != REG_MAGIC) { + if (UCHARAT(progi->program) != REG_MAGIC) { Perl_croak(aTHX_ "corrupted regexp program"); } @@ -1684,7 +1697,7 @@ Perl_regexec_flags(pTHX_ register regexp *prog, char *stringarg, register char * PL_reg_eval_set = 0; PL_reg_maxiter = 0; - if (prog->reganch & ROPT_UTF8) + if (prog->extflags & RXf_UTF8) PL_reg_flags |= RF_utf8; /* Mark beginning of line for ^ and lookbehind. */ @@ -1701,26 +1714,49 @@ Perl_regexec_flags(pTHX_ register regexp *prog, char *stringarg, register char * /* If there is a "must appear" string, look for it. */ s = startpos; - if (prog->reganch & ROPT_GPOS_SEEN) { /* Need to set reginfo->ganch */ + if (prog->extflags & RXf_GPOS_SEEN) { /* Need to set reginfo->ganch */ MAGIC *mg; if (flags & REXEC_IGNOREPOS) /* Means: check only at start */ - reginfo.ganch = startpos; + reginfo.ganch = startpos + prog->gofs; else if (sv && SvTYPE(sv) >= SVt_PVMG && SvMAGIC(sv) && (mg = mg_find(sv, PERL_MAGIC_regex_global)) && mg->mg_len >= 0) { reginfo.ganch = strbeg + mg->mg_len; /* Defined pos() */ - if (prog->reganch & ROPT_ANCH_GPOS) { + if (prog->extflags & RXf_ANCH_GPOS) { if (s > reginfo.ganch) goto phooey; - s = reginfo.ganch; + s = reginfo.ganch - prog->gofs; } } - else /* pos() not defined */ + else if (data) { + reginfo.ganch = strbeg + PTR2UV(data); + } else /* pos() not defined */ reginfo.ganch = strbeg; } - + if (PL_curpm && (PM_GETRE(PL_curpm) == prog)) { + I32 *t; + if (!progi->swap) { + /* We have to be careful. If the previous successful match + was from this regex we don't want a subsequent paritally + successful match to clobber the old results. + So when we detect this possibility we add a swap buffer + to the re, and switch the buffer each match. If we fail + we switch it back, otherwise we leave it swapped. + */ + Newxz(progi->swap, 1, regexp_paren_ofs); + /* no need to copy these */ + Newxz(progi->swap->startp, prog->nparens + 1, I32); + Newxz(progi->swap->endp, prog->nparens + 1, I32); + } + t = progi->swap->startp; + progi->swap->startp = prog->startp; + prog->startp = t; + t = progi->swap->endp; + progi->swap->endp = prog->endp; + prog->endp = t; + } if (!(flags & REXEC_CHECKED) && (prog->check_substr != NULL || prog->check_utf8 != NULL)) { re_scream_pos_data d; @@ -1737,11 +1773,11 @@ Perl_regexec_flags(pTHX_ register regexp *prog, char *stringarg, register char * /* Simplest case: anchored match need be tried only once. */ /* [unless only anchor is BOL and multiline is set] */ - if (prog->reganch & (ROPT_ANCH & ~ROPT_ANCH_GPOS)) { - if (s == startpos && regtry(®info, startpos)) + if (prog->extflags & (RXf_ANCH & ~RXf_ANCH_GPOS)) { + if (s == startpos && regtry(®info, &startpos)) goto got_it; - else if (multiline || (prog->reganch & ROPT_IMPLICIT) - || (prog->reganch & ROPT_ANCH_MBOL)) /* XXXX SBOL? */ + else if (multiline || (prog->intflags & PREGf_IMPLICIT) + || (prog->extflags & RXf_ANCH_MBOL)) /* XXXX SBOL? */ { char *end; @@ -1753,12 +1789,12 @@ Perl_regexec_flags(pTHX_ register regexp *prog, char *stringarg, register char * if (s == startpos) goto after_try; while (1) { - if (regtry(®info, s)) + if (regtry(®info, &s)) goto got_it; after_try: if (s >= end) goto phooey; - if (prog->reganch & RE_USE_INTUIT) { + if (prog->extflags & RXf_USE_INTUIT) { s = re_intuit_start(prog, sv, s + 1, strend, flags, NULL); if (!s) goto phooey; @@ -1771,25 +1807,26 @@ Perl_regexec_flags(pTHX_ register regexp *prog, char *stringarg, register char * s--; while (s < end) { if (*s++ == '\n') { /* don't need PL_utf8skip here */ - if (regtry(®info, s)) + if (regtry(®info, &s)) goto got_it; } } } } goto phooey; - } else if (ROPT_GPOS_CHECK == (prog->reganch & ROPT_GPOS_CHECK)) + } else if (RXf_GPOS_CHECK == (prog->extflags & RXf_GPOS_CHECK)) { /* the warning about reginfo.ganch being used without intialization - is bogus -- we set it above, when prog->reganch & ROPT_GPOS_SEEN + is bogus -- we set it above, when prog->extflags & RXf_GPOS_SEEN and we only enter this block when the same bit is set. */ - if (regtry(®info, reginfo.ganch)) + char *tmp_s = reginfo.ganch - prog->gofs; + if (regtry(®info, &tmp_s)) goto got_it; goto phooey; } /* Messy cases: unanchored match. */ - if ((prog->anchored_substr || prog->anchored_utf8) && prog->reganch & ROPT_SKIP) { + if ((prog->anchored_substr || prog->anchored_utf8) && prog->intflags & PREGf_SKIP) { /* we have /x+whatever/ */ /* it must be a one character string (XXXX Except UTF?) */ char ch; @@ -1804,7 +1841,7 @@ Perl_regexec_flags(pTHX_ register regexp *prog, char *stringarg, register char * REXEC_FBC_SCAN( if (*s == ch) { DEBUG_EXECUTE_r( did_match = 1 ); - if (regtry(®info, s)) goto got_it; + if (regtry(®info, &s)) goto got_it; s += UTF8SKIP(s); while (s < strend && *s == ch) s += UTF8SKIP(s); @@ -1815,7 +1852,7 @@ Perl_regexec_flags(pTHX_ register regexp *prog, char *stringarg, register char * REXEC_FBC_SCAN( if (*s == ch) { DEBUG_EXECUTE_r( did_match = 1 ); - if (regtry(®info, s)) goto got_it; + if (regtry(®info, &s)) goto got_it; s++; while (s < strend && *s == ch) s++; @@ -1897,14 +1934,14 @@ Perl_regexec_flags(pTHX_ register regexp *prog, char *stringarg, register char * } if (do_utf8) { while (s <= last1) { - if (regtry(®info, s)) + if (regtry(®info, &s)) goto got_it; s += UTF8SKIP(s); } } else { while (s <= last1) { - if (regtry(®info, s)) + if (regtry(®info, &s)) goto got_it; s++; } @@ -1920,9 +1957,9 @@ Perl_regexec_flags(pTHX_ register regexp *prog, char *stringarg, register char * }); goto phooey; } - else if ( (c = prog->regstclass) ) { + else if ( (c = progi->regstclass) ) { if (minlen) { - const OPCODE op = OP(prog->regstclass); + const OPCODE op = OP(progi->regstclass); /* don't bother with what can't match */ if (PL_regkind[op] != EXACT && op != CANY && PL_regkind[op] != TRIE) strend = HOPc(strend, -(minlen - 1)); @@ -1998,7 +2035,7 @@ Perl_regexec_flags(pTHX_ register regexp *prog, char *stringarg, register char * /* We don't know much -- general case. */ if (do_utf8) { for (;;) { - if (regtry(®info, s)) + if (regtry(®info, &s)) goto got_it; if (s >= strend) break; @@ -2007,7 +2044,7 @@ Perl_regexec_flags(pTHX_ register regexp *prog, char *stringarg, register char * } else { do { - if (regtry(®info, s)) + if (regtry(®info, &s)) goto got_it; } while (s++ < strend); } @@ -2027,6 +2064,8 @@ got_it: the same. */ restore_pos(aTHX_ prog); } + if (prog->paren_names) + (void)hv_iterinit(prog->paren_names); /* make sure $`, $&, $', and $digit will work later */ if ( !(flags & REXEC_NOT_FIRST) ) { @@ -2066,6 +2105,16 @@ phooey: PL_colors[4], PL_colors[5])); if (PL_reg_eval_set) restore_pos(aTHX_ prog); + if (progi->swap) { + /* we failed :-( roll it back */ + I32 *t; + t = progi->swap->startp; + progi->swap->startp = prog->startp; + prog->startp = t; + t = progi->swap->endp; + progi->swap->endp = prog->endp; + prog->endp = t; + } return 0; } @@ -2074,16 +2123,18 @@ phooey: - regtry - try match at specific point */ STATIC I32 /* 0 failure, 1 success */ -S_regtry(pTHX_ const regmatch_info *reginfo, char *startpos) +S_regtry(pTHX_ regmatch_info *reginfo, char **startpos) { dVAR; register I32 *sp; register I32 *ep; CHECKPOINT lastcp; regexp *prog = reginfo->prog; + RXi_GET_DECL(prog,progi); GET_RE_DEBUG_FLAGS_DECL; + reginfo->cutpoint=NULL; - if ((prog->reganch & ROPT_EVAL_SEEN) && !PL_reg_eval_set) { + if ((prog->extflags & RXf_EVAL_SEEN) && !PL_reg_eval_set) { MAGIC *mg; PL_reg_eval_set = RS_init; @@ -2091,7 +2142,7 @@ S_regtry(pTHX_ const regmatch_info *reginfo, char *startpos) PerlIO_printf(Perl_debug_log, " setting stack tmpbase at %"IVdf"\n", (IV)(PL_stack_sp - PL_stack_base)); )); - SAVEI32(cxstack[cxstack_ix].blk_oldsp); + SAVESTACK_CXPOS(); cxstack[cxstack_ix].blk_oldsp = PL_stack_sp - PL_stack_base; /* Otherwise OP_NEXTSTATE will free whatever on stack now. */ SAVETMPS; @@ -2110,8 +2161,8 @@ S_regtry(pTHX_ const regmatch_info *reginfo, char *startpos) && (mg = mg_find(reginfo->sv, PERL_MAGIC_regex_global)))) { /* prepare for quick setting of pos */ #ifdef PERL_OLD_COPY_ON_WRITE - if (SvIsCOW(sv)) - sv_force_normal_flags(sv, 0); + if (SvIsCOW(reginfo->sv)) + sv_force_normal_flags(reginfo->sv, 0); #endif mg = sv_magicext(reginfo->sv, NULL, PERL_MAGIC_regex_global, &PL_vtbl_mglob, NULL, 0); @@ -2153,9 +2204,9 @@ S_regtry(pTHX_ const regmatch_info *reginfo, char *startpos) prog->subbeg = PL_bostr; prog->sublen = PL_regeol - PL_bostr; /* strend may have been modified */ } - DEBUG_EXECUTE_r(PL_reg_starttry = startpos); - prog->startp[0] = startpos - PL_bostr; - PL_reginput = startpos; + DEBUG_EXECUTE_r(PL_reg_starttry = *startpos); + prog->startp[0] = *startpos - PL_bostr; + PL_reginput = *startpos; PL_reglastparen = &prog->lastparen; PL_reglastcloseparen = &prog->lastcloseparen; prog->lastparen = 0; @@ -2186,8 +2237,8 @@ S_regtry(pTHX_ const regmatch_info *reginfo, char *startpos) * on those tests seems to be returning null fields from matches. * --jhi */ #if 1 - sp = prog->startp; - ep = prog->endp; + sp = PL_regstartp; + ep = PL_regendp; if (prog->nparens) { register I32 i; for (i = prog->nparens; i > (I32)*PL_reglastparen; i--) { @@ -2197,10 +2248,12 @@ S_regtry(pTHX_ const regmatch_info *reginfo, char *startpos) } #endif REGCP_SET(lastcp); - if (regmatch(reginfo, prog->program + 1)) { - prog->endp[0] = PL_reginput - PL_bostr; + if (regmatch(reginfo, progi->program + 1)) { + PL_regendp[0] = PL_reginput - PL_bostr; return 1; } + if (reginfo->cutpoint) + *startpos= reginfo->cutpoint; REGCP_UNWIND(lastcp); return 0; } @@ -2209,14 +2262,12 @@ S_regtry(pTHX_ const regmatch_info *reginfo, char *startpos) #define sayYES goto yes #define sayNO goto no #define sayNO_SILENT goto no_silent -#define saySAME(x) if (x) goto yes; else goto no /* we dont use STMT_START/END here because it leads to "unreachable code" warnings, which are bogus, but distracting. */ #define CACHEsayNO \ - if (st->u.whilem.cache_offset | st->u.whilem.cache_bit) \ - PL_reg_poscache[st->u.whilem.cache_offset] |= \ - (1<u.whilem.cache_bit); \ + if (ST.cache_mask) \ + PL_reg_poscache[ST.cache_offset] |= ST.cache_mask; \ sayNO /* this is used to determine how far from the left messages like @@ -2254,14 +2305,6 @@ S_push_slab(pTHX) return SLAB_FIRST(s); } -/* simulate a recursive