X-Git-Url: https://perl5.git.perl.org/perl5.git/blobdiff_plain/a3621e74372f5d2c10ed0d2a21195cab42a5be54..fd40a117d944c5311ff8dd0bde49755b81f72661:/regexec.c diff --git a/regexec.c b/regexec.c index 192396f..c4d88cc 100644 --- a/regexec.c +++ b/regexec.c @@ -13,7 +13,6 @@ * it's built with -DPERL_EXT_RE_BUILD -DPERL_EXT_RE_DEBUG -DPERL_EXT. * This causes the main functions to be compiled under new names and with * debugging support added, which makes "use re 'debug'" work. - */ /* NOTE: this is derived from Henry Spencer's regexp code, and should not @@ -31,31 +30,9 @@ */ #ifdef PERL_EXT_RE_BUILD -/* need to replace pregcomp et al, so enable that */ -# ifndef PERL_IN_XSUB_RE -# define PERL_IN_XSUB_RE -# endif -/* need access to debugger hooks */ -# if defined(PERL_EXT_RE_DEBUG) && !defined(DEBUGGING) -# define DEBUGGING -# endif -#endif - -#ifdef PERL_IN_XSUB_RE -/* We *really* need to overwrite these symbols: */ -# define Perl_regexec_flags my_regexec -# define Perl_regdump my_regdump -# define Perl_regprop my_regprop -# define Perl_re_intuit_start my_re_intuit_start -/* *These* symbols are masked to allow static link. */ -# define Perl_pregexec my_pregexec -# define Perl_reginitcolors my_reginitcolors -# define Perl_regclass_swash my_regclass_swash - -# define PERL_NO_GET_CONTEXT +#include "re_top.h" #endif -/*SUPPRESS 112*/ /* * pregcomp and pregexec -- regsub and regerror are not used in perl * @@ -79,7 +56,7 @@ **** Alterations to Henry's code are... **** **** Copyright (C) 1991, 1992, 1993, 1994, 1995, 1996, 1997, 1998, 1999, - **** 2000, 2001, 2002, 2003, 2004, 2005, by Larry Wall and others + **** 2000, 2001, 2002, 2003, 2004, 2005, 2006, by Larry Wall and others **** **** You may distribute under the terms of either the GNU General Public **** License or the Artistic License, as specified in the README file. @@ -92,13 +69,16 @@ #define PERL_IN_REGEXEC_C #include "perl.h" -#include "regcomp.h" +#ifdef PERL_IN_XSUB_RE +# include "re_comp.h" +#else +# include "regcomp.h" +#endif #define RF_tainted 1 /* tainted information used? */ #define RF_warned 2 /* warned about big count? */ -#define RF_evaled 4 /* Did an EVAL with setting? */ -#define RF_utf8 8 /* String contains multibyte chars? */ -#define RF_false 16 /* odd number of nested negatives */ + +#define RF_utf8 8 /* Pattern contains multibyte chars? */ #define UTF ((PL_reg_flags & RF_utf8) != 0) @@ -109,7 +89,7 @@ #define STATIC static #endif -#define REGINCLASS(p,c) (ANYOF_FLAGS(p) ? reginclass(p,c,0,0) : ANYOF_BITMAP_TEST(p,*(c))) +#define REGINCLASS(prog,p,c) (ANYOF_FLAGS(p) ? reginclass(prog,p,c,0,0) : ANYOF_BITMAP_TEST(p,*(c))) /* * Forwards. @@ -118,41 +98,39 @@ #define CHR_SVLEN(sv) (do_utf8 ? sv_len_utf8(sv) : SvCUR(sv)) #define CHR_DIST(a,b) (PL_reg_match_utf8 ? utf8_distance(a,b) : a - b) -#define reghop_c(pos,off) ((char*)reghop((U8*)pos, off)) -#define reghopmaybe_c(pos,off) ((char*)reghopmaybe((U8*)pos, off)) -#define HOP(pos,off) (PL_reg_match_utf8 ? reghop((U8*)pos, off) : (U8*)(pos + off)) -#define HOPMAYBE(pos,off) (PL_reg_match_utf8 ? reghopmaybe((U8*)pos, off) : (U8*)(pos + off)) -#define HOPc(pos,off) ((char*)HOP(pos,off)) -#define HOPMAYBEc(pos,off) ((char*)HOPMAYBE(pos,off)) - -#define HOPBACK(pos, off) ( \ - (PL_reg_match_utf8) \ - ? reghopmaybe((U8*)pos, -off) \ - : (pos - off >= PL_bostr) \ - ? (U8*)(pos - off) \ - : (U8*)NULL \ -) -#define HOPBACKc(pos, off) (char*)HOPBACK(pos, off) - -#define reghop3_c(pos,off,lim) ((char*)reghop3((U8*)pos, off, (U8*)lim)) -#define reghopmaybe3_c(pos,off,lim) ((char*)reghopmaybe3((U8*)pos, off, (U8*)lim)) -#define HOP3(pos,off,lim) (PL_reg_match_utf8 ? reghop3((U8*)pos, off, (U8*)lim) : (U8*)(pos + off)) -#define HOPMAYBE3(pos,off,lim) (PL_reg_match_utf8 ? reghopmaybe3((U8*)pos, off, (U8*)lim) : (U8*)(pos + off)) +#define HOPc(pos,off) \ + (char *)(PL_reg_match_utf8 \ + ? reghop3((U8*)pos, off, (U8*)(off >= 0 ? PL_regeol : PL_bostr)) \ + : (U8*)(pos + off)) +#define HOPBACKc(pos, off) \ + (char*)(PL_reg_match_utf8\ + ? reghopmaybe3((U8*)pos, -off, (U8*)PL_bostr) \ + : (pos - off >= PL_bostr) \ + ? (U8*)pos - off \ + : NULL) + +#define HOP3(pos,off,lim) (PL_reg_match_utf8 ? reghop3((U8*)(pos), off, (U8*)(lim)) : (U8*)(pos + off)) #define HOP3c(pos,off,lim) ((char*)HOP3(pos,off,lim)) -#define HOPMAYBE3c(pos,off,lim) ((char*)HOPMAYBE3(pos,off,lim)) -#define LOAD_UTF8_CHARCLASS(a,b) STMT_START { if (!CAT2(PL_utf8_,a)) { ENTER; save_re_context(); (void)CAT2(is_utf8_, a)((U8*)b); LEAVE; } } STMT_END +#define LOAD_UTF8_CHARCLASS(class,str) STMT_START { \ + if (!CAT2(PL_utf8_,class)) { bool ok; ENTER; save_re_context(); ok=CAT2(is_utf8_,class)((const U8*)str); assert(ok); LEAVE; } } STMT_END +#define LOAD_UTF8_CHARCLASS_ALNUM() LOAD_UTF8_CHARCLASS(alnum,"a") +#define LOAD_UTF8_CHARCLASS_DIGIT() LOAD_UTF8_CHARCLASS(digit,"0") +#define LOAD_UTF8_CHARCLASS_SPACE() LOAD_UTF8_CHARCLASS(space," ") +#define LOAD_UTF8_CHARCLASS_MARK() LOAD_UTF8_CHARCLASS(mark, "\xcd\x86") + +/* TODO: Combine JUMPABLE and HAS_TEXT to cache OP(rn) */ /* for use after a quantifier and before an EXACT-like node -- japhy */ #define JUMPABLE(rn) ( \ OP(rn) == OPEN || OP(rn) == CLOSE || OP(rn) == EVAL || \ OP(rn) == SUSPEND || OP(rn) == IFMATCH || \ OP(rn) == PLUS || OP(rn) == MINMOD || \ - (PL_regkind[(U8)OP(rn)] == CURLY && ARG1(rn) > 0) \ + (PL_regkind[OP(rn)] == CURLY && ARG1(rn) > 0) \ ) #define HAS_TEXT(rn) ( \ - PL_regkind[(U8)OP(rn)] == EXACT || PL_regkind[(U8)OP(rn)] == REF \ + PL_regkind[OP(rn)] == EXACT || PL_regkind[OP(rn)] == REF \ ) /* @@ -160,30 +138,35 @@ follow but for lookbehind (rn->flags != 0) we skip to the next step. */ #define FIND_NEXT_IMPT(rn) STMT_START { \ - while (JUMPABLE(rn)) \ - if (OP(rn) == SUSPEND || PL_regkind[(U8)OP(rn)] == CURLY) \ + while (JUMPABLE(rn)) { \ + const OPCODE type = OP(rn); \ + if (type == SUSPEND || PL_regkind[type] == CURLY) \ rn = NEXTOPER(NEXTOPER(rn)); \ - else if (OP(rn) == PLUS) \ + else if (type == PLUS) \ rn = NEXTOPER(rn); \ - else if (OP(rn) == IFMATCH) \ + else if (type == IFMATCH) \ rn = (rn->flags == 0) ? NEXTOPER(NEXTOPER(rn)) : rn + ARG(rn); \ else rn += NEXT_OFF(rn); \ + } \ } STMT_END + static void restore_pos(pTHX_ void *arg); STATIC CHECKPOINT S_regcppush(pTHX_ I32 parenfloor) { - int retval = PL_savestack_ix; + dVAR; + const int retval = PL_savestack_ix; #define REGCP_PAREN_ELEMS 4 - int paren_elems_to_push = (PL_regsize - parenfloor) * REGCP_PAREN_ELEMS; + const int paren_elems_to_push = (PL_regsize - parenfloor) * REGCP_PAREN_ELEMS; int p; + GET_RE_DEBUG_FLAGS_DECL; if (paren_elems_to_push < 0) Perl_croak(aTHX_ "panic: paren_elems_to_push < 0"); -#define REGCP_OTHER_ELEMS 6 +#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. */ @@ -191,8 +174,16 @@ S_regcppush(pTHX_ I32 parenfloor) SSPUSHINT(PL_regstartp[p]); SSPUSHPTR(PL_reg_start_tmp[p]); SSPUSHINT(p); + DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, + " saving \\%"UVuf" %"IVdf"(%"IVdf")..%"IVdf"\n", + (UV)p, (IV)PL_regstartp[p], + (IV)(PL_reg_start_tmp[p] - PL_bostr), + (IV)PL_regendp[p] + )); } /* REGCP_OTHER_ELEMS are pushed in any case, parentheses or no. */ + SSPUSHPTR(PL_regstartp); + SSPUSHPTR(PL_regendp); SSPUSHINT(PL_regsize); SSPUSHINT(*PL_reglastparen); SSPUSHINT(*PL_reglastcloseparen); @@ -207,22 +198,27 @@ S_regcppush(pTHX_ I32 parenfloor) } /* These are needed since we do not localize EVAL nodes: */ -# define REGCP_SET(cp) DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, \ - " Setting an EVAL scope, savestack=%"IVdf"\n", \ - (IV)PL_savestack_ix)); cp = PL_savestack_ix - -# define REGCP_UNWIND(cp) DEBUG_EXECUTE_r(cp != PL_savestack_ix ? \ - PerlIO_printf(Perl_debug_log, \ - " Clearing an EVAL scope, savestack=%"IVdf"..%"IVdf"\n", \ - (IV)(cp), (IV)PL_savestack_ix) : 0); regcpblow(cp) +#define REGCP_SET(cp) \ + DEBUG_STATE_r( \ + PerlIO_printf(Perl_debug_log, \ + " Setting an EVAL scope, savestack=%"IVdf"\n", \ + (IV)PL_savestack_ix)); \ + cp = PL_savestack_ix + +#define REGCP_UNWIND(cp) \ + DEBUG_STATE_r( \ + if (cp != PL_savestack_ix) \ + PerlIO_printf(Perl_debug_log, \ + " Clearing an EVAL scope, savestack=%"IVdf"..%"IVdf"\n", \ + (IV)(cp), (IV)PL_savestack_ix)); \ + regcpblow(cp) STATIC char * -S_regcppop(pTHX) +S_regcppop(pTHX_ const regexp *rex) { + dVAR; I32 i; - U32 paren = 0; char *input; - I32 tmps; GET_RE_DEBUG_FLAGS_DECL; @@ -234,11 +230,15 @@ S_regcppop(pTHX) *PL_reglastcloseparen = SSPOPINT; *PL_reglastparen = SSPOPINT; PL_regsize = SSPOPINT; + PL_regendp=(I32 *) SSPOPPTR; + PL_regstartp=(I32 *) SSPOPPTR; + /* Now restore the parentheses context. */ for (i -= (REGCP_OTHER_ELEMS - REGCP_FRAME_ELEMS); i > 0; i -= REGCP_PAREN_ELEMS) { - paren = (U32)SSPOPINT; + I32 tmps; + U32 paren = (U32)SSPOPINT; PL_reg_start_tmp[paren] = (char *) SSPOPPTR; PL_regstartp[paren] = SSPOPINT; tmps = SSPOPINT; @@ -254,10 +254,10 @@ S_regcppop(pTHX) ); } DEBUG_EXECUTE_r( - if ((I32)(*PL_reglastparen + 1) <= PL_regnpar) { + if (*PL_reglastparen + 1 <= rex->nparens) { PerlIO_printf(Perl_debug_log, " restoring \\%"IVdf"..\\%"IVdf" to undef\n", - (IV)(*PL_reglastparen + 1), (IV)PL_regnpar); + (IV)(*PL_reglastparen + 1), (IV)rex->nparens); } ); #if 1 @@ -271,57 +271,22 @@ S_regcppop(pTHX) * building DynaLoader will fail: * "Error: '*' not in typemap in DynaLoader.xs, line 164" * --jhi */ - for (paren = *PL_reglastparen + 1; (I32)paren <= PL_regnpar; paren++) { - if ((I32)paren > PL_regsize) - PL_regstartp[paren] = -1; - PL_regendp[paren] = -1; + for (i = *PL_reglastparen + 1; (U32)i <= rex->nparens; i++) { + if (i > PL_regsize) + PL_regstartp[i] = -1; + PL_regendp[i] = -1; } #endif return input; } -STATIC char * -S_regcp_set_to(pTHX_ I32 ss) -{ - I32 tmp = PL_savestack_ix; - - PL_savestack_ix = ss; - regcppop(); - PL_savestack_ix = tmp; - return Nullch; -} - -typedef struct re_cc_state -{ - I32 ss; - regnode *node; - struct re_cc_state *prev; - CURCUR *cc; - regexp *re; -} re_cc_state; - #define regcpblow(cp) LEAVE_SCOPE(cp) /* Ignores regcppush()ed data. */ -#define TRYPAREN(paren, n, input) { \ - if (paren) { \ - if (n) { \ - PL_regstartp[paren] = HOPc(input, -1) - PL_bostr; \ - PL_regendp[paren] = input - PL_bostr; \ - } \ - else \ - PL_regendp[paren] = -1; \ - } \ - if (regmatch(next)) \ - sayYES; \ - if (paren && n) \ - PL_regendp[paren] = -1; \ -} - - /* * pregexec and friends */ +#ifndef PERL_IN_XSUB_RE /* - pregexec - match a regexp against a string */ @@ -337,18 +302,7 @@ Perl_pregexec(pTHX_ register regexp *prog, char *stringarg, register char *stren regexec_flags(prog, stringarg, strend, strbeg, minend, screamer, NULL, nosave ? 0 : REXEC_COPY_STR); } - -STATIC void -S_cache_re(pTHX_ regexp *prog) -{ - PL_regprecomp = prog->precomp; /* Needed for FAIL. */ -#ifdef DEBUGGING - PL_regprogram = prog->program; #endif - PL_regnpar = prog->nparens; - PL_regdata = prog->data; - PL_reg_re = prog; -} /* * Need to implement the following flags for reg_anch: @@ -366,7 +320,7 @@ S_cache_re(pTHX_ regexp *prog) /* Assumptions: if ANCH_GPOS, then strpos is anchored. XXXX Check GPOS logic */ -/* If SCREAM, then SvPVX(sv) should be compatible with strpos and strend. +/* If SCREAM, then SvPVX_const(sv) should be compatible with strpos and strend. Otherwise, only SvCUR(sv) is used to get strbeg. */ /* XXXX We assume that strpos is strbeg unless sv. */ @@ -400,6 +354,7 @@ char * Perl_re_intuit_start(pTHX_ regexp *prog, SV *sv, char *strpos, char *strend, U32 flags, re_scream_pos_data *data) { + dVAR; register I32 start_shift = 0; /* Should be nonnegative! */ register I32 end_shift = 0; @@ -407,14 +362,13 @@ Perl_re_intuit_start(pTHX_ regexp *prog, SV *sv, char *strpos, register SV *check; char *strbeg; char *t; - int do_utf8 = sv ? SvUTF8(sv) : 0; /* if no sv we have to assume bytes */ + const bool do_utf8 = (sv && SvUTF8(sv)) ? 