call to regmatch */ - -#define REGMATCH(ns, where) \ - st->scan = scan; \ - scan = (ns); \ - st->resume_state = resume_##where; \ - goto start_recurse; \ - resume_point_##where: /* push a new state then goto it */ @@ -2280,118 +2323,151 @@ S_push_slab(pTHX) /* - - regmatch - main matching routine - * - * Conceptually the strategy is simple: check to see whether the current - * node matches, call self recursively to see whether the rest matches, - * and then act accordingly. In practice we make some effort to avoid - * recursion, in particular by going through "ordinary" nodes (that don't - * need to know whether the rest of the match failed) by a loop instead of - * by recursion. - */ -/* [lwall] I've hoisted the register declarations to the outer block in order to - * maybe save a little bit of pushing and popping on the stack. It also takes - * advantage of machines that use a register save mask on subroutine entry. - * - * This function used to be heavily recursive, but since this had the - * effect of blowing the CPU stack on complex regexes, it has been - * restructured to be iterative, and to save state onto the heap rather - * than the stack. Essentially whereever regmatch() used to be called, it - * pushes the current state, notes where to return, then jumps back into - * the main loop. - * - * Originally the structure of this function used to look something like - S_regmatch() { - int a = 1, b = 2; +regmatch() - main matching routine + +This is basically one big switch statement in a loop. We execute an op, +set 'next' to point the next op, and continue. If we come to a point which +we may need to backtrack to on failure such as (A|B|C), we push a +backtrack state onto the backtrack stack. On failure, we pop the top +state, and re-enter the loop at the state indicated. If there are no more +states to pop, we return failure. + +Sometimes we also need to backtrack on success; for example /A+/, where +after successfully matching one A, we need to go back and try to +match another one; similarly for lookahead assertions: if the assertion +completes successfully, we backtrack to the state just before the assertion +and then carry on. In these cases, the pushed state is marked as +'backtrack on success too'. This marking is in fact done by a chain of +pointers, each pointing to the previous 'yes' state. On success, we pop to +the nearest yes state, discarding any intermediate failure-only states. +Sometimes a yes state is pushed just to force some cleanup code to be +called at the end of a successful match or submatch; e.g. (??{$re}) uses +it to free the inner regex. + +Note that failure backtracking rewinds the cursor position, while +success backtracking leaves it alone. + +A pattern is complete when the END op is executed, while a subpattern +such as (?=foo) is complete when the SUCCESS op is executed. Both of these +ops trigger the "pop to last yes state if any, otherwise return true" +behaviour. + +A common convention in this function is to use A and B to refer to the two +subpatterns (or to the first nodes thereof) in patterns like /A*B/: so A is +the subpattern to be matched possibly multiple times, while B is the entire +rest of the pattern. Variable and state names reflect this convention. + +The states in the main switch are the union of ops and failure/success of +substates associated with with that op. For example, IFMATCH is the op +that does lookahead assertions /(?=A)B/ and so the IFMATCH state means +'execute IFMATCH'; while IFMATCH_A is a state saying that we have just +successfully matched A and IFMATCH_A_fail is a state saying that we have +just failed to match A. Resume states always come in pairs. The backtrack +state we push is marked as 'IFMATCH_A', but when that is popped, we resume +at IFMATCH_A or IFMATCH_A_fail, depending on whether we are backtracking +on success or failure. + +The struct that holds a backtracking state is actually a big union, with +one variant for each major type of op. The variable st points to the +top-most backtrack struct. To make the code clearer, within each +block of code we #define ST to alias the relevant union. + +Here's a concrete example of a (vastly oversimplified) IFMATCH +implementation: + + switch (state) { + .... + +#define ST st->u.ifmatch + + case IFMATCH: // we are executing the IFMATCH op, (?=A)B + ST.foo = ...; // some state we wish to save ... - while (scan != NULL) { - a++; // do stuff with a and b - ... - switch (OP(scan)) { - case FOO: { - int local = 3; - ... - if (regmatch(...)) // recurse - goto yes; - } - ... - } - } - yes: - return 1; - } + // push a yes backtrack state with a resume value of + // IFMATCH_A/IFMATCH_A_fail, then continue execution at the + // first node of A: + PUSH_YES_STATE_GOTO(IFMATCH_A, A); + // NOTREACHED + + case IFMATCH_A: // we have successfully executed A; now continue with B + next = B; + bar = ST.foo; // do something with the preserved value + break; + + case IFMATCH_A_fail: // A failed, so the assertion failed + ...; // do some housekeeping, then ... + sayNO; // propagate the failure - * Now it looks something like this: +#undef ST + + ... + } - typedef struct { - int a, b, local; - int resume_state; - } regmatch_state; +For any old-timers reading this who are familiar with the old recursive +approach, the code above is equivalent to: - S_regmatch() { - regmatch_state *st = new(); - int depth=0; - st->a++; // do stuff with a and b + case IFMATCH: // we are executing the IFMATCH op, (?=A)B + { + int foo = ... ... - while (scan != NULL) { - ... - switch (OP(scan)) { - case FOO: { - st->local = 3; - ... - st->scan = scan; - scan = ...; - st->resume_state = resume_FOO; - goto start_recurse; // recurse - - resume_point_FOO: - if (result) - goto yes; - } - ... - } - start_recurse: - st = new(); push a new state - st->a = 1; st->b = 2; - depth++; - } - yes: - result = 1; - if (depth--) { - st = pop(); - switch (resume_state) { - case resume_FOO: - goto resume_point_FOO; - ... - } + if (regmatch(A)) { + next = B; + bar = foo; + break; } - return result + ...; // do some housekeeping, then ... + sayNO; // propagate the failure } - - * WARNING: this means that any line in this function that contains a - * REGMATCH() or TRYPAREN() is actually simulating a recursive call to - * regmatch() using gotos instead. Thus the values of any local variables - * not saved in the regmatch_state structure will have been lost when - * execution resumes on the next line . - * - * States (ie the st pointer) are allocated in slabs of about 4K in size. - * PL_regmatch_state always points to the currently active state, and - * PL_regmatch_slab points to the slab currently containing PL_regmatch_state. - * The first time regmatch is called, the first slab is allocated, and is - * never freed until interpreter desctruction. When the slab is full, - * a new one is allocated chained to the end. At exit from regmatch, slabs - * allocated since entry are freed. - */ + +The topmost backtrack state, pointed to by st, is usually free. If you +want to claim it, populate any ST.foo fields in it with values you wish to +save, then do one of + + PUSH_STATE_GOTO(resume_state, node); + PUSH_YES_STATE_GOTO(resume_state, node); + +which sets that backtrack state's resume value to 'resume_state', pushes a +new free entry to the top of the backtrack stack, then goes to 'node'. +On backtracking, the free slot is popped, and the saved state becomes the +new free state. An ST.foo field in this new top state can be temporarily +accessed to retrieve values, but once the main loop is re-entered, it +becomes available for reuse. + +Note that the depth of the backtrack stack constantly increases during the +left-to-right execution of the pattern, rather than going up and down with +the pattern nesting. For example the stack is at its maximum at Z at the +end of the pattern, rather than at X in the following: + + /(((X)+)+)+....(Y)+....Z/ + +The only exceptions to this are lookahead/behind assertions and the cut, +(?>A), which pop all the backtrack states associated with A before +continuing. + +Bascktrack state structs are allocated in slabs of about 4K in size. +PL_regmatch_state and st always point to the currently active state, +and PL_regmatch_slab points to the slab currently containing +PL_regmatch_state. The first time regmatch() is called, the first slab is +allocated, and is never freed until interpreter destruction. When the slab +is full, a new one is allocated and chained to the end. At exit from +regmatch(), slabs allocated since entry are freed. + +*/ #define DEBUG_STATE_pp(pp) \ DEBUG_STATE_r({ \ DUMP_EXEC_POS(locinput, scan, do_utf8); \ PerlIO_printf(Perl_debug_log, \ - " %*s"pp" %s\n", \ + " %*s"pp" %s%s%s%s%s\n", \ depth*2, "", \ - reg_name[st->resume_state] ); \ + reg_name[st->resume_state], \ + ((st==yes_state||st==mark_state) ? "[" : ""), \ + ((st==yes_state) ? "Y" : ""), \ + ((st==mark_state) ? "M" : ""), \ + ((st==yes_state||st==mark_state) ? "]" : "") \ + ); \ }); @@ -2403,7 +2479,7 @@ STATIC void S_debug_start_match(pTHX_ const regexp *prog, const bool do_utf8, const char *start, const char *end, const char *blurb) { - const bool utf8_pat= prog->reganch & ROPT_UTF8 ? 1 : 0; + const bool utf8_pat= prog->extflags & RXf_UTF8 ? 1 : 0; if (!PL_colorset) reginitcolors(); { @@ -2487,8 +2563,33 @@ S_dump_exec_pos(pTHX_ const char *locinput, #endif +/* reg_check_named_buff_matched() + * Checks to see if a named buffer has matched. The data array of + * buffer numbers corresponding to the buffer is expected to reside + * in the regexp->data->data array in the slot stored in the ARG() of + * node involved. Note that this routine doesn't actually care about the + * name, that information is not preserved from compilation to execution. + * Returns the index of the leftmost defined buffer with the given name + * or 0 if non of the buffers matched. + */ +STATIC I32 +S_reg_check_named_buff_matched(pTHX_ const regexp *rex, const regnode *scan) { + I32 n; + RXi_GET_DECL(rex,rexi); + SV *sv_dat=(SV*)rexi->data->data[ ARG( scan ) ]; + I32 *nums=(I32*)SvPVX(sv_dat); + for ( n=0; n= nums[n] && + PL_regendp[nums[n]] != -1) + { + return nums[n]; + } + } + return 0; +} + STATIC I32 /* 0 failure, 1 success */ -S_regmatch(pTHX_ const regmatch_info *reginfo, regnode *prog) +S_regmatch(pTHX_ regmatch_info *reginfo, regnode *prog) { #if PERL_VERSION < 9 dMY_CXT; @@ -2498,7 +2599,8 @@ S_regmatch(pTHX_ const regmatch_info *reginfo, regnode *prog) const U32 uniflags = UTF8_ALLOW_DEFAULT; regexp *rex = reginfo->prog; - + RXi_GET_DECL(rex,rexi); + regmatch_slab *orig_slab; regmatch_state *orig_state; @@ -2508,26 +2610,56 @@ S_regmatch(pTHX_ const regmatch_info *reginfo, regnode *prog) /* cache heavy used fields of st in registers */ register regnode *scan; register regnode *next; - register I32 n = 0; /* initialize to shut up compiler warning */ + register U32 n = 0; /* general value; init to avoid compiler warning */ + register I32 ln = 0; /* len or last; init to avoid compiler warning */ register char *locinput = PL_reginput; - - /* these variables are NOT saved during a recusive RFEGMATCH: */ register I32 nextchr; /* is always set to UCHARAT(locinput) */ + bool result = 0; /* return value of S_regmatch */ - int depth = 0; /* depth of recursion */ - int nochange_depth = 0; /* depth of RECURSE recursion with nochange*/ + int depth = 0; /* depth of backtrack stack */ + int nochange_depth = 0; /* depth of GOSUB recursion with nochange*/ regmatch_state *yes_state = NULL; /* state to pop to on success of subpattern */ + /* mark_state piggy backs on the yes_state logic so that when we unwind + the stack on success we can update the mark_state as we go */ + regmatch_state *mark_state = NULL; /* last mark state we have seen */ + regmatch_state *cur_eval = NULL; /* most recent EVAL_AB state */ struct regmatch_state *cur_curlyx = NULL; /* most recent curlyx */ U32 state_num; + bool no_final = 0; /* prevent failure from backtracking? */ + bool do_cutgroup = 0; /* no_final only until next branch/trie entry */ + char *startpoint = PL_reginput; + SV *popmark = NULL; /* are we looking for a mark? */ + SV *sv_commit = NULL; /* last mark name seen in failure */ + SV *sv_yes_mark = NULL; /* last mark name we have seen + during a successfull match */ + U32 lastopen = 0; /* last open we saw */ + bool has_cutgroup = RX_HAS_CUTGROUP(rex) ? 1 : 0; + - I32 parenfloor = 0; + /* these three flags are set by various ops to signal information to + * the very next op. They have a useful lifetime of exactly one loop + * iteration, and are not preserved or restored by state pushes/pops + */ + bool sw = 0; /* the condition value in (?(cond)a|b) */ + bool minmod = 0; /* the next "{n,m}" is a "{n,m}?" */ + int logical = 0; /* the following EVAL is: + 0: (?{...}) + 1: (?(?{...})X|Y) + 2: (??{...}) + or the following IFMATCH/UNLESSM is: + false: plain (?=foo) + true: used as a condition: (?(?=foo)) + */ #ifdef DEBUGGING GET_RE_DEBUG_FLAGS_DECL; #endif + DEBUG_OPTIMISE_r( { + PerlIO_printf(Perl_debug_log,"regmatch start\n"); + }); /* on first ever call to regmatch, allocate first slab */ if (!PL_regmatch_slab) { Newx(PL_regmatch_slab, 1, regmatch_slab); @@ -2546,11 +2678,6 @@ S_regmatch(pTHX_ const regmatch_info *reginfo, regnode *prog) if (st > SLAB_LAST(PL_regmatch_slab)) st = PL_regmatch_state = S_push_slab(aTHX); - st->minmod = 0; - st->sw = 0; - st->logical = 0; - cur_curlyx = NULL; - /* Note that nextchr is a byte even in UTF */ nextchr = UCHARAT(locinput); scan = prog; @@ -2564,10 +2691,10 @@ S_regmatch(pTHX_ const regmatch_info *reginfo, regnode *prog) PerlIO_printf(Perl_debug_log, "%3"IVdf":%*s%s(%"IVdf")\n", - (IV)(scan - rex->program), depth*2, "", + (IV)(scan - rexi->program), depth*2, "", SvPVX_const(prop), (PL_regkind[OP(scan)] == END || !rnext) ? - 0 : (IV)(rnext - rex->program)); + 0 : (IV)(rnext - rexi->program)); }); next = scan + NEXT_OFF(scan); @@ -2674,7 +2801,8 @@ S_regmatch(pTHX_ const regmatch_info *reginfo, regnode *prog) /* what trie are we using right now */ reg_trie_data * const trie - = (reg_trie_data*)rex->data->data[ ARG( scan ) ]; + = (reg_trie_data*)rexi->data->data[ ARG( scan ) ]; + HV * widecharmap = (HV *)rexi->data->data[ ARG( scan ) + 1 ]; U32 state = trie->startstate; if (trie->bitmap && trie_type != trie_utf8_fold && @@ -2710,13 +2838,8 @@ S_regmatch(pTHX_ const regmatch_info *reginfo, regnode *prog) ST.accepted = 0; /* how many accepting states we have seen */ ST.B = next; ST.jump = trie->jump; - -#ifdef DEBUGGING ST.me = scan; -#endif - - /* traverse the TRIE keeping track of all accepting states we transition through until we get to a failing node. @@ -2775,8 +2898,9 @@ S_regmatch(pTHX_ const regmatch_info *reginfo, regnode *prog) }); if ( base ) { - REXEC_TRIE_READ_CHAR(trie_type, trie, uc, uscan, len, - uvc, charid, foldlen, foldbuf, uniflags); + REXEC_TRIE_READ_CHAR(trie_type, trie, widecharmap, uc, + uscan, len, uvc, charid, foldlen, + foldbuf, uniflags); if (charid && (base + charid > trie->uniquecharcount ) @@ -2815,44 +2939,63 @@ S_regmatch(pTHX_ const regmatch_info *reginfo, regnode *prog) }} /* FALL THROUGH */ - case TRIE_next_fail: /* we failed - try next alterative */ - + if (do_cutgroup) { + do_cutgroup = 0; + no_final = 0; + } if ( ST.accepted == 1 ) { /* only one choice left - just continue */ DEBUG_EXECUTE_r({ - reg_trie_data * const trie - = (reg_trie_data*)rex->data->data[ ARG(ST.me) ]; - SV ** const tmp = RX_DEBUG(reginfo->prog) - ? av_fetch( trie->words, ST.accept_buff[ 0 ].wordnum-1, 0 ) - : NULL; + AV *const trie_words + = (AV *) rexi->data->data[ARG(ST.me)+TRIE_WORDS_OFFSET]; + SV ** const tmp = av_fetch( trie_words, + ST.accept_buff[ 0 ].wordnum-1, 0 ); + SV *sv= tmp ? sv_newmortal() : NULL; + PerlIO_printf( Perl_debug_log, "%*s %sonly one match left: #%d <%s>%s\n", REPORT_CODE_OFF+depth*2, "", PL_colors[4], ST.accept_buff[ 0 ].wordnum, - tmp ? SvPV_nolen_const( *tmp ) : "not compiled under -Dr", + tmp ? pv_pretty(sv, SvPV_nolen_const(*tmp), SvCUR(*tmp), 0, + PL_colors[0], PL_colors[1], + (SvUTF8(*tmp) ? PERL_PV_ESCAPE_UNI : 0) + ) + : "not compiled under -Dr", PL_colors[5] ); }); PL_reginput = (char *)ST.accept_buff[ 0 ].endpos; /* in this case we free tmps/leave before we call regmatch as we wont be using accept_buff again. */ - FREETMPS; - LEAVE; + locinput = PL_reginput; nextchr = UCHARAT(locinput); - - if ( !ST.jump ) - scan = ST.B; - else - scan = ST.B - ST.jump[ST.accept_buff[0].wordnum]; + if ( !ST.jump || !ST.jump[ST.accept_buff[0].wordnum]) + scan = ST.B; + else + scan = ST.me + ST.jump[ST.accept_buff[0].wordnum]; + if (!has_cutgroup) { + FREETMPS; + LEAVE; + } else { + ST.accepted--; + PUSH_YES_STATE_GOTO(TRIE_next, scan); + } continue; /* execute rest of RE */ } if (!ST.accepted-- ) { + DEBUG_EXECUTE_r({ + PerlIO_printf( Perl_debug_log, + "%*s %sTRIE failed...%s\n", + REPORT_CODE_OFF+depth*2, "", + PL_colors[4], + PL_colors[5] ); + }); FREETMPS; LEAVE; - sayNO; + sayNO_SILENT; } /* @@ -2885,19 +3028,23 @@ S_regmatch(pTHX_ const regmatch_info *reginfo, regnode *prog) } DEBUG_EXECUTE_r({ - reg_trie_data * const trie - = (reg_trie_data*)rex->data->data[ ARG(ST.me) ]; - SV ** const tmp = RX_DEBUG(reginfo->prog) - ? av_fetch( trie->words, ST.accept_buff[ best ].wordnum - 1, 0 ) - : NULL; - regnode *nextop=!ST.jump ? + AV *const trie_words + = (AV *) rexi->data->data[ARG(ST.me)+TRIE_WORDS_OFFSET]; + SV ** const tmp = av_fetch( trie_words, + ST.accept_buff[ best ].wordnum - 1, 0 ); + regnode *nextop=(!ST.jump || !ST.jump[ST.accept_buff[best].wordnum]) ? ST.B : - ST.B - ST.jump[ST.accept_buff[best].wordnum]; + ST.me + ST.jump[ST.accept_buff[best].wordnum]; + SV *sv= tmp ? sv_newmortal() : NULL; + PerlIO_printf( Perl_debug_log, "%*s %strying alternation #%d <%s> at node #%d %s\n", REPORT_CODE_OFF+depth*2, "", PL_colors[4], ST.accept_buff[best].wordnum, - tmp ? SvPV_nolen_const( *tmp ) : "not compiled under -Dr", + tmp ? pv_pretty(sv, SvPV_nolen_const(*tmp), SvCUR(*tmp), 0, + PL_colors[0], PL_colors[1], + (SvUTF8(*tmp) ? PERL_PV_ESCAPE_UNI : 0) + ) : "not compiled under -Dr", REG_NODE_NUM(nextop), PL_colors[5] ); }); @@ -2909,26 +3056,36 @@ S_regmatch(pTHX_ const regmatch_info *reginfo, regnode *prog) best = ST.accepted; } PL_reginput = (char *)ST.accept_buff[ best ].endpos; - if ( !ST.jump ) { - PUSH_STATE_GOTO(TRIE_next, ST.B); + if ( !ST.jump || !ST.jump[ST.accept_buff[best].wordnum]) { + scan = ST.B; /* NOTREACHED */ } else { - PUSH_STATE_GOTO(TRIE_next, ST.B - ST.jump[ST.accept_buff[best].wordnum]); + scan = ST.me + ST.jump[ST.accept_buff[best].wordnum]; /* NOTREACHED */ } + if (has_cutgroup) { + PUSH_YES_STATE_GOTO(TRIE_next, scan); + /* NOTREACHED */ + } else { + PUSH_STATE_GOTO(TRIE_next, scan); + /* NOTREACHED */ + } /* NOTREACHED */ } /* NOTREACHED */ - + case TRIE_next: + FREETMPS; + LEAVE; + sayYES; #undef ST case EXACT: { char *s = STRING(scan); - st->ln = STR_LEN(scan); + ln = STR_LEN(scan); if (do_utf8 != UTF) { /* The target and the pattern have differing utf8ness. */ char *l = locinput; - const char * const e = s + st->ln; + const char * const e = s + ln; if (do_utf8) { /* The target is utf8, the pattern is not utf8. */ @@ -2966,11 +3123,11 @@ S_regmatch(pTHX_ const regmatch_info *reginfo, regnode *prog) /* Inline the first character, for speed. */ if (UCHARAT(s) != nextchr) sayNO; - if (PL_regeol - locinput < st->ln) + if (PL_regeol - locinput < ln) sayNO; - if (st->ln > 1 && memNE(s, locinput, st->ln)) + if (ln > 1 && memNE(s, locinput, ln)) sayNO; - locinput += st->ln; + locinput += ln; nextchr = UCHARAT(locinput); break; } @@ -2979,14 +3136,14 @@ S_regmatch(pTHX_ const regmatch_info *reginfo, regnode *prog) /* FALL THROUGH */ case EXACTF: { char * const s = STRING(scan); - st->ln = STR_LEN(scan); + ln = STR_LEN(scan); if (do_utf8 || UTF) { /* Either target or the pattern are utf8. */ const char * const l = locinput; char *e = PL_regeol; - if (ibcmp_utf8(s, 0, st->ln, (bool)UTF, + if (ibcmp_utf8(s, 0, ln, (bool)UTF, l, &e, 0, do_utf8)) { /* One more case for the sharp s: * pack("U0U*", 0xDF) =~ /ss/i, @@ -2994,7 +3151,7 @@ S_regmatch(pTHX_ const regmatch_info *reginfo, regnode *prog) * byte sequence for the U+00DF. */ if (!