1 : 0; /* if no sv we have to assume bytes */ I32 ml_anch; - register char *other_last = Nullch; /* other substr checked before this */ - char *check_at = Nullch; /* check substr found at this pos */ - I32 multiline = prog->reganch & PMf_MULTILINE; + 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; #ifdef DEBUGGING - char *i_strpos = strpos; - SV *dsv = PERL_DEBUG_PAD_ZERO(0); + const char * const i_strpos = strpos; #endif GET_RE_DEBUG_FLAGS_DECL; @@ -422,34 +376,13 @@ Perl_re_intuit_start(pTHX_ regexp *prog, SV *sv, char *strpos, RX_MATCH_UTF8_set(prog,do_utf8); if (prog->reganch & ROPT_UTF8) { - DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, - "UTF-8 regex...\n")); PL_reg_flags |= RF_utf8; } - - DEBUG_EXECUTE_r({ - char *s = PL_reg_match_utf8 ? - sv_uni_display(dsv, sv, 60, UNI_DISPLAY_REGEX) : - strpos; - int len = PL_reg_match_utf8 ? - strlen(s) : strend - strpos; - if (!PL_colorset) - reginitcolors(); - if (PL_reg_match_utf8) - DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, - "UTF-8 target...\n")); - PerlIO_printf(Perl_debug_log, - "%sGuessing start of match, REx%s `%s%.60s%s%s' against `%s%.*s%s%s'...\n", - PL_colors[4],PL_colors[5],PL_colors[0], - prog->precomp, - PL_colors[1], - (strlen(prog->precomp) > 60 ? "..." : ""), - PL_colors[0], - (int)(len > 60 ? 60 : len), - s, PL_colors[1], - (len > 60 ? "..." : "") + DEBUG_EXECUTE_r( + debug_start_match(prog, do_utf8, strpos, strend, + sv ? "Guessing start of match in sv for" + : "Guessing start of match in string for"); ); - }); /* CHR_DIST() would be more correct here but it makes things slow. */ if (prog->minlen > strend - strpos) { @@ -457,6 +390,7 @@ Perl_re_intuit_start(pTHX_ regexp *prog, SV *sv, char *strpos, "String too short... [re_intuit_start]\n")); goto fail; } + strbeg = (sv && SvPOK(sv)) ? strend - SvCUR(sv) : strpos; PL_regeol = strend; if (do_utf8) { @@ -468,9 +402,9 @@ Perl_re_intuit_start(pTHX_ regexp *prog, SV *sv, char *strpos, to_byte_substr(prog); check = prog->check_substr; } - if (check == &PL_sv_undef) { + if (check == &PL_sv_undef) { DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, - "Non-utf string cannot match utf check string\n")); + "Non-utf8 string cannot match utf8 check string\n")); goto fail; } if (prog->reganch & ROPT_ANCH) { /* Match at beg-of-str or after \n */ @@ -481,7 +415,7 @@ Perl_re_intuit_start(pTHX_ regexp *prog, SV *sv, char *strpos, if (!ml_anch) { if ( !(prog->reganch & (ROPT_ANCH_GPOS /* Checked by the caller */ | ROPT_IMPLICIT)) /* not a real BOL */ - /* SvCUR is not set on references: SvRV and SvPVX overlap */ + /* SvCUR is not set on references: SvRV and SvPVX_const overlap */ && sv && !SvROK(sv) && (strpos != strbeg)) { DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "Not at start...\n")); @@ -493,6 +427,7 @@ Perl_re_intuit_start(pTHX_ regexp *prog, SV *sv, char *strpos, I32 slen; s = HOP3c(strpos, prog->check_offset_min, strend); + if (SvTAIL(check)) { slen = SvCUR(check); /* >= 1 */ @@ -503,30 +438,31 @@ Perl_re_intuit_start(pTHX_ regexp *prog, SV *sv, char *strpos, } /* Now should match s[0..slen-2] */ slen--; - if (slen && (*SvPVX(check) != *s + if (slen && (*SvPVX_const(check) != *s || (slen > 1 - && memNE(SvPVX(check), s, slen)))) { + && memNE(SvPVX_const(check), s, slen)))) { report_neq: DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "String not equal...\n")); goto fail_finish; } } - else if (*SvPVX(check) != *s + else if (*SvPVX_const(check) != *s || ((slen = SvCUR(check)) > 1 - && memNE(SvPVX(check), s, slen))) + && memNE(SvPVX_const(check), s, slen))) goto report_neq; + check_at = s; goto success_at_start; } } /* Match is anchored, but substr is not anchored wrt beg-of-str. */ s = strpos; start_shift = prog->check_offset_min; /* okay to underestimate on CC */ - end_shift = prog->minlen - start_shift - - CHR_SVLEN(check) + (SvTAIL(check) != 0); + end_shift = prog->check_end_shift; + if (!ml_anch) { - I32 end = prog->check_offset_max + CHR_SVLEN(check) + const I32 end = prog->check_offset_max + CHR_SVLEN(check) - (SvTAIL(check) != 0); - I32 eshift = CHR_DIST((U8*)strend, (U8*)s) - end; + const I32 eshift = CHR_DIST((U8*)strend, (U8*)s) - end; if (end_shift < eshift) end_shift = eshift; @@ -535,76 +471,118 @@ Perl_re_intuit_start(pTHX_ regexp *prog, SV *sv, char *strpos, else { /* Can match at random position */ ml_anch = 0; s = strpos; - start_shift = prog->check_offset_min; /* okay to underestimate on CC */ - /* Should be nonnegative! */ - end_shift = prog->minlen - start_shift - - CHR_SVLEN(check) + (SvTAIL(check) != 0); + start_shift = prog->check_offset_min; /* okay to underestimate on CC */ + end_shift = prog->check_end_shift; + + /* end shift should be non negative here */ } #ifdef DEBUGGING /* 7/99: reports of failure (with the older version) */ if (end_shift < 0) - Perl_croak(aTHX_ "panic: end_shift"); + Perl_croak(aTHX_ "panic: end_shift: %"IVdf" pattern:\n%s\n ", + (IV)end_shift, prog->precomp); #endif restart: /* Find a possible match in the region s..strend by looking for the "check" substring in the region corrected by start/end_shift. */ + + { + I32 srch_start_shift = start_shift; + I32 srch_end_shift = end_shift; + if (srch_start_shift < 0 && strbeg - s > srch_start_shift) { + srch_end_shift -= ((strbeg - s) - srch_start_shift); + srch_start_shift = strbeg - s; + } + DEBUG_OPTIMISE_MORE_r({ + PerlIO_printf(Perl_debug_log, "Check offset min: %"IVdf" Start shift: %"IVdf" End shift %"IVdf" Real End Shift: %"IVdf"\n", + (IV)prog->check_offset_min, + (IV)srch_start_shift, + (IV)srch_end_shift, + (IV)prog->check_end_shift); + }); + if (flags & REXEC_SCREAM) { I32 p = -1; /* Internal iterator of scream. */ - I32 *pp = data ? data->scream_pos : &p; + I32 * const pp = data ? data->scream_pos : &p; if (PL_screamfirst[BmRARE(check)] >= 0 || ( BmRARE(check) == '\n' && (BmPREVIOUS(check) == SvCUR(check) - 1) && SvTAIL(check) )) s = screaminstr(sv, check, - start_shift + (s - strbeg), end_shift, pp, 0); + srch_start_shift + (s - strbeg), srch_end_shift, pp, 0); else goto fail_finish; /* we may be pointing at the wrong string */ if (s && RX_MATCH_COPIED(prog)) - s = strbeg + (s - SvPVX(sv)); + s = strbeg + (s - SvPVX_const(sv)); if (data) *data->scream_olds = s; } - else if (prog->reganch & ROPT_CANY_SEEN) - s = fbm_instr((U8*)(s + start_shift), - (U8*)(strend - end_shift), - check, multiline ? FBMrf_MULTILINE : 0); - else - s = fbm_instr(HOP3(s, start_shift, strend), - HOP3(strend, -end_shift, strbeg), + else { + U8* start_point; + U8* end_point; + if (prog->reganch & ROPT_CANY_SEEN) { + start_point= (U8*)(s + srch_start_shift); + end_point= (U8*)(strend - srch_end_shift); + } else { + start_point= HOP3(s, srch_start_shift, srch_start_shift < 0 ? strbeg : strend); + end_point= HOP3(strend, -srch_end_shift, strbeg); + } + DEBUG_OPTIMISE_MORE_r({ + PerlIO_printf(Perl_debug_log, "fbm_instr len=%d str=<%.*s>\n", + (int)(end_point - start_point), + (int)(end_point - start_point) > 20 ? 20 : (int)(end_point - start_point), + start_point); + }); + + s = fbm_instr( start_point, end_point, check, multiline ? FBMrf_MULTILINE : 0); - + } + } /* Update the count-of-usability, remove useless subpatterns, unshift s. */ - DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "%s %s substr `%s%.*s%s'%s%s", + DEBUG_EXECUTE_r({ + RE_PV_QUOTED_DECL(quoted, do_utf8, PERL_DEBUG_PAD_ZERO(0), + SvPVX_const(check), RE_SV_DUMPLEN(check), 30); + PerlIO_printf(Perl_debug_log, "%s %s substr %s%s%s", (s ? "Found" : "Did not find"), - (check == (do_utf8 ? prog->anchored_utf8 : prog->anchored_substr) ? "anchored" : "floating"), - PL_colors[0], - (int)(SvCUR(check) - (SvTAIL(check)!=0)), - SvPVX(check), - PL_colors[1], (SvTAIL(check) ? "$" : ""), - (s ? " at offset " : "...\n") ) ); + (check == (do_utf8 ? prog->anchored_utf8 : prog->anchored_substr) + ? "anchored" : "floating"), + quoted, + RE_SV_TAIL(check), + (s ? " at offset " : "...\n") ); + }); if (!s) goto fail_finish; - - check_at = s; - /* Finish the diagnostic message */ DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "%ld...\n", (long)(s - i_strpos)) ); + /* XXX dmq: first branch is for positive lookbehind... + Our check string is offset from the beginning of the pattern. + So we need to do any stclass tests offset forward from that + point. I think. :-( + */ + + + + check_at=s; + + /* Got a candidate. Check MBOL anchoring, and the *other* substr. Start with the other substr. XXXX no SCREAM optimization yet - and a very coarse implementation - XXXX /ttx+/ results in anchored=`ttx', floating=`x'. floating will + XXXX /ttx+/ results in anchored="ttx", floating="x". floating will *always* match. Probably should be marked during compile... Probably it is right to do no SCREAM here... */ - if (do_utf8 ? (prog->float_utf8 && prog->anchored_utf8) : (prog->float_substr && prog->anchored_substr)) { + if (do_utf8 ? (prog->float_utf8 && prog->anchored_utf8) + : (prog->float_substr && prog->anchored_substr)) + { /* Take into account the "other" substring. */ /* XXXX May be hopelessly wrong for UTF... */ if (!other_last) @@ -612,16 +590,17 @@ Perl_re_intuit_start(pTHX_ regexp *prog, SV *sv, char *strpos, if (check == (do_utf8 ? prog->float_utf8 : prog->float_substr)) { do_other_anchored: { - char *last = HOP3c(s, -start_shift, strbeg), *last1, *last2; - char *s1 = s; + char * const last = HOP3c(s, -start_shift, strbeg); + char *last1, *last2; + char * const saved_s = s; SV* must; t = s - prog->check_offset_max; if (s - strpos > prog->check_offset_max /* signed-corrected t > strpos */ && (!do_utf8 - || ((t = reghopmaybe3_c(s, -(prog->check_offset_max), strpos)) + || ((t = (char*)reghopmaybe3((U8*)s, -(prog->check_offset_max), (U8*)strpos)) && t > strpos))) - /* EMPTY */; + NOOP; else t = strpos; t = HOP3c(t, prog->anchored_offset, strend); @@ -630,12 +609,17 @@ Perl_re_intuit_start(pTHX_ regexp *prog, SV *sv, char *strpos, last2 = last1 = HOP3c(strend, -prog->minlen, strbeg); if (last < last1) last1 = last; - /* XXXX It is not documented what units *_offsets are in. Assume bytes. */ + /* XXXX It is not documented what units *_offsets are in. + We assume bytes, but this is clearly wrong. + Meaning this code needs to be carefully reviewed for errors. + dmq. + */ + /* On end-of-str: see comment below. */ must = do_utf8 ? prog->anchored_utf8 : prog->anchored_substr; if (must == &PL_sv_undef) { s = (char*)NULL; - DEBUG_EXECUTE_r(must = prog->anchored_utf8); /* for debug */ + DEBUG_r(must = prog->anchored_utf8); /* for debug */ } else s = fbm_instr( @@ -645,14 +629,15 @@ Perl_re_intuit_start(pTHX_ regexp *prog, SV *sv, char *strpos, must, multiline ? FBMrf_MULTILINE : 0 ); - DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, - "%s anchored substr `%s%.*s%s'%s", + DEBUG_EXECUTE_r({ + RE_PV_QUOTED_DECL(quoted, do_utf8, PERL_DEBUG_PAD_ZERO(0), + SvPVX_const(must), RE_SV_DUMPLEN(must), 30); + PerlIO_printf(Perl_debug_log, "%s anchored substr %s%s", (s ? "Found" : "Contradicts"), - PL_colors[0], - (int)(SvCUR(must) - - (SvTAIL(must)!=0)), - SvPVX(must), - PL_colors[1], (SvTAIL(must) ? "$" : ""))); + quoted, RE_SV_TAIL(must)); + }); + + if (!s) { if (last1 >= last2) { DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, @@ -661,7 +646,7 @@ Perl_re_intuit_start(pTHX_ regexp *prog, SV *sv, char *strpos, } DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, ", trying floating at offset %ld...\n", - (long)(HOP3c(s1, 1, strend) - i_strpos))); + (long)(HOP3c(saved_s, 1, strend) - i_strpos))); other_last = HOP3c(last1, prog->anchored_offset+1, strend); s = HOP3c(last, 1, strend); goto restart; @@ -671,7 +656,7 @@ Perl_re_intuit_start(pTHX_ regexp *prog, SV *sv, char *strpos, (long)(s - i_strpos))); t = HOP3c(s, -prog->anchored_offset, strbeg); other_last = HOP3c(s, 1, strend); - s = s1; + s = saved_s; if (t == strpos) goto try_at_start; goto try_at_offset; @@ -680,7 +665,7 @@ Perl_re_intuit_start(pTHX_ regexp *prog, SV *sv, char *strpos, } else { /* Take into account the floating substring. */ char *last, *last1; - char *s1 = s; + char * const saved_s = s; SV* must; t = HOP3c(s, -start_shift, strbeg); @@ -698,19 +683,20 @@ Perl_re_intuit_start(pTHX_ regexp *prog, SV *sv, char *strpos, and end-of-str is not later than strend we are OK. */ if (must == &PL_sv_undef) { s = (char*)NULL; - DEBUG_EXECUTE_r(must = prog->float_utf8); /* for debug message */ + DEBUG_r(must = prog->float_utf8); /* for debug message */ } else s = fbm_instr((unsigned char*)s, (unsigned char*)last + SvCUR(must) - (SvTAIL(must)!