(do_utf8 && toLOWER(s[0]) == 's' && - st->ln >= 2 && + ln >= 2 && toLOWER(s[1]) == 's' && (U8)l[0] == 0xC3 && e - l >= 2 && @@ -3013,13 +3170,13 @@ S_regmatch(pTHX_ const regmatch_info *reginfo, regnode *prog) UCHARAT(s) != ((OP(scan) == EXACTF) ? PL_fold : PL_fold_locale)[nextchr]) sayNO; - if (PL_regeol - locinput < st->ln) + if (PL_regeol - locinput < ln) sayNO; - if (st->ln > 1 && (OP(scan) == EXACTF - ? ibcmp(s, locinput, st->ln) - : ibcmp_locale(s, locinput, st->ln))) + if (ln > 1 && (OP(scan) == EXACTF + ? ibcmp(s, locinput, ln) + : ibcmp_locale(s, locinput, ln))) sayNO; - locinput += st->ln; + locinput += ln; nextchr = UCHARAT(locinput); break; } @@ -3111,35 +3268,35 @@ S_regmatch(pTHX_ const regmatch_info *reginfo, regnode *prog) /* was last char in word? */ if (do_utf8) { if (locinput == PL_bostr) - st->ln = '\n'; + ln = '\n'; else { const U8 * const r = reghop3((U8*)locinput, -1, (U8*)PL_bostr); - st->ln = utf8n_to_uvchr(r, UTF8SKIP(r), 0, uniflags); + ln = utf8n_to_uvchr(r, UTF8SKIP(r), 0, uniflags); } if (OP(scan) == BOUND || OP(scan) == NBOUND) { - st->ln = isALNUM_uni(st->ln); + ln = isALNUM_uni(ln); LOAD_UTF8_CHARCLASS_ALNUM(); n = swash_fetch(PL_utf8_alnum, (U8*)locinput, do_utf8); } else { - st->ln = isALNUM_LC_uvchr(UNI_TO_NATIVE(st->ln)); + ln = isALNUM_LC_uvchr(UNI_TO_NATIVE(ln)); n = isALNUM_LC_utf8((U8*)locinput); } } else { - st->ln = (locinput != PL_bostr) ? + ln = (locinput != PL_bostr) ? UCHARAT(locinput - 1) : '\n'; if (OP(scan) == BOUND || OP(scan) == NBOUND) { - st->ln = isALNUM(st->ln); + ln = isALNUM(ln); n = isALNUM(nextchr); } else { - st->ln = isALNUM_LC(st->ln); + ln = isALNUM_LC(ln); n = isALNUM_LC(nextchr); } } - if (((!st->ln) == (!n)) == (OP(scan) == BOUND || + if (((!ln) == (!n)) == (OP(scan) == BOUND || OP(scan) == BOUNDL)) sayNO; break; @@ -3261,22 +3418,42 @@ S_regmatch(pTHX_ const regmatch_info *reginfo, regnode *prog) locinput++; nextchr = UCHARAT(locinput); break; + + case NREFFL: + { + char *s; + char type; + PL_reg_flags |= RF_tainted; + /* FALL THROUGH */ + case NREF: + case NREFF: + type = OP(scan); + n = reg_check_named_buff_matched(rex,scan); + + if ( n ) { + type = REF + ( type - NREF ); + goto do_ref; + } else { + sayNO; + } + /* unreached */ case REFFL: PL_reg_flags |= RF_tainted; /* FALL THROUGH */ case REF: - case REFF: { - char *s; + case REFF: n = ARG(scan); /* which paren pair */ - st->ln = PL_regstartp[n]; + type = OP(scan); + do_ref: + ln = PL_regstartp[n]; PL_reg_leftiter = PL_reg_maxiter; /* Void cache */ - if ((I32)*PL_reglastparen < n || st->ln == -1) + if (*PL_reglastparen < n || ln == -1) sayNO; /* Do not match unless seen CLOSEn. */ - if (st->ln == PL_regendp[n]) + if (ln == PL_regendp[n]) break; - s = PL_bostr + st->ln; - if (do_utf8 && OP(scan) != REF) { /* REF can do byte comparison */ + s = PL_bostr + ln; + if (do_utf8 && type != REF) { /* REF can do byte comparison */ char *l = locinput; const char *e = PL_bostr + PL_regendp[n]; /* @@ -3284,7 +3461,7 @@ S_regmatch(pTHX_ const regmatch_info *reginfo, regnode *prog) * in the 8-bit case (no pun intended) because in Unicode we * have to map both upper and title case to lower case. */ - if (OP(scan) == REFF) { + if (type == REFF) { while (s < e) { STRLEN ulen1, ulen2; U8 tmpbuf1[UTF8_MAXBYTES_CASE+1]; @@ -3307,24 +3484,23 @@ S_regmatch(pTHX_ const regmatch_info *reginfo, regnode *prog) /* Inline the first character, for speed. */ if (UCHARAT(s) != nextchr && - (OP(scan) == REF || - (UCHARAT(s) != ((OP(scan) == REFF - ? PL_fold : PL_fold_locale)[nextchr])))) + (type == REF || + (UCHARAT(s) != (type == REFF + ? PL_fold : PL_fold_locale)[nextchr]))) sayNO; - st->ln = PL_regendp[n] - st->ln; - if (locinput + st->ln > PL_regeol) + ln = PL_regendp[n] - ln; + if (locinput + ln > PL_regeol) sayNO; - if (st->ln > 1 && (OP(scan) == REF - ? memNE(s, locinput, st->ln) - : (OP(scan) == REFF - ? ibcmp(s, locinput, st->ln) - : ibcmp_locale(s, locinput, st->ln)))) + if (ln > 1 && (type == REF + ? memNE(s, locinput, ln) + : (type == REFF + ? ibcmp(s, locinput, ln) + : ibcmp_locale(s, locinput, ln)))) sayNO; - locinput += st->ln; + locinput += ln; nextchr = UCHARAT(locinput); break; - } - + } case NOTHING: case TAIL: break; @@ -3336,25 +3512,29 @@ S_regmatch(pTHX_ const regmatch_info *reginfo, regnode *prog) { SV *ret; regexp *re; - regnode *startpoint; - - case SRECURSE: - case RECURSE: /* /(...(?1))/ */ + regexp_internal *rei; + regnode *startpoint; + + case GOSTART: + case GOSUB: /* /(...(?1))/ */ if (cur_eval && cur_eval->locinput==locinput) { - if (cur_eval->u.eval.close_paren == ARG(scan)) - Perl_croak(aTHX_ "Infinite recursion in RECURSE in regexp"); + if (cur_eval->u.eval.close_paren == (U32)ARG(scan)) + Perl_croak(aTHX_ "Infinite recursion in regex"); if ( ++nochange_depth > MAX_RECURSE_EVAL_NOCHANGE_DEPTH ) - Perl_croak(aTHX_ "RECURSE without pos change exceeded limit in regexp"); + Perl_croak(aTHX_ + "Pattern subroutine nesting without pos change" + " exceeded limit in regex"); } else { nochange_depth = 0; - } + } re = rex; + rei = rexi; (void)ReREFCNT_inc(rex); - if (OP(scan)==RECURSE) { + if (OP(scan)==GOSUB) { startpoint = scan + ARG2L(scan); ST.close_paren = ARG(scan); } else { - startpoint = re->program+1; + startpoint = rei->program+1; ST.close_paren = 0; } goto eval_recurse_doit; @@ -3362,7 +3542,7 @@ S_regmatch(pTHX_ const regmatch_info *reginfo, regnode *prog) case EVAL: /* /(?{A})B/ /(??{A})B/ and /(?(?{A})X|Y)B/ */ if (cur_eval && cur_eval->locinput==locinput) { if ( ++nochange_depth > MAX_RECURSE_EVAL_NOCHANGE_DEPTH ) - Perl_croak(aTHX_ "EVAL without pos change exceeded limit in regexp"); + Perl_croak(aTHX_ "EVAL without pos change exceeded limit in regex"); } else { nochange_depth = 0; } @@ -3375,11 +3555,17 @@ S_regmatch(pTHX_ const regmatch_info *reginfo, regnode *prog) PAD *old_comppad; n = ARG(scan); - PL_op = (OP_4tree*)rex->data->data[n]; - DEBUG_EXECUTE_r( PerlIO_printf(Perl_debug_log, " re_eval 0x%"UVxf"\n", PTR2UV(PL_op)) ); - PAD_SAVE_LOCAL(old_comppad, (PAD*)rex->data->data[n + 2]); + PL_op = (OP_4tree*)rexi->data->data[n]; + DEBUG_STATE_r( PerlIO_printf(Perl_debug_log, + " re_eval 0x%"UVxf"\n", PTR2UV(PL_op)) ); + PAD_SAVE_LOCAL(old_comppad, (PAD*)rexi->data->data[n + 2]); PL_regendp[0] = PL_reg_magic->mg_len = locinput - PL_bostr; + if (sv_yes_mark) { + SV *sv_mrk = get_sv("REGMARK", 1); + sv_setsv(sv_mrk, sv_yes_mark); + } + CALLRUNOPS(aTHX); /* Scalar context. */ SPAGAIN; if (SP == before) @@ -3392,14 +3578,14 @@ S_regmatch(pTHX_ const regmatch_info *reginfo, regnode *prog) PL_op = oop; PAD_RESTORE_LOCAL(old_comppad); PL_curcop = ocurcop; - if (!st->logical) { + if (!logical) { /* /(?{...})/ */ sv_setsv(save_scalar(PL_replgv), ret); break; } } - if (st->logical == 2) { /* Postponed subexpression: /(??{...})/ */ - + if (logical == 2) { /* Postponed subexpression: /(??{...})/ */ + logical = 0; { /* extract RE object from returned value; compiling if * necessary */ @@ -3436,12 +3622,13 @@ S_regmatch(pTHX_ const regmatch_info *reginfo, regnode *prog) PL_regsize = osize; } } + rei = RXi_GET(re); DEBUG_EXECUTE_r( debug_start_match(re, do_utf8, locinput, PL_regeol, "Matching embedded"); ); - startpoint = re->program + 1; - ST.close_paren = 0; /* only used for RECURSE */ + startpoint = rei->program + 1; + ST.close_paren = 0; /* only used for GOSUB */ /* borrowed from regtry */ if (PL_reg_start_tmpl <= re->nparens) { PL_reg_start_tmpl = re->nparens*3/2 + 3; @@ -3451,24 +3638,24 @@ S_regmatch(pTHX_ const regmatch_info *reginfo, regnode *prog) Newx(PL_reg_start_tmp, PL_reg_start_tmpl, char*); } - eval_recurse_doit: /* Share code with RECURSE below this line */ + eval_recurse_doit: /* Share code with GOSUB below this line */ /* run the pattern returned from (??{...}) */ ST.cp = regcppush(0); /* Save *all* the positions. */ REGCP_SET(ST.lastcp); - PL_regstartp = re->startp; /* essentially NOOP on RECURSE */ - PL_regendp = re->endp; /* essentially NOOP on RECURSE */ + PL_regstartp = re->startp; /* essentially NOOP on GOSUB */ + PL_regendp = re->endp; /* essentially NOOP on GOSUB */ *PL_reglastparen = 0; *PL_reglastcloseparen = 0; PL_reginput = locinput; + PL_regsize = 0; /* XXXX This is too dramatic a measure... */ PL_reg_maxiter = 0; - st->logical = 0; ST.toggle_reg_flags = PL_reg_flags; - if (re->reganch & ROPT_UTF8) + if (re->extflags & RXf_UTF8) PL_reg_flags |= RF_utf8; else PL_reg_flags &= ~RF_utf8; @@ -3477,6 +3664,7 @@ S_regmatch(pTHX_ const regmatch_info *reginfo, regnode *prog) ST.prev_rex = rex; ST.prev_curlyx = cur_curlyx; rex = re; + rexi = rei; cur_curlyx = NULL; ST.B = next; ST.prev_eval = cur_eval; @@ -3485,9 +3673,9 @@ S_regmatch(pTHX_ const regmatch_info *reginfo, regnode *prog) PUSH_YES_STATE_GOTO(EVAL_AB, startpoint); /* NOTREACHED */ } - /* /(?(?{...})X|Y)/ */ - st->sw = (bool)SvTRUE(ret); - st->logical = 0; + /* logical is 1, /(?(?{...})X|Y)/ */ + sw = (bool)SvTRUE(ret); + logical = 0; break; } @@ -3496,6 +3684,7 @@ S_regmatch(pTHX_ const regmatch_info *reginfo, regnode *prog) PL_reg_flags ^= ST.toggle_reg_flags; ReREFCNT_dec(rex); rex = ST.prev_rex; + rexi = RXi_GET(rex); regcpblow(ST.cp); cur_eval = ST.prev_eval; cur_curlyx = ST.prev_curlyx; @@ -3509,6 +3698,7 @@ S_regmatch(pTHX_ const regmatch_info *reginfo, regnode *prog) PL_reg_flags ^= ST.toggle_reg_flags; ReREFCNT_dec(rex); rex = ST.prev_rex; + rexi = RXi_GET(rex); PL_reginput = locinput; REGCP_UNWIND(ST.lastcp); regcppop(rex); @@ -3524,25 +3714,65 @@ S_regmatch(pTHX_ const regmatch_info *reginfo, regnode *prog) PL_reg_start_tmp[n] = locinput; if (n > PL_regsize) PL_regsize = n; + lastopen = n; break; case CLOSE: n = ARG(scan); /* which paren pair */ PL_regstartp[n] = PL_reg_start_tmp[n] - PL_bostr; PL_regendp[n] = locinput - PL_bostr; - if (n > (I32)*PL_reglastparen) + /*if (n > PL_regsize) + PL_regsize = n;*/ + if (n > *PL_reglastparen) *PL_reglastparen = n; *PL_reglastcloseparen = n; - if (cur_eval && cur_eval->u.eval.close_paren == (U32)n) { + if (cur_eval && cur_eval->u.eval.close_paren == n) { goto fake_end; } break; + case ACCEPT: + if (ARG(scan)){ + regnode *cursor; + for (cursor=scan; + cursor && OP(cursor)!