=0), must, multiline ? FBMrf_MULTILINE : 0); - DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "%s floating substr `%s%.*s%s'%s", + DEBUG_EXECUTE_r({ + RE_PV_QUOTED_DECL(quoted, do_utf8, PERL_DEBUG_PAD_ZERO(0), + SvPVX_const(must), RE_SV_DUMPLEN(must), 30); + PerlIO_printf(Perl_debug_log, "%s floating substr %s%s", (s ? "Found" : "Contradicts"), - PL_colors[0], - (int)(SvCUR(must) - (SvTAIL(must)!=0)), - SvPVX(must), - PL_colors[1], (SvTAIL(must) ? "$" : ""))); + quoted, RE_SV_TAIL(must)); + }); if (!s) { if (last1 == last) { DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, @@ -719,7 +705,7 @@ Perl_re_intuit_start(pTHX_ regexp *prog, SV *sv, char *strpos, } DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, ", trying anchored starting at offset %ld...\n", - (long)(s1 + 1 - i_strpos))); + (long)(saved_s + 1 - i_strpos))); other_last = last; s = HOP3c(t, 1, strend); goto restart; @@ -728,7 +714,7 @@ Perl_re_intuit_start(pTHX_ regexp *prog, SV *sv, char *strpos, DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, " at offset %ld...\n", (long)(s - i_strpos))); other_last = s; /* Fix this later. --Hugo */ - s = s1; + s = saved_s; if (t == strpos) goto try_at_start; goto try_at_offset; @@ -736,11 +722,26 @@ Perl_re_intuit_start(pTHX_ regexp *prog, SV *sv, char *strpos, } } - t = s - prog->check_offset_max; + + t= (char*)HOP3( s, -prog->check_offset_max, (prog->check_offset_max<0) ? strend : strpos); + + DEBUG_OPTIMISE_MORE_r( + PerlIO_printf(Perl_debug_log, + "Check offset min:%"IVdf" max:%"IVdf" S:%"IVdf" t:%"IVdf" D:%"IVdf" end:%"IVdf"\n", + (IV)prog->check_offset_min, + (IV)prog->check_offset_max, + (IV)(s-strpos), + (IV)(t-strpos), + (IV)(t-s), + (IV)(strend-strpos) + ) + ); + if (s - strpos > prog->check_offset_max /* signed-corrected t > strpos */ && (!do_utf8 - || ((t = reghopmaybe3_c(s, -prog->check_offset_max, strpos)) - && t > strpos))) { + || ((t = (char*)reghopmaybe3((U8*)s, -prog->check_offset_max, (U8*) ((prog->check_offset_max<0) ? strend : strpos))) + && t > strpos))) + { /* Fixed substring is found far enough so that the match cannot start at strpos. */ try_at_offset: @@ -765,33 +766,33 @@ Perl_re_intuit_start(pTHX_ regexp *prog, SV *sv, char *strpos, */ strpos = t + 1; DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "Found /%s^%s/m at offset %ld, rescanning for anchored from offset %ld...\n", - PL_colors[0],PL_colors[1], (long)(strpos - i_strpos), (long)(strpos - i_strpos + prog->anchored_offset))); + PL_colors[0], PL_colors[1], (long)(strpos - i_strpos), (long)(strpos - i_strpos + prog->anchored_offset))); goto do_other_anchored; } /* We don't contradict the found floating substring. */ /* XXXX Why not check for STCLASS? */ s = t + 1; DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "Found /%s^%s/m at offset %ld...\n", - PL_colors[0],PL_colors[1], (long)(s - i_strpos))); + PL_colors[0], PL_colors[1], (long)(s - i_strpos))); goto set_useful; } /* Position contradicts check-string */ /* XXXX probably better to look for check-string than for "\n", so one should lower the limit for t? */ DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "Found /%s^%s/m, restarting lookup for check-string at offset %ld...\n", - PL_colors[0],PL_colors[1], (long)(t + 1 - i_strpos))); + PL_colors[0], PL_colors[1], (long)(t + 1 - i_strpos))); other_last = strpos = s = t + 1; goto restart; } t++; } DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "Did not find /%s^%s/m...\n", - PL_colors[0],PL_colors[1])); + PL_colors[0], PL_colors[1])); goto fail_finish; } else { DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "Starting position does not contradict /%s^%s/m...\n", - PL_colors[0],PL_colors[1])); + PL_colors[0], PL_colors[1])); } s = t; set_useful: @@ -815,7 +816,7 @@ Perl_re_intuit_start(pTHX_ regexp *prog, SV *sv, char *strpos, } 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? */ @@ -834,9 +835,9 @@ Perl_re_intuit_start(pTHX_ regexp *prog, SV *sv, char *strpos, SvREFCNT_dec(do_utf8 ? prog->check_utf8 : prog->check_substr); if (do_utf8 ? prog->check_substr : prog->check_utf8) SvREFCNT_dec(do_utf8 ? prog->check_substr : prog->check_utf8); - prog->check_substr = prog->check_utf8 = Nullsv; /* disable */ - prog->float_substr = prog->float_utf8 = Nullsv; /* clear */ - check = Nullsv; /* abort */ + prog->check_substr = prog->check_utf8 = NULL; /* disable */ + prog->float_substr = prog->float_utf8 = NULL; /* clear */ + check = NULL; /* abort */ s = strpos; /* XXXX This is a remnant of the old implementation. It looks wasteful, since now INTUIT can use many @@ -849,7 +850,9 @@ Perl_re_intuit_start(pTHX_ regexp *prog, SV *sv, char *strpos, /* Last resort... */ /* XXXX BmUSEFUL already changed, maybe multiple change is meaningful... */ - if (prog->regstclass) { + /* 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) { /* 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; @@ -858,24 +861,26 @@ 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(). */ - U8* str = (U8*)STRING(prog->regstclass); - int cl_l = (PL_regkind[(U8)OP(prog->regstclass)] == EXACT + 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) : 1); - char *endpos = (prog->anchored_substr || prog->anchored_utf8 || ml_anch) - ? HOP3c(s, (prog->minlen ? cl_l : 0), strend) - : (prog->float_substr || prog->float_utf8 - ? HOP3c(HOP3c(check_at, -start_shift, strbeg), - cl_l, strend) - : strend); - char *startpos = strbeg; - + char * endpos; + if (prog->anchored_substr || prog->anchored_utf8 || ml_anch) + endpos= HOP3c(s, (prog->minlen ? cl_l : 0), strend); + else if (prog->float_substr || prog->float_utf8) + endpos= HOP3c(HOP3c(check_at, -start_shift, strbeg), cl_l, strend); + 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)); + t = s; - cache_re(prog); - s = find_byclass(prog, prog->regstclass, s, endpos, startpos, 1); + s = find_byclass(prog, prog->regstclass, s, endpos, NULL); if (!s) { #ifdef DEBUGGING - char *what = 0; + const char *what = NULL; #endif if (endpos == strend) { DEBUG_EXECUTE_r( PerlIO_printf(Perl_debug_log, @@ -925,7 +930,7 @@ Perl_re_intuit_start(pTHX_ regexp *prog, SV *sv, char *strpos, goto giveup; DEBUG_EXECUTE_r( PerlIO_printf(Perl_debug_log, "Looking for /%s^%s/m starting at offset %ld...\n", - PL_colors[0],PL_colors[1], (long)(t - i_strpos)) ); + PL_colors[0], PL_colors[1], (long)(t - i_strpos)) ); goto try_at_offset; } if (!(do_utf8 ? prog->float_utf8 : prog->float_substr)) /* Could have been deleted */ @@ -959,15 +964,164 @@ Perl_re_intuit_start(pTHX_ regexp *prog, SV *sv, char *strpos, BmUSEFUL(do_utf8 ? prog->check_utf8 : prog->check_substr) += 5; /* hooray */ fail: DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "%sMatch rejected by optimizer%s\n", - PL_colors[4],PL_colors[5])); - return Nullch; + PL_colors[4], PL_colors[5])); + return NULL; } + + +#define REXEC_TRIE_READ_CHAR(trie_type, trie, uc, uscan, len, uvc, charid, \ +foldlen, foldbuf, uniflags) STMT_START { \ + switch (trie_type) { \ + case trie_utf8_fold: \ + if ( foldlen>0 ) { \ + uvc = utf8n_to_uvuni( uscan, UTF8_MAXLEN, &len, uniflags ); \ + foldlen -= len; \ + uscan += len; \ + len=0; \ + } else { \ + uvc = utf8n_to_uvuni( (U8*)uc, UTF8_MAXLEN, &len, uniflags ); \ + uvc = to_uni_fold( uvc, foldbuf, &foldlen ); \ + foldlen -= UNISKIP( uvc ); \ + uscan = foldbuf + UNISKIP( uvc ); \ + } \ + break; \ + case trie_utf8: \ + uvc = utf8n_to_uvuni( (U8*)uc, UTF8_MAXLEN, &len, uniflags ); \ + break; \ + case trie_plain: \ + uvc = (UV)*uc; \ + len = 1; \ + } \ + \ + if (uvc < 256) { \ + charid = trie->charmap[ uvc ]; \ + } \ + else { \ + charid = 0; \ + if (trie->widecharmap) { \ + SV** const svpp = hv_fetch(trie->widecharmap, \ + (char*)&uvc, sizeof(UV), 0); \ + if (svpp) \ + charid = (U16)SvIV(*svpp); \ + } \ + } \ +} STMT_END + +#define REXEC_FBC_EXACTISH_CHECK(CoNd) \ + if ( (CoNd) \ + && (ln == len || \ + ibcmp_utf8(s, NULL, 0, do_utf8, \ + m, NULL, ln, (bool)UTF)) \ + && (!reginfo || regtry(reginfo, s)) ) \ + goto got_it; \ + else { \ + U8 foldbuf[UTF8_MAXBYTES_CASE+1]; \ + uvchr_to_utf8(tmpbuf, c); \ + f = to_utf8_fold(tmpbuf, foldbuf, &foldlen); \ + if ( f != c \ + && (f == c1 || f == c2) \ + && (ln == foldlen || \ + !ibcmp_utf8((char *) foldbuf, \ + NULL, foldlen, do_utf8, \ + m, \ + NULL, ln, (bool)UTF)) \ + && (!reginfo || regtry(reginfo, s)) ) \ + goto got_it; \ + } \ + s += len + +#define REXEC_FBC_EXACTISH_SCAN(CoNd) \ +STMT_START { \ + while (s <= e) { \ + if ( (CoNd) \ + && (ln == 1 || !(OP(c) == EXACTF \ + ? ibcmp(s, m, ln) \ + : ibcmp_locale(s, m, ln))) \ + && (!reginfo || regtry(reginfo, s)) ) \ + goto got_it; \ + s++; \ + } \ +} STMT_END + +#define REXEC_FBC_UTF8_SCAN(CoDe) \ +STMT_START { \ + while (s + (uskip = UTF8SKIP(s)) <= strend) { \ + CoDe \ + s += uskip; \ + } \ +} STMT_END + +#define REXEC_FBC_SCAN(CoDe) \ +STMT_START { \ + while (s < strend) { \ + CoDe \ + s++; \ + } \ +} STMT_END + +#define REXEC_FBC_UTF8_CLASS_SCAN(CoNd) \ +REXEC_FBC_UTF8_SCAN( \ + if (CoNd) { \ + if (tmp && (!reginfo || regtry(reginfo, s))) \ + goto got_it; \ + else \ + tmp = doevery; \ + } \ + else \ + tmp = 1; \ +) + +#define REXEC_FBC_CLASS_SCAN(CoNd) \ +REXEC_FBC_SCAN( \ + if (CoNd) { \ + if (tmp && (!reginfo || regtry(reginfo, s))) \ + goto got_it; \ + else \ + tmp = doevery; \ + } \ + else \ + tmp = 1; \ +) + +#define REXEC_FBC_TRYIT \ +if ((!reginfo || regtry(reginfo, s))) \ + goto got_it + +#define REXEC_FBC_CSCAN_PRELOAD(UtFpReLoAd,CoNdUtF8,CoNd) \ + if (do_utf8) { \ + UtFpReLoAd; \ + REXEC_FBC_UTF8_CLASS_SCAN(CoNdUtF8); \ + } \ + else { \ + REXEC_FBC_CLASS_SCAN(CoNd); \ + } \ + break + +#define REXEC_FBC_CSCAN_TAINT(CoNdUtF8,CoNd) \ + PL_reg_flags |= RF_tainted; \ + if (do_utf8) { \ + REXEC_FBC_UTF8_CLASS_SCAN(CoNdUtF8); \ + } \ + else { \ + REXEC_FBC_CLASS_SCAN(CoNd); \ + } \ + break + +#define DUMP_EXEC_POS(li,s,doutf8) \ + dump_exec_pos(li,s,(PL_regeol),(PL_bostr),(PL_reg_starttry),doutf8) + /* We know what class REx starts with. Try to find this position... */ +/* if reginfo is NULL, its a dryrun */ +/* annoyingly all the vars in this routine have different names from their counterparts + in regmatch. /grrr */ + STATIC char * -S_find_byclass(pTHX_ regexp * prog, regnode *c, char *s, char *strend, char *startpos, I32 norun) +S_find_byclass(pTHX_ regexp * prog, const regnode *c, char *s, + const char *strend, const regmatch_info *reginfo) { - I32 doevery = (prog->reganch & ROPT_SKIP) == 0; + dVAR; + const I32 doevery = (prog->reganch & ROPT_SKIP) == 0; char *m; STRLEN ln; STRLEN lnc; @@ -976,37 +1130,27 @@ S_find_byclass(pTHX_ regexp * prog, regnode *c, char *s, char *strend, char *sta unsigned int c2; char *e; register I32 tmp = 1; /* Scratch variable? */ - register bool do_utf8 = PL_reg_match_utf8; + register const bool do_utf8 = PL_reg_match_utf8; /* We know what class it must start with. */ switch (OP(c)) { case ANYOF: if (do_utf8) { - while (s + (uskip = UTF8SKIP(s)) <= strend) { - if ((ANYOF_FLAGS(c) & ANYOF_UNICODE) || + REXEC_FBC_UTF8_CLASS_SCAN((ANYOF_FLAGS(c) & ANYOF_UNICODE) || !UTF8_IS_INVARIANT((U8)s[0]) ? - reginclass(c, (U8*)s, 0, do_utf8) : - REGINCLASS(c, (U8*)s)) { - if (tmp && (norun || regtry(prog, s))) - goto got_it; - else - tmp = doevery; - } - else - tmp = 1; - s += uskip; - } + reginclass(prog, c, (U8*)s, 0, do_utf8) : + REGINCLASS(prog, c, (U8*)s)); } else { while (s < strend) { STRLEN skip = 1; - if (REGINCLASS(c, (U8*)s) || + if (REGINCLASS(prog, c, (U8*)s) || (ANYOF_FOLD_SHARP_S(c, s, strend) && /* The assignment of 2 is intentional: * for the folded sharp s, the skip is 2. */ (skip = SHARP_S_SKIP))) { - if (tmp && (norun || regtry(prog, s))) + if (tmp && (!reginfo || regtry(reginfo, s))) goto got_it; else tmp = doevery; @@ -1018,13 +1162,12 @@ S_find_byclass(pTHX_ regexp * prog, regnode *c, char *s, char *strend, char *sta } break; case CANY: - while (s < strend) { - if (tmp && (norun || regtry(prog, s))) + REXEC_FBC_SCAN( + if (tmp && (!reginfo || regtry(reginfo, s))) goto got_it; else tmp = doevery; - s++; - } + ); break; case EXACTF: m = STRING(c); @@ -1035,14 +1178,15 @@ S_find_byclass(pTHX_ regexp * prog, regnode *c, char *s, char *strend, char *sta U8 *sm = (U8 *) m; U8 tmpbuf1[UTF8_MAXBYTES_CASE+1]; U8 tmpbuf2[UTF8_MAXBYTES_CASE+1]; + const U32 uniflags = UTF8_ALLOW_DEFAULT; to_utf8_lower((U8*)m, tmpbuf1, &ulen1); to_utf8_upper((U8*)m, tmpbuf2, &ulen2); c1 = utf8n_to_uvchr(tmpbuf1, UTF8_MAXBYTES_CASE, - 0, ckWARN(WARN_UTF8) ? 0 : UTF8_ALLOW_ANY); + 0, uniflags); c2 = utf8n_to_uvchr(tmpbuf2, UTF8_MAXBYTES_CASE, - 0, ckWARN(WARN_UTF8) ? 0 : UTF8_ALLOW_ANY); + 0, uniflags); lnc = 0; while (sm < ((U8 *) m + ln)) { lnc++; @@ -1063,7 +1207,7 @@ S_find_byclass(pTHX_ regexp * prog, regnode *c, char *s, char *strend, char *sta do_exactf: e = HOP3c(strend, -((I32)lnc), s); - if (norun && e < s) + if (!reginfo && e < s) e = s; /* Due to minlen logic of intuit() */ /* The idea in the EXACTF* cases is to first find the @@ -1080,43 +1224,21 @@ S_find_byclass(pTHX_ regexp * prog, regnode *c, char *s, char *strend, char *sta if (do_utf8) { UV c, f; U8 tmpbuf [UTF8_MAXBYTES+1]; - U8 foldbuf[UTF8_MAXBYTES_CASE+1]; STRLEN len, foldlen; - + const U32 uniflags = UTF8_ALLOW_DEFAULT; if (c1 == c2) { /* Upper and lower of 1st char are equal - * probably not a "letter". */ while (s <= e) { c = utf8n_to_uvchr((U8*)s, UTF8_MAXBYTES, &len, - ckWARN(WARN_UTF8) ? - 0 : UTF8_ALLOW_ANY); - if ( c == c1 - && (ln == len || - ibcmp_utf8(s, (char **)0, 0, do_utf8, - m, (char **)0, ln, (bool)UTF)) - && (norun || regtry(prog, s)) ) - goto got_it; - else { - uvchr_to_utf8(tmpbuf, c); - f = to_utf8_fold(tmpbuf, foldbuf, &foldlen); - if ( f != c - && (f == c1 || f == c2) - && (ln == foldlen || - !ibcmp_utf8((char *) foldbuf, - (char **)0, foldlen, do_utf8, - m, - (char **)0, ln, (bool)UTF)) - && (norun || regtry(prog, s)) ) - goto got_it; - } - s += len; + uniflags); + REXEC_FBC_EXACTISH_CHECK(c == c1); } } else { while (s <= e) { c = utf8n_to_uvchr((U8*)s, UTF8_MAXBYTES, &len, - ckWARN(WARN_UTF8) ? - 0 : UTF8_ALLOW_ANY); + uniflags); /* Handle some of the three Greek sigmas cases. * Note that not all the possible combinations @@ -1129,50 +1251,15 @@ S_find_byclass(pTHX_ regexp * prog, regnode *c, char *s, char *strend, char *sta c == (UV)UNICODE_GREEK_SMALL_LETTER_FINAL_SIGMA) c = (UV)UNICODE_GREEK_SMALL_LETTER_SIGMA; - if ( (c == c1 || c == c2) - && (ln == len || - ibcmp_utf8(s, (char **)0, 0, do_utf8, - m, (char **)0, ln, (bool)UTF)) - && (norun || regtry(prog, s)) ) - goto got_it; - else { - uvchr_to_utf8(tmpbuf, c); - f = to_utf8_fold(tmpbuf, foldbuf, &foldlen); - if ( f != c - && (f == c1 || f == c2) - && (ln == foldlen || - !ibcmp_utf8((char *) foldbuf, - (char **)0, foldlen, do_utf8, - m, - (char **)0, ln, (bool)UTF)) - && (norun || regtry(prog, s)) ) - goto got_it; - } - s += len; + REXEC_FBC_EXACTISH_CHECK(c == c1 || c == c2); } } } else { if (c1 == c2) - while (s <= e) { - if ( *(U8*)s == c1 - && (ln == 1 || !(OP(c) == EXACTF - ? ibcmp(s, m, ln) - : ibcmp_locale(s, m, ln))) - && (norun || regtry(prog, s)) ) - goto got_it; - s++; - } + REXEC_FBC_EXACTISH_SCAN(*(U8*)s == c1); else - while (s <= e) { - if ( (*(U8*)s == c1 || *(U8*)s == c2) - && (ln == 1 || !(OP(c) == EXACTF - ? ibcmp(s, m, ln) - : ibcmp_locale(s, m, ln))) - && (norun || regtry(prog, s)) ) - goto got_it; - s++; - } + REXEC_FBC_EXACTISH_SCAN(*(U8*)s == c1 || *(U8*)s == c2); } break; case BOUNDL: @@ -1183,39 +1270,34 @@ S_find_byclass(pTHX_ regexp * prog, regnode *c, char *s, char *strend, char *sta if (s == PL_bostr) tmp = '\n'; else { - U8 *r = reghop3((U8*)s, -1, (U8*)PL_bostr); - - tmp = utf8n_to_uvchr(r, UTF8SKIP(r), 0, 0); + U8 * const r = reghop3((U8*)s, -1, (U8*)PL_bostr); + tmp = utf8n_to_uvchr(r, UTF8SKIP(r), 0, UTF8_ALLOW_DEFAULT); } tmp = ((OP(c) == BOUND ? isALNUM_uni(tmp) : isALNUM_LC_uvchr(UNI_TO_NATIVE(tmp))) != 0); - LOAD_UTF8_CHARCLASS(alnum,"a"); - while (s + (uskip = UTF8SKIP(s)) <= strend) { + LOAD_UTF8_CHARCLASS_ALNUM(); + REXEC_FBC_UTF8_SCAN( if (tmp == !(OP(c) == BOUND ? - swash_fetch(PL_utf8_alnum, (U8*)s, do_utf8) : + (bool)swash_fetch(PL_utf8_alnum, (U8*)s, do_utf8) : isALNUM_LC_utf8((U8*)s))) { tmp = !tmp; - if ((norun || regtry(prog, s))) - goto got_it; - } - s += uskip; + REXEC_FBC_TRYIT; } + ); } else { tmp = (s != PL_bostr) ? UCHARAT(s - 1) : '\n'; tmp = ((OP(c) == BOUND ? isALNUM(tmp) : isALNUM_LC(tmp)) != 0); - while (s < strend) { + REXEC_FBC_SCAN( if (tmp == !(OP(c) == BOUND ? isALNUM(*s) : isALNUM_LC(*s))) { tmp = !tmp; - if ((norun || regtry(prog, s))) - goto got_it; - } - s++; + REXEC_FBC_TRYIT; } + ); } - if ((!prog->minlen && tmp) && (norun || regtry(prog, s))) + if ((!prog->minlen && tmp) && (!reginfo || regtry(reginfo, s))) goto got_it; break; case NBOUNDL: @@ -1226,385 +1308,308 @@ S_find_byclass(pTHX_ regexp * prog, regnode *c, char *s, char *strend, char *sta if (s == PL_bostr) tmp = '\n'; else { - U8 *r = reghop3((U8*)s, -1, (U8*)PL_bostr); - - tmp = utf8n_to_uvchr(r, UTF8SKIP(r), 0, 0); + U8 * const r = reghop3((U8*)s, -1, (U8*)PL_bostr); + tmp = utf8n_to_uvchr(r, UTF8SKIP(r), 0, UTF8_ALLOW_DEFAULT); } tmp = ((OP(c) == NBOUND ? isALNUM_uni(tmp) : isALNUM_LC_uvchr(UNI_TO_NATIVE(tmp))) != 0); - LOAD_UTF8_CHARCLASS(alnum,"a"); - while (s + (uskip = UTF8SKIP(s)) <= strend) { + LOAD_UTF8_CHARCLASS_ALNUM(); + REXEC_FBC_UTF8_SCAN( if (tmp == !(OP(c) == NBOUND ? - swash_fetch(PL_utf8_alnum, (U8*)s, do_utf8) : + (bool)swash_fetch(PL_utf8_alnum, (U8*)s, do_utf8) : isALNUM_LC_utf8((U8*)s))) tmp = !tmp; - else if ((norun || regtry(prog, s))) - goto got_it; - s += uskip; - } + else REXEC_FBC_TRYIT; + ); } else { tmp = (s != PL_bostr) ? UCHARAT(s - 1) : '\n'; tmp = ((OP(c) == NBOUND ? isALNUM(tmp) : isALNUM_LC(tmp)) != 0); - while (s < strend) { + REXEC_FBC_SCAN( if (tmp == !(OP(c) == NBOUND ? isALNUM(*s) : isALNUM_LC(*s))) tmp = !tmp; - else if ((norun || regtry(prog, s))) - goto got_it; - s++; - } + else REXEC_FBC_TRYIT; + ); } - if ((!prog->minlen && !tmp) && (norun || regtry(prog, s))) + if ((!prog->minlen && !tmp) && (!reginfo || regtry(reginfo, s))) goto got_it; break; case ALNUM: - if (do_utf8) { - LOAD_UTF8_CHARCLASS(alnum,"a"); - while (s + (uskip = UTF8SKIP(s)) <= strend) { - if (swash_fetch(PL_utf8_alnum, (U8*)s, do_utf8)) { - if (tmp && (norun || regtry(prog, s))) - goto got_it; - else - tmp = doevery; - } - else - tmp = 1; - s += uskip; - } - } - else { - while (s < strend) { - if (isALNUM(*s)) { - if (tmp && (norun || regtry(prog, s))) - goto got_it; - else - tmp = doevery; - } - else - tmp = 1; - s++; - } - } - break; + REXEC_FBC_CSCAN_PRELOAD( + LOAD_UTF8_CHARCLASS_ALNUM(), + swash_fetch(PL_utf8_alnum, (U8*)s, do_utf8), + isALNUM(*s) + ); case ALNUML: - PL_reg_flags |= RF_tainted; - if (do_utf8) { - while (s + (uskip = UTF8SKIP(s)) <= strend) { - if (isALNUM_LC_utf8((U8*)s)) { - if (tmp && (norun || regtry(prog, s))) - goto got_it; - else - tmp = doevery; - } - else - tmp = 1; - s += uskip; - } - } - else { - while (s < strend) { - if (isALNUM_LC(*s)) { - if (tmp && (norun || regtry(prog, s))) - goto got_it; - else - tmp = doevery; - } - else - tmp = 1; - s++; - } - } - break; + REXEC_FBC_CSCAN_TAINT( + isALNUM_LC_utf8((U8*)s), + isALNUM_LC(*s) + ); case NALNUM: - if (do_utf8) { - LOAD_UTF8_CHARCLASS(alnum,"a"); - while (s + (uskip = UTF8SKIP(s)) <= strend) { - if (!swash_fetch(PL_utf8_alnum, (U8*)s, do_utf8)) { - if (tmp && (norun || regtry(prog, s))) - goto got_it; - else - tmp = doevery; - } - else - tmp = 1; - s += uskip; - } - } - else { - while (s < strend) { - if (!isALNUM(*s)) { - if (tmp && (norun || regtry(prog, s))) - goto got_it; - else - tmp = doevery; - } - else - tmp = 1; - s++; - } - } - break; + REXEC_FBC_CSCAN_PRELOAD( + LOAD_UTF8_CHARCLASS_ALNUM(), + !swash_fetch(PL_utf8_alnum, (U8*)s, do_utf8), + !isALNUM(*s) + ); case NALNUML: - PL_reg_flags |= RF_tainted; - if (do_utf8) { - while (s + (uskip = UTF8SKIP(s)) <= strend) { - if (!isALNUM_LC_utf8((U8*)s)) { - if (tmp && (norun || regtry(prog, s))) - goto got_it; - else - tmp = doevery; - } - else - tmp = 1; - s += uskip; - } - } - else { - while (s < strend) { - if (!isALNUM_LC(*s)) { - if (tmp && (norun || regtry(prog, s))) - goto got_it; - else - tmp = doevery; - } - else - tmp = 1; - s++; - } - } - break; + REXEC_FBC_CSCAN_TAINT( + !isALNUM_LC_utf8((U8*)s), + !isALNUM_LC(*s) + ); case SPACE: - if (do_utf8) { - LOAD_UTF8_CHARCLASS(space," "); - while (s + (uskip = UTF8SKIP(s)) <= strend) { - if (*s == ' ' || swash_fetch(PL_utf8_space,(U8*)s, do_utf8)) { - if (tmp && (norun || regtry(prog, s))) - goto got_it; - else - tmp = doevery; - } - else - tmp = 1; - s += uskip; - } - } - else { - while (s < strend) { - if (isSPACE(*s)) { - if (tmp && (norun || regtry(prog, s))) - goto got_it; - else - tmp = doevery; - } - else - tmp = 1; - s++; - } - } - break; + REXEC_FBC_CSCAN_PRELOAD( + LOAD_UTF8_CHARCLASS_SPACE(), + *s == ' ' || swash_fetch(PL_utf8_space,(U8*)s, do_utf8), + isSPACE(*s) + ); case SPACEL: - PL_reg_flags |= RF_tainted; - if (do_utf8) { - while (s + (uskip = UTF8SKIP(s)) <= strend) { - if (*s == ' ' || isSPACE_LC_utf8((U8*)s)) { - if (tmp && (norun || regtry(prog, s))) - goto got_it; - else - tmp = doevery; - } - else - tmp = 1; - s += uskip; - } - } - else { - while (s < strend) { - if (isSPACE_LC(*s)) { - if (tmp && (norun || regtry(prog, s))) - goto got_it; - else - tmp = doevery; - } - else - tmp = 1; - s++; - } - } - break; + REXEC_FBC_CSCAN_TAINT( + *s == ' ' || isSPACE_LC_utf8((U8*)s), + isSPACE_LC(*s) + ); case NSPACE: - if (do_utf8) { - LOAD_UTF8_CHARCLASS(space," "); - while (s + (uskip = UTF8SKIP(s)) <= strend) { - if (!(*s == ' ' || swash_fetch(PL_utf8_space,(U8*)s, do_utf8))) { - if (tmp && (norun || regtry(prog, s))) - goto got_it; - else - tmp = doevery; - } - else - tmp = 1; - s += uskip; - } - } - else { - while (s < strend) { - if (!isSPACE(*s)) { - if (tmp && (norun || regtry(prog, s))) - goto got_it; - else - tmp = doevery; - } - else - tmp = 1; - s++; - } - } - break; + REXEC_FBC_CSCAN_PRELOAD( + LOAD_UTF8_CHARCLASS_SPACE(), + !(*s == ' ' || swash_fetch(PL_utf8_space,(U8*)s, do_utf8)), + !isSPACE(*s) + ); case NSPACEL: - PL_reg_flags |= RF_tainted; - if (do_utf8) { - while (s + (uskip = UTF8SKIP(s)) <= strend) { - if (!(*s == ' ' || isSPACE_LC_utf8((U8*)s))) { - if (tmp && (norun || regtry(prog, s))) - goto got_it; - else - tmp = doevery; - } - else - tmp = 1; - s += uskip; - } - } - else { - while (s < strend) { - if (!isSPACE_LC(*s)) { - if (tmp && (norun || regtry(prog, s))) - goto got_it; - else - tmp = doevery; - } - else - tmp = 1; - s++; - } - } - break; + REXEC_FBC_CSCAN_TAINT( + !