=END; + cursor=regnext(cursor)) + { + if ( OP(cursor)==CLOSE ){ + n = ARG(cursor); + if ( n <= lastopen ) { + PL_regstartp[n] = PL_reg_start_tmp[n] - PL_bostr; + PL_regendp[n] = locinput - PL_bostr; + /*if (n > PL_regsize) + PL_regsize = n;*/ + if (n > *PL_reglastparen) + *PL_reglastparen = n; + *PL_reglastcloseparen = n; + if ( n == ARG(scan) || (cur_eval && + cur_eval->u.eval.close_paren == n)) + break; + } + } + } + } + goto fake_end; + /*NOTREACHED*/ case GROUPP: n = ARG(scan); /* which paren pair */ - st->sw = (bool)((I32)*PL_reglastparen >= n && PL_regendp[n] != -1); + sw = (bool)(*PL_reglastparen >= n && PL_regendp[n] != -1); break; + case NGROUPP: + /* reg_check_named_buff_matched returns 0 for no match */ + sw = (bool)(0 < reg_check_named_buff_matched(rex,scan)); + break; + case INSUBP: + n = ARG(scan); + sw = (cur_eval && (!n || cur_eval->u.eval.close_paren == n)); + break; + case DEFINEP: + sw = 0; + break; case IFTHEN: PL_reg_leftiter = PL_reg_maxiter; /* Void cache */ - if (st->sw) + if (sw) next = NEXTOPER(NEXTOPER(scan)); else { next = scan + ARG(scan); @@ -3551,212 +3781,206 @@ S_regmatch(pTHX_ const regmatch_info *reginfo, regnode *prog) } break; case LOGICAL: - st->logical = scan->flags; + logical = scan->flags; break; + /******************************************************************* - cur_curlyx points to the regmatch_state associated with the most recent CURLYX. - This struct contains info about the innermost (...)* loop (an - "infoblock"), and a pointer to the next outer cur_curlyx. - Here is how Y(A)*Z is processed (if it is compiled into CURLYX/WHILEM): +The CURLYX/WHILEM pair of ops handle the most generic case of the /A*B/ +pattern, where A and B are subpatterns. (For simple A, CURLYM or +STAR/PLUS/CURLY/CURLYN are used instead.) - 1) After matching Y, regnode for CURLYX is processed; +A*B is compiled as - 2) This regnode populates cur_curlyx, and calls regmatch() recursively - with the starting point at WHILEM node; +On entry to the subpattern, CURLYX is called. This pushes a CURLYX +state, which contains the current count, initialised to -1. It also sets +cur_curlyx to point to this state, with any previous value saved in the +state block. - 3) Each hit of WHILEM node tries to match A and Z (in the order - depending on the current iteration, min/max of {min,max} and - greediness). The information about where are nodes for "A" - and "Z" is read from cur_curlyx, as is info on how many times "A" - was already matched, and greediness. +CURLYX then jumps straight to the WHILEM op, rather than executing A, +since the pattern may possibly match zero times (i.e. it's a while {} loop +rather than a do {} while loop). - 4) After A matches, the same WHILEM node is hit again. +Each entry to WHILEM represents a successful match of A. The count in the +CURLYX block is incremented, another WHILEM state is pushed, and execution +passes to A or B depending on greediness and the current count. - 5) Each time WHILEM is hit, cur_curlyx is the infoblock created by CURLYX - of the same pair. Thus when WHILEM tries to match Z, it temporarily - resets cur_curlyx, since this Y(A)*Z can be a part of some other loop: - as in (Y(A)*Z)*. If Z matches, the automaton will hit the WHILEM node - of the external loop. +For example, if matching against the string a1a2a3b (where the aN are +substrings that match /A/), then the match progresses as follows: (the +pushed states are interspersed with the bits of strings matched so far): - Currently present infoblocks form a tree with a stem formed by cur_curlyx - and whatever it mentions via ->next, and additional attached trees - corresponding to temporarily unset infoblocks as in "5" above. + + + a1 + a1 a2 + a1 a2 a3 + a1 a2 a3 b - In the following picture, infoblocks for outer loop of - (Y(A)*?Z)*?T are denoted O, for inner I. NULL starting block - is denoted by x. The matched string is YAAZYAZT. Temporarily postponed - infoblocks are drawn below the "reset" infoblock. +(Contrast this with something like CURLYM, which maintains only a single +backtrack state: - In fact in the picture below we do not show failed matches for Z and T - by WHILEM blocks. [We illustrate minimal matches, since for them it is - more obvious *why* one needs to *temporary* unset infoblocks.] + a1 + a1 a2 + a1 a2 a3 + a1 a2 a3 b +) - Matched REx position InfoBlocks Comment - (Y(A)*?Z)*?T x - Y(A)*?Z)*?T x <- O - Y (A)*?Z)*?T x <- O - Y A)*?Z)*?T x <- O <- I - YA )*?Z)*?T x <- O <- I - YA A)*?Z)*?T x <- O <- I - YAA )*?Z)*?T x <- O <- I - YAA Z)*?T x <- O # Temporary unset I - I +Each WHILEM state block marks a point to backtrack to upon partial failure +of A or B, and also contains some minor state data related to that +iteration. The CURLYX block, pointed to by cur_curlyx, contains the +overall state, such as the count, and pointers to the A and B ops. - YAAZ Y(A)*?Z)*?T x <- O - I +This is complicated slightly by nested CURLYX/WHILEM's. Since cur_curlyx +must always point to the *current* CURLYX block, the rules are: - YAAZY (A)*?Z)*?T x <- O - I +When executing CURLYX, save the old cur_curlyx in the CURLYX state block, +and set cur_curlyx to point the new block. - YAAZY A)*?Z)*?T x <- O <- I - I +When popping the CURLYX block after a successful or unsuccessful match, +restore the previous cur_curlyx. - YAAZYA )*?Z)*?T x <- O <- I - I +When WHILEM is about to execute B, save the current cur_curlyx, and set it +to the outer one saved in the CURLYX block. - YAAZYA Z)*?T x <- O # Temporary unset I - I,I +When popping the WHILEM block after a successful or unsuccessful B match, +restore the previous cur_curlyx. - YAAZYAZ )*?T x <- O - I,I +Here's an example for the pattern (AI* BI)*BO +I and O refer to inner and outer, C and W refer to CURLYX and WHILEM: - YAAZYAZ T x # Temporary unset O - O - I,I +cur_ +curlyx backtrack stack +------ --------------- +NULL +CO +CI ai +CO ai bi +NULL ai bi bo - YAAZYAZT x - O - I,I - *******************************************************************/ +At this point the pattern succeeds, and we work back down the stack to +clean up, restoring as we go: - case CURLYX: { - /* No need to save/restore up to this paren */ - parenfloor = scan->flags; - - /* Dave says: - - CURLYX and WHILEM are always paired: they're the moral - equivalent of pp_enteriter anbd pp_iter. - - The only time next could be null is if the node tree is - corrupt. This was mentioned on p5p a few days ago. - - See http://www.xray.mpe.mpg.de/mailing-lists/perl5-porters/2006-04/msg00556.html - So we'll assert that this is true: - */ - assert(next); - if (OP(PREVOPER(next)) == NOTHING) /* LONGJMP */ - next += ARG(next); - /* XXXX Probably it is better to teach regpush to support - parenfloor > PL_regsize... */ - if (parenfloor > (I32)*PL_reglastparen) - parenfloor = *PL_reglastparen; /* Pessimization... */ - - st->u.curlyx.cp = PL_savestack_ix; - st->u.curlyx.outercc = cur_curlyx; - cur_curlyx = st; - /* these fields contain the state of the current curly. - * they are accessed by subsequent WHILEMs; - * cur and lastloc are also updated by WHILEM */ - st->u.curlyx.parenfloor = parenfloor; - st->u.curlyx.cur = -1; /* this will be updated by WHILEM */ - st->u.curlyx.min = ARG1(scan); - st->u.curlyx.max = ARG2(scan); - st->u.curlyx.scan = NEXTOPER(scan) + EXTRA_STEP_2ARGS; - st->u.curlyx.lastloc = 0; - /* st->next and st->minmod are also read by WHILEM */ +CO ai bi +CI ai +CO +NULL - PL_reginput = locinput; - REGMATCH(PREVOPER(next), CURLYX); /* start on the WHILEM */ - /*** all unsaved local vars undefined at this point */ - regcpblow(st->u.curlyx.cp); - cur_curlyx = st->u.curlyx.outercc; - saySAME(result); +*******************************************************************/ + +#define ST st->u.curlyx + + case CURLYX: /* start of /A*B/ (for complex A) */ + { + /* No need to save/restore up to this paren */ + I32 parenfloor = scan->flags; + + assert(next); /* keep Coverity happy */ + if (OP(PREVOPER(next)) == NOTHING) /* LONGJMP */ + next += ARG(next); + + /* XXXX Probably it is better to teach regpush to support + parenfloor > PL_regsize... */ + if (parenfloor > (I32)*PL_reglastparen) + parenfloor = *PL_reglastparen; /* Pessimization... */ + + ST.prev_curlyx= cur_curlyx; + cur_curlyx = st; + ST.cp = PL_savestack_ix; + + /* these fields contain the state of the current curly. + * they are accessed by subsequent WHILEMs */ + ST.parenfloor = parenfloor; + ST.min = ARG1(scan); + ST.max = ARG2(scan); + ST.A = NEXTOPER(scan) + EXTRA_STEP_2ARGS; + ST.B = next; + ST.minmod = minmod; + minmod = 0; + ST.count = -1; /* this will be updated by WHILEM */ + ST.lastloc = NULL; /* this will be updated by WHILEM */ + + PL_reginput = locinput; + PUSH_YES_STATE_GOTO(CURLYX_end, PREVOPER(next)); + /* NOTREACHED */ + } + + case CURLYX_end: /* just finished matching all of A*B */ + if (PL_reg_eval_set){ + SV *pres= GvSV(PL_replgv); + SvREFCNT_inc(pres); + regcpblow(ST.cp); + sv_setsv(GvSV(PL_replgv), pres); + SvREFCNT_dec(pres); + } else { + regcpblow(ST.cp); } + cur_curlyx = ST.prev_curlyx; + sayYES; /* NOTREACHED */ - case WHILEM: { - /* - * This is really hard to understand, because after we match - * what we're trying to match, we must make sure the rest of - * the REx is going to match for sure, and to do that we have - * to go back UP the parse tree by recursing ever deeper. And - * if it fails, we have to reset our parent's current state - * that we can try again after backing off. - */ - /* Dave says: - - cur_curlyx gets initialised by CURLYX ready for use by WHILEM. - So again, unless somethings been corrupted, cur_curlyx cannot - be null at that point in WHILEM. - - See http://www.xray.mpe.mpg.de/mailing-lists/perl5-porters/2006-04/msg00556.html - So we'll assert that this is true: - */ - assert(cur_curlyx); - st->u.whilem.lastloc = cur_curlyx->u.curlyx.lastloc; /* Detection of 0-len. */ - st->u.whilem.cache_offset = 0; - st->u.whilem.cache_bit = 0; - - n = cur_curlyx->u.curlyx.cur + 1; /* how many we know we matched */ - PL_reginput = locinput; + case CURLYX_end_fail: /* just failed to match all of A*B */ + regcpblow(ST.cp); + cur_curlyx = ST.prev_curlyx; + sayNO; + /* NOTREACHED */ - DEBUG_EXECUTE_r( - PerlIO_printf(Perl_debug_log, - "%*s %ld out of %ld..%ld cc=%"UVxf"\n", - REPORT_CODE_OFF+depth*2, "", - (long)n, (long)cur_curlyx->u.curlyx.min, - (long)cur_curlyx->u.curlyx.max, - PTR2UV(cur_curlyx)) - ); - /* If degenerate scan matches "", assume scan done. */ +#undef ST +#define ST st->u.whilem - if (locinput == cur_curlyx->u.curlyx.lastloc && n >= - cur_curlyx->u.curlyx.min) - { - st->u.whilem.savecc = cur_curlyx; - cur_curlyx = cur_curlyx->u.curlyx.outercc; - DEBUG_EXECUTE_r( - PerlIO_printf(Perl_debug_log, - "%*s empty match detected, try continuation...\n", - REPORT_CODE_OFF+depth*2, "") - ); - REGMATCH(st->u.whilem.savecc->next, WHILEM1); - /*** all unsaved local vars undefined at this point */ - cur_curlyx = st->u.whilem.savecc; - if (result) - sayYES; - sayNO; - } + case WHILEM: /* just matched an A in /A*B/ (for complex A) */ + { + /* see the discussion above about CURLYX/WHILEM */ + I32 n; + assert(cur_curlyx); /* keep Coverity happy */ + n = ++cur_curlyx->u.curlyx.count; /* how many A's matched */ + ST.save_lastloc = cur_curlyx->u.curlyx.lastloc; + ST.cache_offset = 0; + ST.cache_mask = 0; + + PL_reginput = locinput; - /* First just match a string of min scans. */ - - if (n < cur_curlyx->u.curlyx.min) { - cur_curlyx->u.curlyx.cur = n; - cur_curlyx->u.curlyx.lastloc = locinput; - REGMATCH(cur_curlyx->u.curlyx.scan, WHILEM2); - /*** all unsaved local vars undefined at this point */ - if (result) - sayYES; - cur_curlyx->u.curlyx.cur = n - 1; - cur_curlyx->u.curlyx.lastloc = st->u.whilem.lastloc; - sayNO; - } + DEBUG_EXECUTE_r( PerlIO_printf(Perl_debug_log, + "%*s whilem: matched %ld out of %ld..%ld\n", + REPORT_CODE_OFF+depth*2, "", (long)n, + (long)cur_curlyx->u.curlyx.min, + (long)cur_curlyx->u.curlyx.max) + ); + + /* First just match a string of min A's. */ + + if (n < cur_curlyx->u.curlyx.min) { + cur_curlyx->u.curlyx.lastloc = locinput; + PUSH_STATE_GOTO(WHILEM_A_pre, cur_curlyx->u.curlyx.A); + /* NOTREACHED */ + } + + /* If degenerate A matches "", assume A done. */ + + if (locinput == cur_curlyx->u.curlyx.lastloc) { + DEBUG_EXECUTE_r( PerlIO_printf(Perl_debug_log, + "%*s whilem: empty match detected, trying continuation...\n", + REPORT_CODE_OFF+depth*2, "") + ); + goto do_whilem_B_max; + } + + /* super-linear cache processing */ + + if (scan->flags) { - if (scan->flags) { - /* Check whether we already were at this position. - Postpone detection until we know the match is not - *that* much linear. */ if (!PL_reg_maxiter) { + /* start the countdown: Postpone detection until we + * know the match is not *that* much linear. */ PL_reg_maxiter = (PL_regeol - PL_bostr + 1) * (scan->flags>>4); /* possible overflow for long strings and many CURLYX's */ if (PL_reg_maxiter < 0) PL_reg_maxiter = I32_MAX; PL_reg_leftiter = PL_reg_maxiter; } + if (PL_reg_leftiter-- == 0) { + /* initialise cache */ const I32 size = (PL_reg_maxiter + 7)/8; if (PL_reg_poscache) { if ((I32)PL_reg_poscache_size < size) { @@ -3769,124 +3993,133 @@ S_regmatch(pTHX_ const regmatch_info *reginfo, regnode *prog) PL_reg_poscache_size = size; Newxz(PL_reg_poscache, size, char); } - DEBUG_EXECUTE_r( - PerlIO_printf(Perl_debug_log, - "%sDetected a super-linear match, switching on caching%s...\n", - PL_colors[4], PL_colors[5]) - ); + DEBUG_EXECUTE_r( PerlIO_printf(Perl_debug_log, + "%swhilem: Detected a super-linear match, switching on caching%s...\n", + PL_colors[4], PL_colors[5]) + ); } + if (PL_reg_leftiter < 0) { - st->u.whilem.cache_offset = locinput - PL_bostr; - - st->u.whilem.cache_offset = (scan->flags & 0xf) - 1 - + st->u.whilem.cache_offset * (scan->flags>>4); - st->u.whilem.cache_bit = st->u.whilem.cache_offset % 8; - st->u.whilem.cache_offset /= 8; - if (PL_reg_poscache[st->u.whilem.cache_offset] & (1<u.whilem.cache_bit)) { - DEBUG_EXECUTE_r( - PerlIO_printf(Perl_debug_log, - "%*s already tried at this position...\n", - REPORT_CODE_OFF+depth*2, "") + /* have we already failed at this position? */ + I32 offset, mask; + offset = (scan->flags & 0xf) - 1 + + (locinput - PL_bostr) * (scan->flags>>4); + mask = 1 << (offset % 8); + offset /= 8; + if (PL_reg_poscache[offset] & mask) { + DEBUG_EXECUTE_r( PerlIO_printf(Perl_debug_log, + "%*s whilem: (cache) already tried at this position...\n", + REPORT_CODE_OFF+depth*2, "") ); sayNO; /* cache records failure */ } + ST.cache_offset = offset; + ST.cache_mask = mask; } - } + } - /* Prefer next over scan for minimal matching. */ - - if (cur_curlyx->minmod) { - st->u.whilem.savecc = cur_curlyx; - cur_curlyx = cur_curlyx->u.curlyx.outercc; - st->u.whilem.cp = regcppush(st->u.whilem.savecc->u.curlyx.parenfloor); - REGCP_SET(st->u.whilem.lastcp); - REGMATCH(st->u.whilem.savecc->next, WHILEM3); - /*** all unsaved local vars undefined at this point */ - cur_curlyx = st->u.whilem.savecc; - if (result) { - regcpblow(st->u.whilem.cp); - sayYES; /* All done. */ - } - REGCP_UNWIND(st->u.whilem.lastcp); - regcppop(rex); - - if (n >= cur_curlyx->u.curlyx.max) { /* Maximum greed exceeded? */ - if (ckWARN(WARN_REGEXP) && n >= REG_INFTY - && !(PL_reg_flags & RF_warned)) { - PL_reg_flags |= RF_warned; - Perl_warner(aTHX_ packWARN(WARN_REGEXP), "%s limit (%d) exceeded", - "Complex regular subexpression recursion", - REG_INFTY - 1); - } - CACHEsayNO; - } + /* Prefer B over A for minimal matching. */ - DEBUG_EXECUTE_r( - PerlIO_printf(Perl_debug_log, - "%*s trying longer...\n", - REPORT_CODE_OFF+depth*2, "") - ); - /* Try scanning more and see if it helps. */ - PL_reginput = locinput; - cur_curlyx->u.curlyx.cur = n; - cur_curlyx->u.curlyx.lastloc = locinput; - st->u.whilem.cp = regcppush(cur_curlyx->u.curlyx.parenfloor); - REGCP_SET(st->u.whilem.lastcp); - REGMATCH(cur_curlyx->u.curlyx.scan, WHILEM4); - /*** all unsaved local vars undefined at this point */ - if (result) { - regcpblow(st->u.whilem.cp); - sayYES; - } - REGCP_UNWIND(st->u.whilem.lastcp); - regcppop(rex); - cur_curlyx->u.curlyx.cur = n - 1; - cur_curlyx->u.curlyx.lastloc = st->u.whilem.lastloc; - CACHEsayNO; - } + if (cur_curlyx->u.curlyx.minmod) { + ST.save_curlyx = cur_curlyx; + cur_curlyx = cur_curlyx->u.curlyx.prev_curlyx; + ST.cp = regcppush(ST.save_curlyx->u.curlyx.parenfloor); + REGCP_SET(ST.lastcp); + PUSH_YES_STATE_GOTO(WHILEM_B_min, ST.save_curlyx->u.curlyx.B); + /* NOTREACHED */ + } - /* Prefer scan over next for maximal matching. */ - - if (n < cur_curlyx->u.curlyx.max) { /* More greed allowed? */ - st->u.whilem.cp = regcppush(cur_curlyx->u.curlyx.parenfloor); - cur_curlyx->u.curlyx.cur = n; - cur_curlyx->u.curlyx.lastloc = locinput; - REGCP_SET(st->u.whilem.lastcp); - REGMATCH(cur_curlyx->u.curlyx.scan, WHILEM5); - /*** all unsaved local vars undefined at this point */ - if (result) { - regcpblow(st->u.whilem.cp); - sayYES; - } - REGCP_UNWIND(st->u.whilem.lastcp); - regcppop(rex); /* Restore some previous $s? */ - PL_reginput = locinput; - DEBUG_EXECUTE_r( - PerlIO_printf(Perl_debug_log, - "%*s failed, try continuation...\n", - REPORT_CODE_OFF+depth*2, "") - ); - } - if (ckWARN(WARN_REGEXP) && n >= REG_INFTY - && !(PL_reg_flags & RF_warned)) { + /* Prefer A over B for maximal matching. */ + + if (n < cur_curlyx->u.curlyx.max) { /* More greed allowed? */ + ST.cp = regcppush(cur_curlyx->u.curlyx.parenfloor); + cur_curlyx->u.curlyx.lastloc = locinput; + REGCP_SET(ST.lastcp); + PUSH_STATE_GOTO(WHILEM_A_max, cur_curlyx->u.curlyx.A); + /* NOTREACHED */ + } + goto do_whilem_B_max; + } + /* NOTREACHED */ + + case WHILEM_B_min: /* just matched B in a minimal match */ + case WHILEM_B_max: /* just matched B in a maximal match */ + cur_curlyx = ST.save_curlyx; + sayYES; + /* NOTREACHED */ + + case WHILEM_B_max_fail: /* just failed to match B in a maximal match */ + cur_curlyx = ST.save_curlyx; + cur_curlyx->u.curlyx.lastloc = ST.save_lastloc; + cur_curlyx->u.curlyx.count--; + CACHEsayNO; + /* NOTREACHED */ + + case WHILEM_A_min_fail: /* just failed to match A in a minimal match */ + REGCP_UNWIND(ST.lastcp); + regcppop(rex); + /* FALL THROUGH */ + case WHILEM_A_pre_fail: /* just failed to match even minimal A */ + cur_curlyx->u.curlyx.lastloc = ST.save_lastloc; + cur_curlyx->u.curlyx.count--; + CACHEsayNO; + /* NOTREACHED */ + + case WHILEM_A_max_fail: /* just failed to match A in a maximal match */ + REGCP_UNWIND(ST.lastcp); + regcppop(rex); /* Restore some previous $s? */ + PL_reginput = locinput; + DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, + "%*s whilem: failed, trying continuation...\n", + REPORT_CODE_OFF+depth*2, "") + ); + do_whilem_B_max: + if (cur_curlyx->u.curlyx.count >= REG_INFTY + && ckWARN(WARN_REGEXP) + && !(PL_reg_flags & RF_warned)) + { + PL_reg_flags |= RF_warned; + Perl_warner(aTHX_ packWARN(WARN_REGEXP), "%s limit (%d) exceeded", + "Complex regular subexpression recursion", + REG_INFTY - 1); + } + + /* now try B */ + ST.save_curlyx = cur_curlyx; + cur_curlyx = cur_curlyx->u.curlyx.prev_curlyx; + PUSH_YES_STATE_GOTO(WHILEM_B_max, ST.save_curlyx->u.curlyx.B); + /* NOTREACHED */ + + case WHILEM_B_min_fail: /* just failed to match B in a minimal match */ + cur_curlyx = ST.save_curlyx; + REGCP_UNWIND(ST.lastcp); + regcppop(rex); + + if (cur_curlyx->u.curlyx.count >= cur_curlyx->u.curlyx.max) { + /* Maximum greed exceeded */ + if (cur_curlyx->u.curlyx.count >= REG_INFTY + && ckWARN(WARN_REGEXP) + && !(PL_reg_flags & RF_warned)) + { PL_reg_flags |= RF_warned; - Perl_warner(aTHX_ packWARN(WARN_REGEXP), "%s limit (%d) exceeded", - "Complex regular subexpression recursion", - REG_INFTY - 1); + Perl_warner(aTHX_ packWARN(WARN_REGEXP), + "%s limit (%d) exceeded", + "Complex regular subexpression recursion", + REG_INFTY - 1); } - - /* Failed deeper matches of scan, so see if this one works. */ - st->u.whilem.savecc = cur_curlyx; - cur_curlyx = cur_curlyx->u.curlyx.outercc; - REGMATCH(st->u.whilem.savecc->next, WHILEM6); - /*** all unsaved local vars undefined at this point */ - cur_curlyx = st->u.whilem.savecc; - if (result) - sayYES; - cur_curlyx->u.curlyx.cur = n - 1; - cur_curlyx->u.curlyx.lastloc = st->u.whilem.lastloc; + cur_curlyx->u.curlyx.count--; CACHEsayNO; } + + DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, + "%*s trying longer...\n", REPORT_CODE_OFF+depth*2, "") + ); + /* Try grabbing another A and see if it helps. */ + PL_reginput = locinput; + cur_curlyx->u.curlyx.lastloc = locinput; + ST.cp = regcppush(cur_curlyx->u.curlyx.parenfloor); + REGCP_SET(ST.lastcp); + PUSH_STATE_GOTO(WHILEM_A_min, ST.save_curlyx->u.curlyx.A); /* NOTREACHED */ #undef ST @@ -3901,32 +4134,67 @@ S_regmatch(pTHX_ const regmatch_info *reginfo, regnode *prog) case BRANCH: /* /(...