(*s == ' ' || isSPACE_LC_utf8((U8*)s)), + !isSPACE_LC(*s) + ); case DIGIT: - if (do_utf8) { - LOAD_UTF8_CHARCLASS(digit,"0"); - while (s + (uskip = UTF8SKIP(s)) <= strend) { - if (swash_fetch(PL_utf8_digit,(U8*)s, do_utf8)) { - if (tmp && (norun || regtry(prog, s))) - goto got_it; - else - tmp = doevery; - } - else - tmp = 1; - s += uskip; - } - } - else { - while (s < strend) { - if (isDIGIT(*s)) { - if (tmp && (norun || regtry(prog, s))) - goto got_it; - else - tmp = doevery; - } - else - tmp = 1; - s++; - } - } - break; + REXEC_FBC_CSCAN_PRELOAD( + LOAD_UTF8_CHARCLASS_DIGIT(), + swash_fetch(PL_utf8_digit,(U8*)s, do_utf8), + isDIGIT(*s) + ); case DIGITL: - PL_reg_flags |= RF_tainted; - if (do_utf8) { - while (s + (uskip = UTF8SKIP(s)) <= strend) { - if (isDIGIT_LC_utf8((U8*)s)) { - if (tmp && (norun || regtry(prog, s))) - goto got_it; - else - tmp = doevery; - } - else - tmp = 1; - s += uskip; - } - } - else { - while (s < strend) { - if (isDIGIT_LC(*s)) { - if (tmp && (norun || regtry(prog, s))) - goto got_it; - else - tmp = doevery; - } - else - tmp = 1; - s++; - } - } - break; + REXEC_FBC_CSCAN_TAINT( + isDIGIT_LC_utf8((U8*)s), + isDIGIT_LC(*s) + ); case NDIGIT: - if (do_utf8) { - LOAD_UTF8_CHARCLASS(digit,"0"); - while (s + (uskip = UTF8SKIP(s)) <= strend) { - if (!swash_fetch(PL_utf8_digit,(U8*)s, do_utf8)) { - if (tmp && (norun || regtry(prog, s))) - goto got_it; - else - tmp = doevery; - } - else - tmp = 1; - s += uskip; - } - } - else { - while (s < strend) { - if (!isDIGIT(*s)) { - if (tmp && (norun || regtry(prog, s))) - goto got_it; - else - tmp = doevery; - } - else - tmp = 1; - s++; - } - } - break; + REXEC_FBC_CSCAN_PRELOAD( + LOAD_UTF8_CHARCLASS_DIGIT(), + !swash_fetch(PL_utf8_digit,(U8*)s, do_utf8), + !isDIGIT(*s) + ); case NDIGITL: - PL_reg_flags |= RF_tainted; - if (do_utf8) { - while (s + (uskip = UTF8SKIP(s)) <= strend) { - if (!isDIGIT_LC_utf8((U8*)s)) { - if (tmp && (norun || regtry(prog, s))) - goto got_it; - else - tmp = doevery; - } - else - tmp = 1; - s += uskip; - } - } - else { - while (s < strend) { - if (!isDIGIT_LC(*s)) { - if (tmp && (norun || regtry(prog, s))) - goto got_it; - else - tmp = doevery; - } - else - tmp = 1; - s++; - } + REXEC_FBC_CSCAN_TAINT( + !isDIGIT_LC_utf8((U8*)s), + !isDIGIT_LC(*s) + ); + case AHOCORASICKC: + case AHOCORASICK: + { + const enum { trie_plain, trie_utf8, trie_utf8_fold } + trie_type = do_utf8 ? + (c->flags == EXACT ? trie_utf8 : trie_utf8_fold) + : 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; + + const char *last_start = strend - trie->minlen; +#ifdef DEBUGGING + const char *real_start = s; +#endif + STRLEN maxlen = trie->maxlen; + SV *sv_points; + U8 **points; /* map of where we were in the input string + when reading a given char. For ASCII this + is unnecessary overhead as the relationship + is always 1:1, but for unicode, especially + case folded unicode this is not true. */ + U8 foldbuf[ UTF8_MAXBYTES_CASE + 1 ]; + U8 *bitmap=NULL; + + + GET_RE_DEBUG_FLAGS_DECL; + + /* We can't just allocate points here. We need to wrap it in + * an SV so it gets freed properly if there is a croak while + * running the match */ + ENTER; + SAVETMPS; + sv_points=newSV(maxlen * sizeof(U8 *)); + SvCUR_set(sv_points, + maxlen * sizeof(U8 *)); + SvPOK_on(sv_points); + sv_2mortal(sv_points); + points=(U8**)SvPV_nolen(sv_points ); + if ( trie_type != trie_utf8_fold + && (trie->bitmap || OP(c)==AHOCORASICKC) ) + { + if (trie->bitmap) + bitmap=(U8*)trie->bitmap; + else + bitmap=(U8*)ANYOF_BITMAP(c); + } + /* this is the Aho-Corasick algorithm modified a touch + to include special handling for long "unknown char" + sequences. The basic idea being that we use AC as long + as we are dealing with a possible matching char, when + we encounter an unknown char (and we have not encountered + an accepting state) we scan forward until we find a legal + starting char. + AC matching is basically that of trie matching, except + that when we encounter a failing transition, we fall back + to the current states "fail state", and try the current char + again, a process we repeat until we reach the root state, + state 1, or a legal transition. If we fail on the root state + then we can either terminate if we have reached an accepting + state previously, or restart the entire process from the beginning + if we have not. + + */ + while (s <= last_start) { + const U32 uniflags = UTF8_ALLOW_DEFAULT; + U8 *uc = (U8*)s; + U16 charid = 0; + U32 base = 1; + U32 state = 1; + UV uvc = 0; + STRLEN len = 0; + STRLEN foldlen = 0; + U8 *uscan = (U8*)NULL; + U8 *leftmost = NULL; +#ifdef DEBUGGING + U32 accepted_word= 0; +#endif + U32 pointpos = 0; + + while ( state && uc <= (U8*)strend ) { + int failed=0; + U32 word = aho->states[ state ].wordnum; + + if( state==1 ) { + if ( bitmap ) { + DEBUG_TRIE_EXECUTE_r( + if ( uc <= (U8*)last_start && !BITMAP_TEST(bitmap,*uc) ) { + dump_exec_pos( (char *)uc, c, strend, real_start, + (char *)uc, do_utf8 ); + PerlIO_printf( Perl_debug_log, + " Scanning for legal start char...\n"); + } + ); + while ( uc <= (U8*)last_start && !BITMAP_TEST(bitmap,*uc) ) { + uc++; + } + s= (char *)uc; + } + if (uc >(U8*)last_start) break; + } + + if ( word ) { + U8 *lpos= points[ (pointpos - trie->wordlen[word-1] ) % maxlen ]; + if (!leftmost || lpos < leftmost) { + DEBUG_r(accepted_word=word); + leftmost= lpos; + } + if (base==0) break; + + } + points[pointpos++ % maxlen]= uc; + REXEC_TRIE_READ_CHAR(trie_type, trie, uc, uscan, len, + uvc, charid, foldlen, foldbuf, uniflags); + DEBUG_TRIE_EXECUTE_r({ + dump_exec_pos( (char *)uc, c, strend, real_start, + s, do_utf8 ); + PerlIO_printf(Perl_debug_log, + " Charid:%3u CP:%4"UVxf" ", + charid, uvc); + }); + + do { +#ifdef DEBUGGING + word = aho->states[ state ].wordnum; +#endif + base = aho->states[ state ].trans.base; + + DEBUG_TRIE_EXECUTE_r({ + if (failed) + dump_exec_pos( (char *)uc, c, strend, real_start, + s, do_utf8 ); + PerlIO_printf( Perl_debug_log, + "%sState: %4"UVxf", word=%"UVxf, + failed ? " Fail transition to " : "", + (UV)state, (UV)word); + }); + if ( base ) { + U32 tmp; + if (charid && + (base + charid > trie->uniquecharcount ) + && (base + charid - 1 - trie->uniquecharcount + < trie->lasttrans) + && trie->trans[base + charid - 1 - + trie->uniquecharcount].check == state + && (tmp=trie->trans[base + charid - 1 - + trie->uniquecharcount ].next)) + { + DEBUG_TRIE_EXECUTE_r( + PerlIO_printf( Perl_debug_log," - legal\n")); + state = tmp; + break; + } + else { + DEBUG_TRIE_EXECUTE_r( + PerlIO_printf( Perl_debug_log," - fail\n")); + failed = 1; + state = aho->fail[state]; + } + } + else { + /* we must be accepting here */ + DEBUG_TRIE_EXECUTE_r( + PerlIO_printf( Perl_debug_log," - accepting\n")); + failed = 1; + break; + } + } while(state); + uc += len; + if (failed) { + if (leftmost) + break; + if (!state) state = 1; + } + } + if ( aho->states[ state ].wordnum ) { + U8 *lpos = points[ (pointpos - trie->wordlen[aho->states[ state ].wordnum-1]) % maxlen ]; + if (!leftmost || lpos < leftmost) { + DEBUG_r(accepted_word=aho->states[ state ].wordnum); + leftmost = lpos; + } + } + if (leftmost) { + 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 + ); + }); + if (!reginfo || regtry(reginfo, s)) { + FREETMPS; + LEAVE; + goto got_it; + } + s = HOPc(s,1); + DEBUG_TRIE_EXECUTE_r({ + PerlIO_printf( Perl_debug_log,"Pattern failed. Looking for new start point...\n"); + }); + } else { + DEBUG_TRIE_EXECUTE_r( + PerlIO_printf( Perl_debug_log,"No match.\n")); + break; + } + } + FREETMPS; + LEAVE; } break; default: @@ -1628,34 +1633,24 @@ Perl_regexec_flags(pTHX_ register regexp *prog, char *stringarg, register char * /* data: May be used for some additional optimizations. */ /* nosave: For optimizations. */ { + dVAR; register char *s; register regnode *c; 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 start_shift = 0; */ /* Offset of the start to find - constant substr. */ /* CC */ I32 end_shift = 0; /* Same for the end. */ /* CC */ I32 scream_pos = -1; /* Internal iterator of scream. */ - char *scream_olds; - SV* oreplsv = GvSV(PL_replgv); - bool do_utf8 = DO_UTF8(sv); - I32 multiline = prog->reganch & PMf_MULTILINE; -#ifdef DEBUGGING - SV *dsv0 = PERL_DEBUG_PAD_ZERO(0); - SV *dsv1 = PERL_DEBUG_PAD_ZERO(1); -#endif + char *scream_olds = NULL; + SV* const oreplsv = GvSV(PL_replgv); + const bool do_utf8 = (bool)DO_UTF8(sv); + I32 multiline; - GET_RE_DEBUG_FLAGS_DECL; - - RX_MATCH_UTF8_set(prog,do_utf8); + regmatch_info reginfo; /* create some info to pass to regtry etc */ - PL_regcc = 0; + GET_RE_DEBUG_FLAGS_DECL; - cache_re(prog); -#ifdef DEBUGGING - PL_regnarrate = DEBUG_r_TEST; -#endif + PERL_UNUSED_ARG(data); /* Be paranoid... */ if (prog == NULL || startpos == NULL) { @@ -1663,13 +1658,24 @@ Perl_regexec_flags(pTHX_ register regexp *prog, char *stringarg, register char * return 0; } + multiline = prog->reganch & PMf_MULTILINE; + reginfo.prog = prog; + + RX_MATCH_UTF8_set(prog, do_utf8); + DEBUG_EXECUTE_r( + debug_start_match(prog, do_utf8, startpos, strend, + "Matching"); + ); + minlen = prog->minlen; - if (strend - startpos < minlen) { + + if (strend - startpos < (minlen+(prog->check_offset_min<0?prog->check_offset_min:0))) { DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "String too short [regexec_flags]...\n")); goto phooey; } + /* Check validity of program. */ if (UCHARAT(prog->program) != REG_MAGIC) { Perl_croak(aTHX_ "corrupted regexp program"); @@ -1683,43 +1689,40 @@ Perl_regexec_flags(pTHX_ register regexp *prog, char *stringarg, register char * PL_reg_flags |= RF_utf8; /* Mark beginning of line for ^ and lookbehind. */ - PL_regbol = startpos; + reginfo.bol = startpos; /* XXX not used ??? */ PL_bostr = strbeg; - PL_reg_sv = sv; + reginfo.sv = sv; /* Mark end of line for $ (and such) */ PL_regeol = strend; /* see how far we have to get to not match where we matched before */ - PL_regtill = startpos+minend; - - /* We start without call_cc context. */ - PL_reg_call_cc = 0; + reginfo.till = startpos+minend; /* If there is a "must appear" string, look for it. */ s = startpos; - if (prog->reganch & ROPT_GPOS_SEEN) { /* Need to have PL_reg_ganch */ + if (prog->reganch & ROPT_GPOS_SEEN) { /* Need to set reginfo->ganch */ MAGIC *mg; if (flags & REXEC_IGNOREPOS) /* Means: check only at start */ - PL_reg_ganch = startpos; + reginfo.ganch = startpos; else if (sv && SvTYPE(sv) >= SVt_PVMG && SvMAGIC(sv) && (mg = mg_find(sv, PERL_MAGIC_regex_global)) && mg->mg_len >= 0) { - PL_reg_ganch = strbeg + mg->mg_len; /* Defined pos() */ + reginfo.ganch = strbeg + mg->mg_len; /* Defined pos() */ if (prog->reganch & ROPT_ANCH_GPOS) { - if (s > PL_reg_ganch) + if (s > reginfo.ganch) goto phooey; - s = PL_reg_ganch; + s = reginfo.ganch; } } else /* pos() not defined */ - PL_reg_ganch = strbeg; + reginfo.ganch = strbeg; } - if (!(flags & REXEC_CHECKED) && (prog->check_substr != Nullsv || prog->check_utf8 != Nullsv)) { + if (!(flags & REXEC_CHECKED) && (prog->check_substr != NULL || prog->check_utf8 != NULL)) { re_scream_pos_data d; d.scream_olds = &scream_olds; @@ -1731,34 +1734,12 @@ Perl_regexec_flags(pTHX_ register regexp *prog, char *stringarg, register char * } } - DEBUG_EXECUTE_r({ - char *s0 = UTF ? - pv_uni_display(dsv0, (U8*)prog->precomp, prog->prelen, 60, - UNI_DISPLAY_REGEX) : - prog->precomp; - int len0 = UTF ? SvCUR(dsv0) : prog->prelen; - char *s1 = do_utf8 ? sv_uni_display(dsv1, sv, 60, - UNI_DISPLAY_REGEX) : startpos; - int len1 = do_utf8 ? SvCUR(dsv1) : strend - startpos; - if (!PL_colorset) - reginitcolors(); - PerlIO_printf(Perl_debug_log, - "%sMatching REx%s `%s%*.*s%s%s' against `%s%.*s%s%s'\n", - PL_colors[4],PL_colors[5],PL_colors[0], - len0, len0, s0, - PL_colors[1], - len0 > 60 ? "..." : "", - PL_colors[0], - (int)(len1 > 60 ? 60 : len1), - s1, PL_colors[1], - (len1 > 60 ? "..." : "") - ); - }); + /* 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(prog, startpos)) + if (s == startpos && regtry(®info, startpos)) goto got_it; else if (multiline || (prog->reganch & ROPT_IMPLICIT) || (prog->reganch & ROPT_ANCH_MBOL)) /* XXXX SBOL? */ @@ -1773,7 +1754,7 @@ Perl_regexec_flags(pTHX_ register regexp *prog, char *stringarg, register char * if (s == startpos) goto after_try; while (1) { - if (regtry(prog, s)) + if (regtry(®info, s)) goto got_it; after_try: if (s >= end) @@ -1791,15 +1772,19 @@ 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(prog, s)) + if (regtry(®info, s)) goto got_it; } } } } goto phooey; - } else if (prog->reganch & ROPT_ANCH_GPOS) { - if (regtry(prog, PL_reg_ganch)) + } else if (ROPT_GPOS_CHECK == (prog->reganch & ROPT_GPOS_CHECK)) + { + /* the warning about reginfo.ganch being used without intialization + is bogus -- we set it above, when prog->reganch & ROPT_GPOS_SEEN + and we only enter this block when the same bit is set. */ + if (regtry(®info, reginfo.ganch)) goto got_it; goto phooey; } @@ -1814,41 +1799,38 @@ Perl_regexec_flags(pTHX_ register regexp *prog, char *stringarg, register char * #endif if (!