|A|...)/ */ scan = NEXTOPER(scan); /* scan now points to inner node */ - if (!next || (OP(next) != BRANCH && OP(next) != BRANCHJ)) + if ((!next || (OP(next) != BRANCH && OP(next) != BRANCHJ)) + && !has_cutgroup) + { /* last branch; skip state push and jump direct to node */ continue; + } ST.lastparen = *PL_reglastparen; ST.next_branch = next; REGCP_SET(ST.cp); PL_reginput = locinput; /* Now go into the branch */ - PUSH_STATE_GOTO(BRANCH_next, scan); + if (has_cutgroup) { + PUSH_YES_STATE_GOTO(BRANCH_next, scan); + } else { + PUSH_STATE_GOTO(BRANCH_next, scan); + } /* NOTREACHED */ - + case CUTGROUP: + PL_reginput = locinput; + sv_yes_mark = st->u.mark.mark_name = scan->flags ? NULL : + (SV*)rexi->data->data[ ARG( scan ) ]; + PUSH_STATE_GOTO(CUTGROUP_next,next); + /* NOTREACHED */ + case CUTGROUP_next_fail: + do_cutgroup = 1; + no_final = 1; + if (st->u.mark.mark_name) + sv_commit = st->u.mark.mark_name; + sayNO; + /* NOTREACHED */ + case BRANCH_next: + sayYES; + /* NOTREACHED */ case BRANCH_next_fail: /* that branch failed; try the next, if any */ + if (do_cutgroup) { + do_cutgroup = 0; + no_final = 0; + } REGCP_UNWIND(ST.cp); for (n = *PL_reglastparen; n > ST.lastparen; n--) PL_regendp[n] = -1; *PL_reglastparen = n; + /*dmq: *PL_reglastcloseparen = n; */ scan = ST.next_branch; /* no more branches? */ - if (!scan || (OP(scan) != BRANCH && OP(scan) != BRANCHJ)) - sayNO; + if (!scan || (OP(scan) != BRANCH && OP(scan) != BRANCHJ)) { + DEBUG_EXECUTE_r({ + PerlIO_printf( Perl_debug_log, + "%*s %sBRANCH failed...%s\n", + REPORT_CODE_OFF+depth*2, "", + PL_colors[4], + PL_colors[5] ); + }); + sayNO_SILENT; + } continue; /* execute next BRANCH[J] op */ /* NOTREACHED */ case MINMOD: - st->minmod = 1; + minmod = 1; break; #undef ST @@ -3945,10 +4213,10 @@ S_regmatch(pTHX_ const regmatch_info *reginfo, regnode *prog) /* if paren positive, emulate an OPEN/CLOSE around A */ if (ST.me->flags) { - I32 paren = ST.me->flags; + U32 paren = ST.me->flags; if (paren > PL_regsize) PL_regsize = paren; - if (paren > (I32)*PL_reglastparen) + if (paren > *PL_reglastparen) *PL_reglastparen = paren; scan += NEXT_OFF(scan); /* Skip former OPEN. */ } @@ -3956,8 +4224,8 @@ S_regmatch(pTHX_ const regmatch_info *reginfo, regnode *prog) ST.B = next; ST.alen = 0; ST.count = 0; - ST.minmod = st->minmod; - st->minmod = 0; + ST.minmod = minmod; + minmod = 0; ST.c1 = CHRTEST_UNINIT; REGCP_SET(ST.cp); @@ -3997,13 +4265,21 @@ S_regmatch(pTHX_ const regmatch_info *reginfo, regnode *prog) ); locinput = PL_reginput; - if (ST.count < (ST.minmod ? ARG1(ST.me) : ARG2(ST.me))) + + if (cur_eval && cur_eval->u.eval.close_paren && + cur_eval->u.eval.close_paren == (U32)ST.me->flags) + goto fake_end; + + if ( ST.count < (ST.minmod ? ARG1(ST.me) : ARG2(ST.me)) ) goto curlym_do_A; /* try to match another A */ goto curlym_do_B; /* try to match B */ case CURLYM_A_fail: /* just failed to match an A */ REGCP_UNWIND(ST.cp); - if (ST.minmod || ST.count < ARG1(ST.me) /* min*/ ) + + if (ST.minmod || ST.count < ARG1(ST.me) /* min*/ + || (cur_eval && cur_eval->u.eval.close_paren && + cur_eval->u.eval.close_paren == (U32)ST.me->flags)) sayNO; curlym_do_B: /* execute the B in /A{m,n}B/ */ @@ -4052,10 +4328,20 @@ S_regmatch(pTHX_ const regmatch_info *reginfo, regnode *prog) PL_regstartp[paren] = HOPc(PL_reginput, -ST.alen) - PL_bostr; PL_regendp[paren] = PL_reginput - PL_bostr; + /*dmq: *PL_reglastcloseparen = paren; */ } else PL_regendp[paren] = -1; + if (cur_eval && cur_eval->u.eval.close_paren && + cur_eval->u.eval.close_paren == (U32)ST.me->flags) + { + if (ST.count) + goto fake_end; + else + sayNO; + } } + PUSH_STATE_GOTO(CURLYM_B, ST.B); /* match B */ /* NOTREACHED */ @@ -4081,6 +4367,7 @@ S_regmatch(pTHX_ const regmatch_info *reginfo, regnode *prog) if (success) { \ PL_regstartp[paren] = HOPc(locinput, -1) - PL_bostr; \ PL_regendp[paren] = locinput - PL_bostr; \ + *PL_reglastcloseparen = paren; \ } \ else \ PL_regendp[paren] = -1; \ @@ -4102,10 +4389,15 @@ S_regmatch(pTHX_ const regmatch_info *reginfo, regnode *prog) ST.paren = scan->flags; /* Which paren to set */ if (ST.paren > PL_regsize) PL_regsize = ST.paren; - if (ST.paren > (I32)*PL_reglastparen) + if (ST.paren > *PL_reglastparen) *PL_reglastparen = ST.paren; ST.min = ARG1(scan); /* min to match */ ST.max = ARG2(scan); /* max to match */ + if (cur_eval && cur_eval->u.eval.close_paren && + cur_eval->u.eval.close_paren == (U32)ST.paren) { + ST.min=1; + ST.max=1; + } scan = regnext(NEXTOPER(scan) + NODE_STEP_REGNODE); goto repeat; case CURLY: /* /A{m,n}B/ where A is width 1 */ @@ -4185,9 +4477,9 @@ S_regmatch(pTHX_ const regmatch_info *reginfo, regnode *prog) ST.A = scan; ST.B = next; PL_reginput = locinput; - if (st->minmod) { - st->minmod = 0; - if (ST.min && regrepeat(rex, ST.A, ST.min) < ST.min) + if (minmod) { + minmod = 0; + if (ST.min && regrepeat(rex, ST.A, ST.min, depth) < ST.min) sayNO; ST.count = ST.min; locinput = PL_reginput; @@ -4220,7 +4512,7 @@ S_regmatch(pTHX_ const regmatch_info *reginfo, regnode *prog) } else { - ST.count = regrepeat(rex, ST.A, ST.max); + ST.count = regrepeat(rex, ST.A, ST.max, depth); locinput = PL_reginput; if (ST.count < ST.min) sayNO; @@ -4306,11 +4598,15 @@ S_regmatch(pTHX_ const regmatch_info *reginfo, regnode *prog) /* PL_reginput == oldloc now */ if (n) { ST.count += n; - if (regrepeat(rex, ST.A, n) < n) + if (regrepeat(rex, ST.A, n, depth) < n) sayNO; } PL_reginput = locinput; CURLY_SETPAREN(ST.paren, ST.count); + if (cur_eval && cur_eval->u.eval.close_paren && + cur_eval->u.eval.close_paren == (U32)ST.paren) { + goto fake_end; + } PUSH_STATE_GOTO(CURLY_B_min_known, ST.B); } /* NOTREACHED */ @@ -4324,7 +4620,7 @@ S_regmatch(pTHX_ const regmatch_info *reginfo, regnode *prog) REGCP_UNWIND(ST.cp); /* failed -- move forward one */ PL_reginput = locinput; - if (regrepeat(rex, ST.A, 1)) { + if (regrepeat(rex, ST.A, 1, depth)) { ST.count++; locinput = PL_reginput; if (ST.count <= ST.max || (ST.max == REG_INFTY && @@ -4332,6 +4628,10 @@ S_regmatch(pTHX_ const regmatch_info *reginfo, regnode *prog) { curly_try_B_min: CURLY_SETPAREN(ST.paren, ST.count); + if (cur_eval && cur_eval->u.eval.close_paren && + cur_eval->u.eval.close_paren == (U32)ST.paren) { + goto fake_end; + } PUSH_STATE_GOTO(CURLY_B_min, ST.B); } } @@ -4341,6 +4641,10 @@ S_regmatch(pTHX_ const regmatch_info *reginfo, regnode *prog) curly_try_B_max: /* a successful greedy match: now try to match B */ + if (cur_eval && cur_eval->u.eval.close_paren && + cur_eval->u.eval.close_paren == (U32)ST.paren) { + goto fake_end; + } { UV c = 0; if (ST.c1 != CHRTEST_VOID) @@ -4369,7 +4673,6 @@ S_regmatch(pTHX_ const regmatch_info *reginfo, regnode *prog) #undef ST - case END: fake_end: if (cur_eval) { @@ -4382,7 +4685,8 @@ S_regmatch(pTHX_ const regmatch_info *reginfo, regnode *prog) PL_reg_flags ^= st->u.eval.toggle_reg_flags; st->u.eval.prev_rex = rex; /* inner */ - rex = cur_eval->u.eval.prev_rex; /* outer */ + rex = cur_eval->u.eval.prev_rex; /* outer */ + rexi = RXi_GET(rex); cur_curlyx = cur_eval->u.eval.prev_curlyx; ReREFCNT_inc(rex); st->u.eval.cp = regcppush(0); /* Save *all* the positions. */ @@ -4399,8 +4703,8 @@ S_regmatch(pTHX_ const regmatch_info *reginfo, regnode *prog) st->u.eval.prev_eval = cur_eval; cur_eval = cur_eval->u.eval.prev_eval; DEBUG_EXECUTE_r( - PerlIO_printf(Perl_debug_log, "%*s EVAL trying tail ... %x\n", - REPORT_CODE_OFF+depth*2, "",(int)cur_eval);); + PerlIO_printf(Perl_debug_log, "%*s EVAL trying tail ... %"UVxf"\n", + REPORT_CODE_OFF+depth*2, "",PTR2UV(cur_eval));); PUSH_YES_STATE_GOTO(EVAL_AB, st->u.eval.prev_eval->u.eval.B); /* match B */ } @@ -4412,6 +4716,7 @@ S_regmatch(pTHX_ const regmatch_info *reginfo, regnode *prog) (long)(locinput - PL_reg_starttry), (long)(reginfo->till - PL_reg_starttry), PL_colors[5])); + sayNO_SILENT; /* Cannot match: too short. */ } PL_reginput = locinput; /* put where regtry can find it */ @@ -4444,9 +4749,9 @@ S_regmatch(pTHX_ const regmatch_info *reginfo, regnode *prog) char * const s = HOPBACKc(locinput, scan->flags); if (!s) { /* trivial fail */ - if (st->logical) { - st->logical = 0; - st->sw = 1 - (bool)ST.wanted; + if (logical) { + logical = 0; + sw = 1 - (bool)ST.wanted; } else if (ST.wanted) sayNO; @@ -4462,6 +4767,7 @@ S_regmatch(pTHX_ const regmatch_info *reginfo, regnode *prog) do_ifmatch: ST.me = scan; + ST.logical = logical; /* execute body of (?...A) */ PUSH_YES_STATE_GOTO(IFMATCH_A, NEXTOPER(NEXTOPER(scan))); /* NOTREACHED */ @@ -4471,9 +4777,8 @@ S_regmatch(pTHX_ const regmatch_info *reginfo, regnode *prog) /* FALL THROUGH */ case IFMATCH_A: /* body of (?...A) succeeded */ - if (st->logical) { - st->logical = 0; - st->sw = (bool)ST.wanted; + if (ST.logical) { + sw = (bool)ST.wanted; } else if (!ST.wanted) sayNO; @@ -4496,14 +4801,111 @@ S_regmatch(pTHX_ const regmatch_info *reginfo, regnode *prog) if (next == scan) next = NULL; break; + case COMMIT: + reginfo->cutpoint = PL_regeol; + /* FALLTHROUGH */ + case PRUNE: + PL_reginput = locinput; + if (!scan->flags) + sv_yes_mark = sv_commit = (SV*)rexi->data->data[ ARG( scan ) ]; + PUSH_STATE_GOTO(COMMIT_next,next); + /* NOTREACHED */ + case COMMIT_next_fail: + no_final = 1; + /* FALLTHROUGH */ + case OPFAIL: + sayNO; + /* NOTREACHED */ + +#define ST st->u.mark + case MARKPOINT: + ST.prev_mark = mark_state; + ST.mark_name = sv_commit = sv_yes_mark + = (SV*)rexi->data->data[ ARG( scan ) ]; + mark_state = st; + ST.mark_loc = PL_reginput = locinput; + PUSH_YES_STATE_GOTO(MARKPOINT_next,next); + /* NOTREACHED */ + case MARKPOINT_next: + mark_state = ST.prev_mark; + sayYES; + /* NOTREACHED */ + case MARKPOINT_next_fail: + if (popmark && sv_eq(ST.mark_name,popmark)) + { + if (ST.mark_loc > startpoint) + reginfo->cutpoint = HOPBACKc(ST.mark_loc, 1); + popmark = NULL; /* we found our mark */ + sv_commit = ST.mark_name; + + DEBUG_EXECUTE_r({ + PerlIO_printf(Perl_debug_log, + "%*s %ssetting cutpoint to mark:%"SVf"...%s\n", + REPORT_CODE_OFF+depth*2, "", + PL_colors[4], sv_commit, PL_colors[5]); + }); + } + mark_state = ST.prev_mark; + sv_yes_mark = mark_state ? + mark_state->u.mark.mark_name : NULL; + sayNO; + /* NOTREACHED */ + case SKIP: + PL_reginput = locinput; + if (scan->flags) { + /* (*SKIP) : if we fail we cut here*/ + ST.mark_name = NULL; + ST.mark_loc = locinput; + PUSH_STATE_GOTO(SKIP_next,next); + } else { + /* (*SKIP:NAME) : if there is a (*MARK:NAME) fail where it was, + otherwise do nothing. Meaning we need to scan + */ + regmatch_state *cur = mark_state; + SV *find = (SV*)rexi->data->data[ ARG( scan ) ]; + + while (cur) { + if ( sv_eq( cur->u.mark.mark_name, + find ) ) + { + ST.mark_name = find; + PUSH_STATE_GOTO( SKIP_next, next ); + } + cur = cur->u.mark.prev_mark; + } + } + /* Didn't find our (*MARK:NAME) so ignore this (*SKIP:NAME) */ + break; + case SKIP_next_fail: + if (ST.mark_name) { + /* (*CUT:NAME) - Set up to search for the name as we + collapse the stack*/ + popmark = ST.mark_name; + } else { + /* (*CUT) - No name, we cut here.*/ + if (ST.mark_loc > startpoint) + reginfo->cutpoint = HOPBACKc(ST.mark_loc, 1); + /* but we set sv_commit to latest mark_name if there + is one so they can test to see how things lead to this + cut */ + if (mark_state) + sv_commit=mark_state->u.mark.mark_name; + } + no_final = 1; + sayNO; + /* NOTREACHED */ +#undef ST + default: PerlIO_printf(Perl_error_log, "%"UVxf" %d\n", PTR2UV(scan), OP(scan)); Perl_croak(aTHX_ "regexp memory corruption"); - } + + } /* end switch */ - scan = next; - continue; + /* switch break jumps here */ + scan = next; /* prepare to execute the next op and ... */ + continue; /* ... jump back to the top, reusing st */ /* NOTREACHED */ push_yes_state: @@ -4516,18 +4918,33 @@ S_regmatch(pTHX_ const regmatch_info *reginfo, regnode *prog) { regmatch_state *newst; - DEBUG_STATE_pp("push"); + DEBUG_STACK_r({ + regmatch_state *cur = st; + regmatch_state *curyes = yes_state; + int curd = depth; + regmatch_slab *slab = PL_regmatch_slab; + for (;curd > -1;cur--,curd--) { + if (cur < SLAB_FIRST(slab)) { + slab = slab->prev; + cur = SLAB_LAST(slab); + } + PerlIO_printf(Perl_error_log, "%*s#%-3d %-10s %s\n", + REPORT_CODE_OFF + 2 + depth * 2,"", + curd, reg_name[cur->resume_state], + (curyes == cur) ? "yes" : "" + ); + if (curyes == cur) + curyes = cur->u.yes.prev_yes_state; + } + } else + DEBUG_STATE_pp("push") + ); depth++; st->locinput = locinput; newst = st+1; if (newst > SLAB_LAST(PL_regmatch_slab)) newst = S_push_slab(aTHX); PL_regmatch_state = newst; - /* XXX probably don't need to initialise these */ - newst->minmod = 0; - newst->sw = 0; - newst->logical = 0; - locinput = PL_reginput; nextchr = UCHARAT(locinput); @@ -4535,39 +4952,6 @@ S_regmatch(pTHX_ const regmatch_info *reginfo, regnode *prog) continue; /* NOTREACHED */ } - - /* simulate recursively calling regmatch(), but without actually - * recursing - ie save the current state on the heap rather than on - * the stack, then re-enter the loop. This avoids complex regexes - * blowing the processor stack */ - - start_recurse: - { - /* push new state */ - regmatch_state *oldst = st; - - DEBUG_STATE_pp("push"); - depth++; - st->u.yes.prev_yes_state = yes_state; - yes_state = st; - - /* grab the next free state slot */ - st++; - if (st > SLAB_LAST(PL_regmatch_slab)) - st = S_push_slab(aTHX); - PL_regmatch_state = st; - - oldst->next = next; - oldst->n = n; - oldst->locinput = locinput; - - locinput = PL_reginput; - nextchr = UCHARAT(locinput); - st->minmod = 0; - st->sw = 0; - st->logical = 0; - - } } /* @@ -4590,7 +4974,13 @@ yes: PL_regmatch_slab = PL_regmatch_slab->prev; st = SLAB_LAST(PL_regmatch_slab); } - DEBUG_STATE_pp("pop (yes)"); + DEBUG_STATE_r({ + if (no_final) { + DEBUG_STATE_pp("pop (no final)"); + } else { + DEBUG_STATE_pp("pop (yes)"); + } + }); depth--; } #else @@ -4607,53 +4997,13 @@ yes: st = yes_state; yes_state = st->u.yes.prev_yes_state; PL_regmatch_state = st; - - switch (st->resume_state) { - case resume_CURLYX: - case resume_WHILEM1: - case resume_WHILEM2: - case resume_WHILEM3: - case resume_WHILEM4: - case resume_WHILEM5: - case resume_WHILEM6: - result = 1; - /* restore previous state and re-enter */ - scan = st->scan; - next = st->next; - n = st->n; - locinput= st->locinput; - nextchr = UCHARAT(locinput); - switch (st->resume_state) { - case resume_CURLYX: - goto resume_point_CURLYX; - case resume_WHILEM1: - goto resume_point_WHILEM1; - case resume_WHILEM2: - goto resume_point_WHILEM2; - case resume_WHILEM3: - goto resume_point_WHILEM3; - case resume_WHILEM4: - goto resume_point_WHILEM4; - case resume_WHILEM5: - goto resume_point_WHILEM5; - case resume_WHILEM6: - goto resume_point_WHILEM6; - } - Perl_croak(aTHX_ "unexpected whilem resume state"); - - case EVAL_AB: - case IFMATCH_A: - case CURLYM_A: - state_num = st->resume_state; - goto reenter_switch; - - case CURLYM_B: - case BRANCH_next: - case TRIE_next: - case CURLY_B_max: - default: - Perl_croak(aTHX_ "unexpected yes resume state"); - } + + if (no_final) { + locinput= st->locinput; + nextchr = UCHARAT(locinput); + } + state_num = st->resume_state + no_final; + goto reenter_switch; } DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "%sMatch successful!%s\n", @@ -4671,8 +5021,13 @@ no: ); no_silent: - result = 0; - + if (no_final) { + if (yes_state) { + goto yes; + } else { + goto final_exit; + } + } if (depth) { /* there's a previous state to backtrack to */ st--; @@ -4681,9 +5036,6 @@ no_silent: st = SLAB_LAST(PL_regmatch_slab); } PL_regmatch_state = st; - scan = st->scan; - next = st->next; - n = st->n; locinput= st->locinput; nextchr = UCHARAT(locinput); @@ -4692,41 +5044,27 @@ no_silent: if (yes_state == st) yes_state = st->u.yes.prev_yes_state; - switch (st->resume_state) { - case resume_CURLYX: - goto resume_point_CURLYX; - case resume_WHILEM1: - goto resume_point_WHILEM1; - case resume_WHILEM2: - goto resume_point_WHILEM2; - case resume_WHILEM3: - goto resume_point_WHILEM3; - case resume_WHILEM4: - goto resume_point_WHILEM4; - case resume_WHILEM5: - goto resume_point_WHILEM5; - case resume_WHILEM6: - goto resume_point_WHILEM6; - - case TRIE_next: - case EVAL_AB: - case BRANCH_next: - case CURLYM_A: - case CURLYM_B: - case IFMATCH_A: - case CURLY_B_max: - case CURLY_B_min: - case CURLY_B_min_known: - state_num = st->resume_state + 1; /* failure = success + 1 */ - goto reenter_switch; - - default: - Perl_croak(aTHX_ "regexp resume memory corruption"); - } + state_num = st->resume_state + 1; /* failure = success + 1 */ + goto reenter_switch; } + result = 0; final_exit: - + if (rex->intflags & PREGf_VERBARG_SEEN) { + SV *sv_err = get_sv("REGERROR", 1); + SV *sv_mrk = get_sv("REGMARK", 1); + if (result) { + sv_commit = &PL_sv_no; + if (!sv_yes_mark) + sv_yes_mark = &PL_sv_yes; + } else { + if (!sv_commit) + sv_commit = &PL_sv_yes; + sv_yes_mark = &PL_sv_no; + } + sv_setsv(sv_err, sv_commit); + sv_setsv(sv_mrk, sv_yes_mark); + } /* restore original high-water mark */ PL_regmatch_slab = orig_slab; PL_regmatch_state = orig_state; @@ -4754,7 +5092,7 @@ no_silent: * rather than incrementing count on every character. [Er, except utf8.]] */ STATIC I32 -S_regrepeat(pTHX_ const regexp *prog, const regnode *p, I32 max) +S_regrepeat(pTHX_ const regexp *prog, const regnode *p, I32 max, int depth) { dVAR; register char *scan; @@ -4985,7 +5323,7 @@ S_regrepeat(pTHX_ const regexp *prog, const regnode *p, I32 max) regprop(prog, prop, p); PerlIO_printf(Perl_debug_log, "%*s %s can match %"IVdf" times out of %"IVdf"...\n", - REPORT_CODE_OFF+1, "", SvPVX_const(prop),(IV)c,(IV)max); + REPORT_CODE_OFF + depth*2, "", SvPVX_const(prop),(IV)c,(IV)max); }); }); @@ -5005,7 +5343,8 @@ Perl_regclass_swash(pTHX_ const regexp *prog, register const regnode* node, bool SV *sw = NULL; SV *si = NULL; SV *alt = NULL; - const struct reg_data * const data = prog ? prog->data : NULL; + RXi_GET_DECL(prog,progi); + const struct reg_data * const data = prog ? progi->data : NULL; if (data && data->count) { const U32 n = ARG(node); @@ -5278,56 +5617,46 @@ restore_pos(pTHX_ void *arg) STATIC void S_to_utf8_substr(pTHX_ register regexp *prog) { - if (prog->float_substr && !prog->float_utf8) { - SV* const sv = newSVsv(prog->float_substr); - prog->float_utf8 = sv; - sv_utf8_upgrade(sv); - if (SvTAIL(prog->float_substr)) - SvTAIL_on(sv); - if (prog->float_substr == prog->check_substr) - prog->check_utf8 = sv; - } - if (prog->anchored_substr && !prog->anchored_utf8) { - SV* const sv = newSVsv(prog->anchored_substr); - prog->anchored_utf8 = sv; - sv_utf8_upgrade(sv); - if (SvTAIL(prog->anchored_substr)) - SvTAIL_on(sv); - if (prog->anchored_substr == prog->check_substr) - prog->check_utf8 = sv; - } + int i = 1; + do { + if (prog->substrs->data[i].substr + && !prog->substrs->data[i].utf8_substr) { + SV* const sv = newSVsv(prog->substrs->data[i].substr); + prog->substrs->data[i].utf8_substr = sv; + sv_utf8_upgrade(sv); + if (SvVALID(prog->substrs->data[i].substr)) + fbm_compile(sv, 0); + if (SvTAIL(prog->substrs->data[i].substr)) + SvTAIL_on(sv); + if (prog->substrs->data[i].substr == prog->check_substr) + prog->check_utf8 = sv; + } + } while (i--); } STATIC void S_to_byte_substr(pTHX_ register regexp *prog) { dVAR; - if (prog->float_utf8 && !prog->float_substr) { - SV* sv = newSVsv(prog->float_utf8); - prog->float_substr = sv; - if (sv_utf8_downgrade(sv, TRUE)) { - if (SvTAIL(prog->float_utf8)) - SvTAIL_on(sv); - } else { - SvREFCNT_dec(sv); - prog->float_substr = sv = &PL_sv_undef; - } - if (prog->float_utf8 == prog->check_utf8) - prog->check_substr = sv; - } - if (prog->anchored_utf8 && !prog->anchored_substr) { - SV* sv = newSVsv(prog->anchored_utf8); - prog->anchored_substr = sv; - if (sv_utf8_downgrade(sv, TRUE)) { - if (SvTAIL(prog->anchored_utf8)) - SvTAIL_on(sv); - } else { - SvREFCNT_dec(sv); - prog->anchored_substr = sv = &PL_sv_undef; + int i = 1; + do { + if (prog->substrs->data[i].utf8_substr + && !prog->substrs->data[i].substr) { + SV* sv = newSVsv(prog->substrs->data[i].utf8_substr); + if (sv_utf8_downgrade(sv, TRUE)) { + if (SvVALID(prog->substrs->data[i].utf8_substr)) + fbm_compile(sv, 0); + if (SvTAIL(prog->substrs->data[i].utf8_substr)) + SvTAIL_on(sv); + } else { + SvREFCNT_dec(sv); + sv = &PL_sv_undef; + } + prog->substrs->data[i].substr = sv; + if (prog->substrs->data[i].utf8_substr == prog->check_utf8) + prog->check_substr = sv; } - if (prog->anchored_utf8 == prog->check_utf8) - prog->check_substr = sv; - } + } while (i--); } /*