(do_utf8 ? prog->anchored_utf8 : prog->anchored_substr)) do_utf8 ? to_utf8_substr(prog) : to_byte_substr(prog); - ch = SvPVX(do_utf8 ? prog->anchored_utf8 : prog->anchored_substr)[0]; + ch = SvPVX_const(do_utf8 ? prog->anchored_utf8 : prog->anchored_substr)[0]; if (do_utf8) { - while (s < strend) { + REXEC_FBC_SCAN( if (*s == ch) { DEBUG_EXECUTE_r( did_match = 1 ); - if (regtry(prog, s)) goto got_it; + if (regtry(®info, s)) goto got_it; s += UTF8SKIP(s); while (s < strend && *s == ch) s += UTF8SKIP(s); } - s += UTF8SKIP(s); - } + ); } else { - while (s < strend) { + REXEC_FBC_SCAN( if (*s == ch) { DEBUG_EXECUTE_r( did_match = 1 ); - if (regtry(prog, s)) goto got_it; + if (regtry(®info, s)) goto got_it; s++; while (s < strend && *s == ch) s++; } - s++; - } + ); } DEBUG_EXECUTE_r(if (!did_match) PerlIO_printf(Perl_debug_log, "Did not find anchored character...\n") ); } - /*SUPPRESS 560*/ - else if (prog->anchored_substr != Nullsv - || prog->anchored_utf8 != Nullsv - || ((prog->float_substr != Nullsv || prog->float_utf8 != Nullsv) + else if (prog->anchored_substr != NULL + || prog->anchored_utf8 != NULL + || ((prog->float_substr != NULL || prog->float_utf8 != NULL) && prog->float_max_offset < strend - s)) { SV *must; I32 back_max; @@ -1870,107 +1852,101 @@ Perl_regexec_flags(pTHX_ register regexp *prog, char *stringarg, register char * back_max = prog->float_max_offset; back_min = prog->float_min_offset; } + + if (must == &PL_sv_undef) /* could not downgrade utf8 check substring, so must fail */ goto phooey; - last = HOP3c(strend, /* Cannot start after this */ - -(I32)(CHR_SVLEN(must) - - (SvTAIL(must) != 0) + back_min), strbeg); - + if (back_min<0) { + last = strend; + } else { + last = HOP3c(strend, /* Cannot start after this */ + -(I32)(CHR_SVLEN(must) + - (SvTAIL(must) != 0) + back_min), strbeg); + } if (s > PL_bostr) last1 = HOPc(s, -1); else last1 = s - 1; /* bogus */ - /* XXXX check_substr already used to find `s', can optimize if + /* XXXX check_substr already used to find "s", can optimize if check_substr==must. */ scream_pos = -1; dontbother = end_shift; strend = HOPc(strend, -dontbother); while ( (s <= last) && ((flags & REXEC_SCREAM) - ? (s = screaminstr(sv, must, HOP3c(s, back_min, strend) - strbeg, + ? (s = screaminstr(sv, must, HOP3c(s, back_min, (back_min<0 ? strbeg : strend)) - strbeg, end_shift, &scream_pos, 0)) - : (s = fbm_instr((unsigned char*)HOP3(s, back_min, strend), + : (s = fbm_instr((unsigned char*)HOP3(s, back_min, (back_min<0 ? strbeg : strend)), (unsigned char*)strend, must, multiline ? FBMrf_MULTILINE : 0))) ) { /* we may be pointing at the wrong string */ if ((flags & REXEC_SCREAM) && RX_MATCH_COPIED(prog)) - s = strbeg + (s - SvPVX(sv)); + s = strbeg + (s - SvPVX_const(sv)); DEBUG_EXECUTE_r( did_match = 1 ); if (HOPc(s, -back_max) > last1) { last1 = HOPc(s, -back_min); s = HOPc(s, -back_max); } else { - char *t = (last1 >= PL_bostr) ? HOPc(last1, 1) : last1 + 1; + char * const t = (last1 >= PL_bostr) ? HOPc(last1, 1) : last1 + 1; last1 = HOPc(s, -back_min); - s = t; + s = t; } if (do_utf8) { while (s <= last1) { - if (regtry(prog, s)) + if (regtry(®info, s)) goto got_it; s += UTF8SKIP(s); } } else { while (s <= last1) { - if (regtry(prog, s)) + if (regtry(®info, s)) goto got_it; s++; } } } - DEBUG_EXECUTE_r(if (!did_match) - PerlIO_printf(Perl_debug_log, - "Did not find %s substr `%s%.*s%s'%s...\n", + DEBUG_EXECUTE_r(if (!did_match) { + RE_PV_QUOTED_DECL(quoted, do_utf8, PERL_DEBUG_PAD_ZERO(0), + SvPVX_const(must), RE_SV_DUMPLEN(must), 30); + PerlIO_printf(Perl_debug_log, "Did not find %s substr %s%s...\n", ((must == prog->anchored_substr || must == prog->anchored_utf8) ? "anchored" : "floating"), - PL_colors[0], - (int)(SvCUR(must) - (SvTAIL(must)!=0)), - SvPVX(must), - PL_colors[1], (SvTAIL(must) ? "$" : "")) - ); + quoted, RE_SV_TAIL(must)); + }); goto phooey; } - else if ((c = prog->regstclass)) { + else if ( (c = prog->regstclass) ) { if (minlen) { - I32 op = (U8)OP(prog->regstclass); + const OPCODE op = OP(prog->regstclass); /* don't bother with what can't match */ - if (PL_regkind[op] != EXACT && op != CANY) + if (PL_regkind[op] != EXACT && op != CANY && PL_regkind[op] != TRIE) strend = HOPc(strend, -(minlen - 1)); } DEBUG_EXECUTE_r({ - SV *prop = sv_newmortal(); - char *s0; - char *s1; - int len0; - int len1; - - regprop(prop, c); - s0 = UTF ? - pv_uni_display(dsv0, (U8*)SvPVX(prop), SvCUR(prop), 60, - UNI_DISPLAY_REGEX) : - SvPVX(prop); - len0 = UTF ? SvCUR(dsv0) : SvCUR(prop); - s1 = UTF ? - sv_uni_display(dsv1, sv, 60, UNI_DISPLAY_REGEX) : s; - len1 = UTF ? SvCUR(dsv1) : strend - s; - PerlIO_printf(Perl_debug_log, - "Matching stclass `%*.*s' against `%*.*s'\n", - len0, len0, s0, - len1, len1, s1); + SV * const prop = sv_newmortal(); + regprop(prog, prop, c); + { + RE_PV_QUOTED_DECL(quoted,UTF,PERL_DEBUG_PAD_ZERO(1), + s,strend-s,60); + PerlIO_printf(Perl_debug_log, + "Matching stclass %.*s against %s (%d chars)\n", + (int)SvCUR(prop), SvPVX_const(prop), + quoted, (int)(strend - s)); + } }); - if (find_byclass(prog, c, s, strend, startpos, 0)) + if (find_byclass(prog, c, s, strend, ®info)) goto got_it; - DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "Contradicts stclass...\n")); + DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "Contradicts stclass... [regexec_flags]\n")); } else { dontbother = 0; - if (prog->float_substr != Nullsv || prog->float_utf8 != Nullsv) { + if (prog->float_substr != NULL || prog->float_utf8 != NULL) { /* Trim the end. */ char *last; SV* float_real; @@ -1986,18 +1962,18 @@ Perl_regexec_flags(pTHX_ register regexp *prog, char *stringarg, register char * last = scream_olds; /* Only one occurrence. */ /* we may be pointing at the wrong string */ else if (RX_MATCH_COPIED(prog)) - s = strbeg + (s - SvPVX(sv)); + s = strbeg + (s - SvPVX_const(sv)); } else { STRLEN len; - char *little = SvPV(float_real, len); + const char * const little = SvPV_const(float_real, len); if (SvTAIL(float_real)) { if (memEQ(strend - len + 1, little, len - 1)) last = strend - len + 1; else if (!multiline) last = memEQ(strend - len, little, len) - ? strend - len : Nullch; + ? strend - len : NULL; else goto find_last; } else { @@ -2005,13 +1981,14 @@ Perl_regexec_flags(pTHX_ register regexp *prog, char *stringarg, register char * if (len) last = rninstr(s, strend, little, little + len); else - last = strend; /* matching `$' */ + last = strend; /* matching "$" */ } } if (last == NULL) { - DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, - "%sCan't trim the tail, match fails (should not happen)%s\n", - PL_colors[4],PL_colors[5])); + DEBUG_EXECUTE_r( + PerlIO_printf(Perl_debug_log, + "%sCan't trim the tail, match fails (should not happen)%s\n", + PL_colors[4], PL_colors[5])); goto phooey; /* Should not happen! */ } dontbother = strend - last + prog->float_min_offset; @@ -2022,7 +1999,7 @@ Perl_regexec_flags(pTHX_ register regexp *prog, char *stringarg, register char * /* We don't know much -- general case. */ if (do_utf8) { for (;;) { - if (regtry(prog, s)) + if (regtry(®info, s)) goto got_it; if (s >= strend) break; @@ -2031,7 +2008,7 @@ Perl_regexec_flags(pTHX_ register regexp *prog, char *stringarg, register char * } else { do { - if (regtry(prog, s)) + if (regtry(®info, s)) goto got_it; } while (s++ < strend); } @@ -2049,15 +2026,17 @@ got_it: sv_setsv(oreplsv, GvSV(PL_replgv));/* So that when GvSV(replgv) is restored, the value remains the same. */ - restore_pos(aTHX_ 0); + 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) ) { RX_MATCH_COPY_FREE(prog); if (flags & REXEC_COPY_STR) { - I32 i = PL_regeol - startpos + (stringarg - strbeg); -#ifdef PERL_COPY_ON_WRITE + const I32 i = PL_regeol - startpos + (stringarg - strbeg); +#ifdef PERL_OLD_COPY_ON_WRITE if ((SvIsCOW(sv) || (SvFLAGS(sv) & CAN_COW_MASK) == CAN_COW_FLAGS)) { if (DEBUG_C_TEST) { @@ -2066,7 +2045,7 @@ got_it: (int) SvTYPE(sv)); } prog->saved_copy = sv_setsv_cow(prog->saved_copy, sv); - prog->subbeg = SvPVX(prog->saved_copy); + prog->subbeg = (char *)SvPVX_const(prog->saved_copy); assert (SvPOKp(prog->saved_copy)); } else #endif @@ -2087,27 +2066,26 @@ got_it: phooey: DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "%sMatch failed%s\n", - PL_colors[4],PL_colors[5])); + PL_colors[4], PL_colors[5])); if (PL_reg_eval_set) - restore_pos(aTHX_ 0); + restore_pos(aTHX_ prog); return 0; } + /* - regtry - try match at specific point */ STATIC I32 /* 0 failure, 1 success */ -S_regtry(pTHX_ regexp *prog, char *startpos) +S_regtry(pTHX_ const regmatch_info *reginfo, char *startpos) { - register I32 i; + dVAR; register I32 *sp; register I32 *ep; CHECKPOINT lastcp; + regexp *prog = reginfo->prog; GET_RE_DEBUG_FLAGS_DECL; -#ifdef DEBUGGING - PL_regindent = 0; /* XXXX Not good when matches are reenterable... */ -#endif if ((prog->reganch & ROPT_EVAL_SEEN) && !PL_reg_eval_set) { MAGIC *mg; @@ -2116,7 +2094,7 @@ S_regtry(pTHX_ regexp *prog, 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; @@ -2124,30 +2102,33 @@ S_regtry(pTHX_ regexp *prog, char *startpos) /* SAVEI8(cxstack[cxstack_ix].blk_gimme); cxstack[cxstack_ix].blk_gimme = G_SCALAR; */ - if (PL_reg_sv) { + if (reginfo->sv) { /* Make $_ available to executed code. */ - if (PL_reg_sv != DEFSV) { + if (reginfo->sv != DEFSV) { SAVE_DEFSV; - DEFSV = PL_reg_sv; + DEFSV = reginfo->sv; } - if (!(SvTYPE(PL_reg_sv) >= SVt_PVMG && SvMAGIC(PL_reg_sv) - && (mg = mg_find(PL_reg_sv, PERL_MAGIC_regex_global)))) { + if (!(SvTYPE(reginfo->sv) >= SVt_PVMG && SvMAGIC(reginfo->sv) + && (mg = mg_find(reginfo->sv, PERL_MAGIC_regex_global)))) { /* prepare for quick setting of pos */ - sv_magic(PL_reg_sv, (SV*)0, - PERL_MAGIC_regex_global, Nullch, 0); - mg = mg_find(PL_reg_sv, PERL_MAGIC_regex_global); +#ifdef PERL_OLD_COPY_ON_WRITE + if (SvIsCOW(sv)) + sv_force_normal_flags(sv, 0); +#endif + mg = sv_magicext(reginfo->sv, NULL, PERL_MAGIC_regex_global, + &PL_vtbl_mglob, NULL, 0); mg->mg_len = -1; } PL_reg_magic = mg; PL_reg_oldpos = mg->mg_len; - SAVEDESTRUCTOR_X(restore_pos, 0); + SAVEDESTRUCTOR_X(restore_pos, prog); } if (!PL_reg_curpm) { - Newz(22,PL_reg_curpm, 1, PMOP); + Newxz(PL_reg_curpm, 1, PMOP); #ifdef USE_ITHREADS { - SV* repointer = newSViv(0); + SV* const repointer = newSViv(0); /* so we know which PL_regex_padav element is PL_reg_curpm */ SvFLAGS(repointer) |= SVf_BREAK; av_push(PL_regex_padav,repointer); @@ -2165,32 +2146,32 @@ S_regtry(pTHX_ regexp *prog, char *startpos) $` inside (?{}) could fail... */ PL_reg_oldsaved = prog->subbeg; PL_reg_oldsavedlen = prog->sublen; -#ifdef PERL_COPY_ON_WRITE +#ifdef PERL_OLD_COPY_ON_WRITE PL_nrs = prog->saved_copy; #endif RX_MATCH_COPIED_off(prog); } else - PL_reg_oldsaved = Nullch; + PL_reg_oldsaved = NULL; 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; - PL_regstartp = prog->startp; - PL_regendp = prog->endp; PL_reglastparen = &prog->lastparen; PL_reglastcloseparen = &prog->lastcloseparen; prog->lastparen = 0; prog->lastcloseparen = 0; PL_regsize = 0; - DEBUG_EXECUTE_r(PL_reg_starttry = startpos); + PL_regstartp = prog->startp; + PL_regendp = prog->endp; if (PL_reg_start_tmpl <= prog->nparens) { PL_reg_start_tmpl = prog->nparens*3/2 + 3; if(PL_reg_start_tmp) Renew(PL_reg_start_tmp, PL_reg_start_tmpl, char*); else - New(22,PL_reg_start_tmp, PL_reg_start_tmpl, char*); + Newx(PL_reg_start_tmp, PL_reg_start_tmpl, char*); } /* XXXX What this code is doing here?!!! There should be no need @@ -2211,6 +2192,7 @@ S_regtry(pTHX_ regexp *prog, char *startpos) sp = prog->startp; ep = prog->endp; if (prog->nparens) { + register I32 i; for (i = prog->nparens; i > (I32)*PL_reglastparen; i--) { *++sp = -1; *++ep = -1; @@ -2218,7 +2200,7 @@ S_regtry(pTHX_ regexp *prog, char *startpos) } #endif REGCP_SET(lastcp); - if (regmatch(prog->program + 1)) { + if (regmatch(reginfo, prog->program + 1)) { prog->endp[0] = PL_reginput - PL_bostr; return 1; } @@ -2226,162 +2208,401 @@ S_regtry(pTHX_ regexp *prog, char *startpos) return 0; } -#define RE_UNWIND_BRANCH 1 -#define RE_UNWIND_BRANCHJ 2 - -union re_unwind_t; - -typedef struct { /* XX: makes sense to enlarge it... */ - I32 type; - I32 prev; - CHECKPOINT lastcp; -} re_unwind_generic_t; - -typedef struct { - I32 type; - I32 prev; - CHECKPOINT lastcp; - I32 lastparen; - regnode *next; - char *locinput; - I32 nextchr; -#ifdef DEBUGGING - int regindent; -#endif -} re_unwind_branch_t; - -typedef union re_unwind_t { - I32 type; - re_unwind_generic_t generic; - re_unwind_branch_t branch; -} re_unwind_t; #define sayYES goto yes #define sayNO goto no -#define sayNO_ANYOF goto no_anyof -#define sayYES_FINAL goto yes_final -#define sayYES_LOUD goto yes_loud -#define sayNO_FINAL goto no_final -#define sayNO_SILENT goto do_no -#define saySAME(x) if (x) goto yes; else goto no +#define sayNO_SILENT goto no_silent + +/* we dont use STMT_START/END here because it leads to + "unreachable code" warnings, which are bogus, but distracting. */ +#define CACHEsayNO \ + 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 - 'failed...' are printed. Currently 29 makes these messages line - up with the opcode they refer to. Earlier perls used 25 which - left these messages outdented making reviewing a debug output - quite difficult. + 'failed...' are printed. It should be set such that messages + are inline with the regop output that created them. */ -#define REPORT_CODE_OFF 29 +#define REPORT_CODE_OFF 32 /* Make sure there is a test for this +1 options in re_tests */ #define TRIE_INITAL_ACCEPT_BUFFLEN 4; -#define TRIE_CHECK_STATE_IS_ACCEPTING STMT_START { \ - if ( trie->states[ state ].wordnum ) { \ - if ( !accepted ) { \ - ENTER; \ - SAVETMPS; \ - bufflen = TRIE_INITAL_ACCEPT_BUFFLEN ; \ - sv_accept_buff=NEWSV( 1234, \ - bufflen * sizeof(reg_trie_accepted) - 1 ); \ - SvCUR_set( sv_accept_buff, sizeof(reg_trie_accepted) ); \ - SvPOK_on( sv_accept_buff ); \ - sv_2mortal( sv_accept_buff ); \ - accept_buff = (reg_trie_accepted*)SvPV_nolen( sv_accept_buff );\ - } else { \ - if ( accepted >= bufflen ) { \ - bufflen *= 2; \ - accept_buff =(reg_trie_accepted*)SvGROW( sv_accept_buff, \ - bufflen * sizeof(reg_trie_accepted) ); \ - } \ - SvCUR_set( sv_accept_buff,SvCUR( sv_accept_buff ) \ - + sizeof( reg_trie_accepted ) ); \ - } \ - accept_buff[ accepted ].wordnum = trie->states[ state ].wordnum; \ - accept_buff[ accepted ].endpos = uc; \ - ++accepted; \ - } } STMT_END - -#define TRIE_HANDLE_CHAR STMT_START { \ - if ( uvc < 256 ) { \ - charid = trie->charmap[ uvc ]; \ - } else { \ - charid = 0; \ - if( trie->widecharmap ) { \ - SV** svpp = (SV**)NULL; \ - svpp = hv_fetch( trie->widecharmap, (char*)&uvc, \ - sizeof( UV ), 0 ); \ - if ( svpp ) { \ - charid = (U16)SvIV( *svpp ); \ - } \ - } \ - } \ - if ( charid && \ - ( base + charid - 1 - trie->uniquecharcount ) >=0 && \ - trie->trans[ base + charid - 1 - trie->uniquecharcount ].check == state ) \ - { \ - state = trie->trans[ base + charid - 1 - trie->uniquecharcount ].next; \ - } else { \ - state = 0; \ - } \ - uc += len; \ - } STMT_END +#define CHRTEST_UNINIT -1001 /* c1/c2 haven't been calculated yet */ +#define CHRTEST_VOID -1000 /* the c1/c2 "next char" test should be skipped */ + +#define SLAB_FIRST(s) (&(s)->states[0]) +#define SLAB_LAST(s) (&(s)->states[PERL_REGMATCH_SLAB_SLOTS-1]) + +/* grab a new slab and return the first slot in it */ + +STATIC regmatch_state * +S_push_slab(pTHX) +{ +#if PERL_VERSION < 9 + dMY_CXT; +#endif + regmatch_slab *s = PL_regmatch_slab->next; + if (!s) { + Newx(s, 1, regmatch_slab); + s->prev = PL_regmatch_slab; + s->next = NULL; + PL_regmatch_slab->next = s; + } + PL_regmatch_slab = s; + return SLAB_FIRST(s); +} + + +/* push a new state then goto it */ + +#define PUSH_STATE_GOTO(state, node) \ + scan = node; \ + st->resume_state = state; \ + goto push_state; + +/* push a new state with success backtracking, then goto it */ + +#define PUSH_YES_STATE_GOTO(state, node) \ + scan = node; \ + st->resume_state = state; \ + goto push_yes_state; + + /* - - 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. + +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 + ... + // 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 + +#undef ST + + ... + } + +For any old-timers reading this who are familiar with the old recursive +approach, the code above is equivalent to: + + case IFMATCH: // we are executing the IFMATCH op, (?=A)B + { + int foo = ... + ... + if (regmatch(A)) { + next = B; + bar = foo; + break; + } + ...; // do some housekeeping, then ... + sayNO; // propagate the failure + } + +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", \ + depth*2, "", \ + reg_name[st->resume_state] ); \ + }); + + +#define REG_NODE_NUM(x) ((x) ? (int)((x)-prog) : -1) + +#ifdef DEBUGGING + +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; + if (!PL_colorset) + reginitcolors(); + { + RE_PV_QUOTED_DECL(s0, utf8_pat, PERL_DEBUG_PAD_ZERO(0), + prog->precomp, prog->prelen, 60); + + RE_PV_QUOTED_DECL(s1, do_utf8, PERL_DEBUG_PAD_ZERO(1), + start, end - start, 60); + + PerlIO_printf(Perl_debug_log, + "%s%s REx%s %s against %s\n", + PL_colors[4], blurb, PL_colors[5], s0, s1); + + if (do_utf8||utf8_pat) + PerlIO_printf(Perl_debug_log, "UTF-8 %s%s%s...\n", + utf8_pat ? "pattern" : "", + utf8_pat && do_utf8 ? " and " : "", + do_utf8 ? "string" : "" + ); + } +} + +STATIC void +S_dump_exec_pos(pTHX_ const char *locinput, + const regnode *scan, + const char *loc_regeol, + const char *loc_bostr, + const char *loc_reg_starttry, + const bool do_utf8) +{ + const int docolor = *PL_colors[0] || *PL_colors[2] || *PL_colors[4]; + const int taill = (docolor ? 10 : 7); /* 3 chars for "> <" */ + int l = (loc_regeol - locinput) > taill ? taill : (loc_regeol - locinput); + /* The part of the string before starttry has one color + (pref0_len chars), between starttry and current + position another one (pref_len - pref0_len chars), + after the current position the third one. + We assume that pref0_len <= pref_len, otherwise we + decrease pref0_len. */ + int pref_len = (locinput - loc_bostr) > (5 + taill) - l + ? (5 + taill) - l : locinput - loc_bostr; + int pref0_len; + + while (do_utf8 && UTF8_IS_CONTINUATION(*(U8*)(locinput - pref_len))) + pref_len++; + pref0_len = pref_len - (locinput - loc_reg_starttry); + if (l + pref_len < (5 + taill) && l < loc_regeol - locinput) + l = ( loc_regeol - locinput > (5 + taill) - pref_len + ? (5 + taill) - pref_len : loc_regeol - locinput); + while (do_utf8 && UTF8_IS_CONTINUATION(*(U8*)(locinput + l))) + l--; + if (pref0_len < 0) + pref0_len = 0; + if (pref0_len > pref_len) + pref0_len = pref_len; + { + const int is_uni = (do_utf8 && OP(scan) != CANY) ? 1 : 0; + + RE_PV_COLOR_DECL(s0,len0,is_uni,PERL_DEBUG_PAD(0), + (locinput - pref_len),pref0_len, 60, 4, 5); + + RE_PV_COLOR_DECL(s1,len1,is_uni,PERL_DEBUG_PAD(1), + (locinput - pref_len + pref0_len), + pref_len - pref0_len, 60, 2, 3); + + RE_PV_COLOR_DECL(s2,len2,is_uni,PERL_DEBUG_PAD(2), + locinput, loc_regeol - locinput, 10, 0, 1); + + const STRLEN tlen=len0+len1+len2; + PerlIO_printf(Perl_debug_log, + "%4"IVdf" <%.*s%.*s%s%.*s>%*s|", + (IV)(locinput - loc_bostr), + len0, s0, + len1, s1, + (docolor ? "" : "> <"), + len2, s2, + (int)(tlen > 19 ? 0 : 19 - tlen), + ""); + } +} + +#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; + SV *sv_dat=(SV*)rex->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_ regnode *prog) +S_regmatch(pTHX_ const regmatch_info *reginfo, regnode *prog) { - register regnode *scan; /* Current node. */ - regnode *next; /* Next node. */ - regnode *inner; /* Next node in internal branch. */ - register I32 nextchr; /* renamed nextchr - nextchar colides with - function of same name */ - register I32 n; /* no or next */ - register I32 ln = 0; /* len or last */ - register char *s = Nullch; /* operand or save */ - register char *locinput = PL_reginput; - register I32 c1 = 0, c2 = 0, paren; /* case fold search, parenth */ - int minmod = 0, sw = 0, logical = 0; - I32 unwind = 0; - - /* used by the trie code */ - SV *sv_accept_buff; /* accepting states we have traversed */ - reg_trie_accepted *accept_buff; /* "" */ - reg_trie_data *trie; /* what trie are we using right now */ - U32 accepted = 0; /* how many accepting states we have seen*/ - -#if 0 - I32 firstcp = PL_savestack_ix; +#if PERL_VERSION < 9 + dMY_CXT; #endif - register bool do_utf8 = PL_reg_match_utf8; -#ifdef DEBUGGING - SV *dsv0 = PERL_DEBUG_PAD_ZERO(0); - SV *dsv1 = PERL_DEBUG_PAD_ZERO(1); - SV *dsv2 = PERL_DEBUG_PAD_ZERO(2); + dVAR; + register const bool do_utf8 = PL_reg_match_utf8; + const U32 uniflags = UTF8_ALLOW_DEFAULT; - SV *re_debug_flags; -#endif + regexp *rex = reginfo->prog; + + regmatch_slab *orig_slab; + regmatch_state *orig_state; - GET_RE_DEBUG_FLAGS; + /* the current state. This is a cached copy of PL_regmatch_state */ + register regmatch_state *st; + + /* cache heavy used fields of st in registers */ + register regnode *scan; + register regnode *next; + register I32 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; + register I32 nextchr; /* is always set to UCHARAT(locinput) */ + + bool result = 0; /* return value of S_regmatch */ + int depth = 0; /* depth of backtrack stack */ + int nochange_depth = 0; /* depth of RECURSE recursion with nochange*/ + regmatch_state *yes_state = NULL; /* state to pop to on success of + subpattern */ + regmatch_state *cur_eval = NULL; /* most recent EVAL_AB state */ + struct regmatch_state *cur_curlyx = NULL; /* most recent curlyx */ + U32 state_num; + + /* 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 - PL_regindent++; + GET_RE_DEBUG_FLAGS_DECL; #endif + /* on first ever call to regmatch, allocate first slab */ + if (!PL_regmatch_slab) { + Newx(PL_regmatch_slab, 1, regmatch_slab); + PL_regmatch_slab->prev = NULL; + PL_regmatch_slab->next = NULL; + PL_regmatch_state = SLAB_FIRST(PL_regmatch_slab); + } + + /* remember current high-water mark for exit */ + /* XXX this should be done with SAVE* instead */ + orig_slab = PL_regmatch_slab; + orig_state = PL_regmatch_state; + + /* grab next free state slot */ + st = ++PL_regmatch_state; + if (st > SLAB_LAST(PL_regmatch_slab)) + st = PL_regmatch_state = S_push_slab(aTHX); /* Note that nextchr is a byte even in UTF */ nextchr = UCHARAT(locinput); @@ -2389,79 +2610,30 @@ S_regmatch(pTHX_ regnode *prog) while (scan != NULL) { DEBUG_EXECUTE_r( { - SV *prop = sv_newmortal(); - int docolor = *PL_colors[0]; - int taill = (docolor ? 10 : 7); /* 3 chars for "> <" */ - int l = (PL_regeol - locinput) > taill ? taill : (PL_regeol - locinput); - /* The part of the string before starttry has one color - (pref0_len chars), between starttry and current - position another one (pref_len - pref0_len chars), - after the current position the third one. - We assume that pref0_len <= pref_len, otherwise we - decrease pref0_len. */ - int pref_len = (locinput - PL_bostr) > (5 + taill) - l - ? (5 + taill) - l : locinput - PL_bostr; - int pref0_len; - - while (do_utf8 && UTF8_IS_CONTINUATION(*(U8*)(locinput - pref_len))) - pref_len++; - pref0_len = pref_len - (locinput - PL_reg_starttry); - if (l + pref_len < (5 + taill) && l < PL_regeol - locinput) - l = ( PL_regeol - locinput > (5 + taill) - pref_len - ? (5 + taill) - pref_len : PL_regeol - locinput); - while (do_utf8 && UTF8_IS_CONTINUATION(*(U8*)(locinput + l))) - l--; - if (pref0_len < 0) - pref0_len = 0; - if (pref0_len > pref_len) - pref0_len = pref_len; - regprop(prop, scan); - { - char *s0 = - do_utf8 && OP(scan) != CANY ? - pv_uni_display(dsv0, (U8*)(locinput - pref_len), - pref0_len, 60, UNI_DISPLAY_REGEX) : - locinput - pref_len; - int len0 = do_utf8 ? strlen(s0) : pref0_len; - char *s1 = do_utf8 && OP(scan) != CANY ? - pv_uni_display(dsv1, (U8*)(locinput - pref_len + pref0_len), - pref_len - pref0_len, 60, UNI_DISPLAY_REGEX) : - locinput - pref_len + pref0_len; - int len1 = do_utf8 ? strlen(s1) : pref_len - pref0_len; - char *s2 = do_utf8 && OP(scan) != CANY ? - pv_uni_display(dsv2, (U8*)locinput, - PL_regeol - locinput, 60, UNI_DISPLAY_REGEX) : - locinput; - int len2 = do_utf8 ? strlen(s2) : l; - PerlIO_printf(Perl_debug_log, - "%4"IVdf" <%s%.*s%s%s%.*s%s%s%s%.*s%s>%*s|%3"IVdf":%*s%s\n", - (IV)(locinput - PL_bostr), - PL_colors[4], - len0, s0, - PL_colors[5], - PL_colors[2], - len1, s1, - PL_colors[3], - (docolor ? "" : "> <"), - PL_colors[0], - len2, s2, - PL_colors[1], - 15 - l - pref_len + 1, - "", - (IV)(scan - PL_regprogram), PL_regindent*2, "", - SvPVX(prop)); - } + SV * const prop = sv_newmortal(); + regnode *rnext=regnext(scan); + DUMP_EXEC_POS( locinput, scan, do_utf8 ); + regprop(rex, prop, scan); + + PerlIO_printf(Perl_debug_log, + "%3"IVdf":%*s%s(%"IVdf")\n", + (IV)(scan - rex->program), depth*2, "", + SvPVX_const(prop), + (PL_regkind[OP(scan)] == END || !rnext) ? + 0 : (IV)(rnext - rex->program)); }); next = scan + NEXT_OFF(scan); if (next == scan) next = NULL; + state_num = OP(scan); - switch (OP(scan)) { + reenter_switch: + switch (state_num) { case BOL: if (locinput == PL_bostr) { - /* regtill = regbol; */ + /* reginfo->till = reginfo->bol; */ break; } sayNO; @@ -2477,7 +2649,7 @@ S_regmatch(pTHX_ regnode *prog) break; sayNO; case GPOS: - if (locinput == PL_reg_ganch) + if (locinput == reginfo->ganch) break; sayNO; case EOL: @@ -2527,248 +2699,299 @@ S_regmatch(pTHX_ regnode *prog) nextchr = UCHARAT(++locinput); break; - - - /* - traverse the TRIE keeping track of all accepting states - we transition through until we get to a failing node. - - we use two slightly different pieces of code to handle - the traversal depending on whether its case sensitive or - not. we reuse the accept code however. (this should probably - be turned into a macro.) - - */ - case TRIEF: - case TRIEFL: +#undef ST +#define ST st->u.trie + case TRIEC: + /* In this case the charclass data is available inline so + we can fail fast without a lot of extra overhead. + */ + if (scan->flags == EXACT || !do_utf8) { + if(!ANYOF_BITMAP_TEST(scan, *locinput)) { + DEBUG_EXECUTE_r( + PerlIO_printf(Perl_debug_log, + "%*s %sfailed to match trie start class...%s\n", + REPORT_CODE_OFF+depth*2, "", PL_colors[4], PL_colors[5]) + ); + sayNO_SILENT; + /* NOTREACHED */ + } + } + /* FALL THROUGH */ + case TRIE: { - - U32 uniflags = ckWARN( WARN_UTF8 ) ? 0 : UTF8_ALLOW_ANY; + /* what type of TRIE am I? (utf8 makes this contextual) */ + const enum { trie_plain, trie_utf8, trie_utf8_fold } + trie_type = do_utf8 ? + (scan->flags == EXACT ? trie_utf8 : trie_utf8_fold) + : trie_plain; + + /* what trie are we using right now */ + reg_trie_data * const trie + = (reg_trie_data*)rex->data->data[ ARG( scan ) ]; + U32 state = trie->startstate; + + if (trie->bitmap && trie_type != trie_utf8_fold && + !TRIE_BITMAP_TEST(trie,*locinput) + ) { + if (trie->states[ state ].wordnum) { + DEBUG_EXECUTE_r( + PerlIO_printf(Perl_debug_log, + "%*s %smatched empty string...%s\n", + REPORT_CODE_OFF+depth*2, "", PL_colors[4], PL_colors[5]) + ); + break; + } else { + DEBUG_EXECUTE_r( + PerlIO_printf(Perl_debug_log, + "%*s %sfailed to match trie start class...%s\n", + REPORT_CODE_OFF+depth*2, "", PL_colors[4], PL_colors[5]) + ); + sayNO_SILENT; + } + } + + { U8 *uc = ( U8* )locinput; - U32 state = 1; - U16 charid = 0; - U32 base = 0; - UV uvc = 0; + STRLEN len = 0; STRLEN foldlen = 0; - U8 foldbuf[ UTF8_MAXBYTES_CASE + 1 ]; U8 *uscan = (U8*)NULL; STRLEN bufflen=0; - accepted = 0; + SV *sv_accept_buff = NULL; + U8 foldbuf[ UTF8_MAXBYTES_CASE + 1 ]; - trie = (reg_trie_data*)PL_regdata->data[ ARG( scan ) ]; + 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. + */ while ( state && uc <= (U8*)PL_regeol ) { + U32 base = trie->states[ state ].trans.base; + UV uvc = 0; + U16 charid; + /* We use charid to hold the wordnum as we don't use it + for charid until after we have done the wordnum logic. + We define an alias just so that the wordnum logic reads + more naturally. */ + +#define got_wordnum charid + got_wordnum = trie->states[ state ].wordnum; + + if ( got_wordnum ) { + if ( ! ST.accepted ) { + ENTER; + SAVETMPS; + bufflen = TRIE_INITAL_ACCEPT_BUFFLEN; + sv_accept_buff=newSV(bufflen * + sizeof(reg_trie_accepted) - 1); + SvCUR_set(sv_accept_buff, 0); + SvPOK_on(sv_accept_buff); + sv_2mortal(sv_accept_buff); + SAVETMPS; + ST.accept_buff = + (reg_trie_accepted*)SvPV_nolen(sv_accept_buff ); + } + do { + if (ST.accepted >= bufflen) { + bufflen *= 2; + ST.accept_buff =(reg_trie_accepted*) + SvGROW(sv_accept_buff, + bufflen * sizeof(reg_trie_accepted)); + } + SvCUR_set(sv_accept_buff,SvCUR(sv_accept_buff) + + sizeof(reg_trie_accepted)); - TRIE_CHECK_STATE_IS_ACCEPTING; - base = trie->states[ state ].trans.base; + ST.accept_buff[ST.accepted].wordnum = got_wordnum; + ST.accept_buff[ST.accepted].endpos = uc; + ++ST.accepted; + } while (trie->nextword && (got_wordnum= trie->nextword[got_wordnum])); + } +#undef got_wordnum - DEBUG_TRIE_EXECUTE_r( + DEBUG_TRIE_EXECUTE_r({ + DUMP_EXEC_POS( (char *)uc, scan, do_utf8 ); PerlIO_printf( Perl_debug_log, - "%*s %sState: %4x, Base: %4x Accepted: %4x ", - REPORT_CODE_OFF + PL_regindent * 2, "", PL_colors[4], - state, base, accepted ); - ); + "%*s %sState: %4"UVxf" Accepted: %4"UVxf" ", + 2+depth * 2, "", PL_colors[4], + (UV)state, (UV)ST.accepted ); + }); if ( base ) { - - if ( do_utf8 || UTF ) { - if ( foldlen>0 ) { - uvc = utf8n_to_uvuni( uscan, UTF8_MAXLEN, &len, uniflags ); - foldlen -= len; - uscan += len; - len=0; - } else { - uvc = utf8n_to_uvuni( (U8*)uc, UTF8_MAXLEN, &len, uniflags ); - uvc = to_uni_fold( uvc, foldbuf, &foldlen ); - foldlen -= UNISKIP( uvc ); - uscan = foldbuf + UNISKIP( uvc ); - } - } else { - uvc = (U32)*uc; - len = 1; + REXEC_TRIE_READ_CHAR(trie_type, trie, uc, uscan, len, + uvc, charid, foldlen, foldbuf, uniflags); + + if (charid && + (base + charid > trie->uniquecharcount ) + && (base + charid - 1 - trie->uniquecharcount + < trie->lasttrans) + && trie->trans[base + charid - 1 - + trie->uniquecharcount].check == state) + { + state = trie->trans[base + charid - 1 - + trie->uniquecharcount ].next; } + else { + state = 0; + } + uc += len; - TRIE_HANDLE_CHAR; - - } else { + } + else { state = 0; } DEBUG_TRIE_EXECUTE_r( - PerlIO_printf( Perl_debug_log, - "Charid:%3x CV:%4x After State: %4x%s\n", - charid, uvc, state, PL_colors[5] ); + PerlIO_printf( Perl_debug_log, + "Charid:%3x CP:%4"UVxf" After State: %4"UVxf"%s\n", + charid, uvc, (UV)state, PL_colors[5] ); ); } - if ( !accepted ) { + if (!ST.accepted ) sayNO; - } else { - goto TrieAccept; - } - } - /* unreached codepoint: we jump into the middle of the next case - from previous if blocks */ - case TRIE: - { - U32 uniflags = ckWARN( WARN_UTF8 ) ? 0 : UTF8_ALLOW_ANY; - U8 *uc = (U8*)locinput; - U32 state = 1; - U16 charid = 0; - U32 base = 0; - UV uvc = 0; - STRLEN len = 0; - STRLEN bufflen = 0; - accepted = 0; - - trie = (reg_trie_data*)PL_regdata->data[ ARG( scan ) ]; - - while ( state && uc <= (U8*)PL_regeol ) { - TRIE_CHECK_STATE_IS_ACCEPTING; - - base = trie->states[ state ].trans.base; + DEBUG_EXECUTE_r( + PerlIO_printf( Perl_debug_log, + "%*s %sgot %"IVdf" possible matches%s\n", + REPORT_CODE_OFF + depth * 2, "", + PL_colors[4], (IV)ST.accepted, PL_colors[5] ); + ); + }} - DEBUG_TRIE_EXECUTE_r( - PerlIO_printf( Perl_debug_log, - "%*s %sState: %4x, Base: %4x Accepted: %4x ", - REPORT_CODE_OFF + PL_regindent * 2, "", PL_colors[4], - state, base, accepted ); - ); + /* FALL THROUGH */ - if ( base ) { + case TRIE_next_fail: /* we failed - try next alterative */ + + 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; + 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", + 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]; + + continue; /* execute rest of RE */ + } - if ( do_utf8 || UTF ) { - uvc = utf8n_to_uvuni( (U8*)uc, UTF8_MAXLEN, &len, uniflags ); - } else { - uvc = (U32)*uc; - len = 1; - } + if (!ST.accepted-- ) { + FREETMPS; + LEAVE; + sayNO; + } - TRIE_HANDLE_CHAR; + /* + There are at least two accepting states left. Presumably + the number of accepting states is going to be low, + typically two. So we simply scan through to find the one + with lowest wordnum. Once we find it, we swap the last + state into its place and decrement the size. We then try to + match the rest of the pattern at the point where the word + ends. If we succeed, control just continues along the + regex; if we fail we return here to try the next accepting + state + */ - } else { - state = 0; - } + { + U32 best = 0; + U32 cur; + for( cur = 1 ; cur <= ST.accepted ; cur++ ) { DEBUG_TRIE_EXECUTE_r( - PerlIO_printf( Perl_debug_log, - "Charid:%3x CV:%4x After State: %4x%s\n", - charid, uvc, state, PL_colors[5] ); + PerlIO_printf( Perl_debug_log, + "%*s %sgot %"IVdf" (%d) as best, looking at %"IVdf" (%d)%s\n", + REPORT_CODE_OFF + depth * 2, "", PL_colors[4], + (IV)best, ST.accept_buff[ best ].wordnum, (IV)cur, + ST.accept_buff[ cur ].wordnum, PL_colors[5] ); ); - } - if ( !accepted ) { - sayNO; - } - } + if (ST.accept_buff[cur].wordnum < + ST.accept_buff[best].wordnum) + best = cur; + } - /* - There was at least one accepting state that we - transitioned through. Presumably the number of accepting - states is going to be low, typically one or two. So we - simply scan through to find the one with lowest wordnum. - Once we find it, we swap the last state into its place - and decrement the size. We then try to match the rest of - the pattern at the point where the word ends, if we - succeed then we end the loop, otherwise the loop - eventually terminates once all of the accepting states - have been tried. - */ - TrieAccept: - { - int gotit = 0; - - if ( accepted == 1 ) { - DEBUG_EXECUTE_r({ - SV **tmp = av_fetch( trie->words, accept_buff[ 0 ].wordnum-1, 0 ); - PerlIO_printf( Perl_debug_log, - "%*s %sonly one match : #%d <%s>%s\n", - REPORT_CODE_OFF+PL_regindent*2, "",PL_colors[4], - accept_buff[ 0 ].wordnum, - tmp ? SvPV_nolen( *tmp ) : "not compiled under -Dr", - PL_colors[5] ); - }); - PL_reginput = 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; - gotit = regmatch( scan + NEXT_OFF( scan ) ); - } else { - DEBUG_EXECUTE_r( - PerlIO_printf( Perl_debug_log,"%*s %sgot %d possible matches%s\n", - REPORT_CODE_OFF + PL_regindent * 2, "",PL_colors[4], accepted, - PL_colors[5] ); - ); - while ( !gotit && accepted-- ) { - U32 best = 0; - U32 cur; - for( cur = 1 ; cur <= accepted ; cur++ ) { - DEBUG_TRIE_EXECUTE_r( - PerlIO_printf( Perl_debug_log, - "%*s %sgot %d (%d) as best, looking at %d (%d)%s\n", - REPORT_CODE_OFF + PL_regindent * 2, "", PL_colors[4], - best, accept_buff[ best ].wordnum, cur, - accept_buff[ cur ].wordnum, PL_colors[5] ); - ); - - if ( accept_buff[ cur ].wordnum < accept_buff[ best ].wordnum ) - best = cur; - } - DEBUG_EXECUTE_r({ - SV **tmp = av_fetch( trie->words, accept_buff[ best ].wordnum - 1, 0 ); - PerlIO_printf( Perl_debug_log, "%*s %strying alternation #%d <%s> at 0x%p%s\n", - REPORT_CODE_OFF+PL_regindent*2, "",PL_colors[4], - accept_buff[best].wordnum, - tmp ? SvPV_nolen( *tmp ) : "not compiled under -Dr",scan, - PL_colors[5] ); - }); - if ( bestdata->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 ? + ST.B : + ST.B - ST.jump[ST.accept_buff[best].wordnum]; + 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", + REG_NODE_NUM(nextop), + PL_colors[5] ); + }); + + if ( best= PL_regeol) sayNO; if (NATIVE_TO_UNI(*(U8*)s) != utf8n_to_uvuni((U8*)l, UTF8_MAXBYTES, &ulen, - ckWARN(WARN_UTF8) ? - 0 : UTF8_ALLOW_ANY)) + uniflags)) sayNO; l += ulen; s ++; @@ -2777,12 +3000,12 @@ S_regmatch(pTHX_ regnode *prog) else { /* The target is not utf8, the pattern is utf8. */ while (s < e) { + STRLEN ulen; if (l >= PL_regeol) sayNO; if (NATIVE_TO_UNI(*((U8*)l)) != utf8n_to_uvuni((U8*)s, UTF8_MAXBYTES, &ulen, - ckWARN(WARN_UTF8) ? - 0 : UTF8_ALLOW_ANY)) + uniflags)) sayNO; s += ulen; l ++; @@ -2803,16 +3026,17 @@ S_regmatch(pTHX_ regnode *prog) locinput += ln; nextchr = UCHARAT(locinput); break; + } case EXACTFL: PL_reg_flags |= RF_tainted; /* FALL THROUGH */ - case EXACTF: - s = STRING(scan); + case EXACTF: { + char * const s = STRING(scan); ln = STR_LEN(scan); if (do_utf8 || UTF) { /* Either target or the pattern are utf8. */ - char *l = locinput; + const char * const l = locinput; char *e = PL_regeol; if (ibcmp_utf8(s, 0, ln, (bool)UTF, @@ -2851,12 +3075,13 @@ S_regmatch(pTHX_ regnode *prog) locinput += ln; nextchr = UCHARAT(locinput); break; + } case ANYOF: if (do_utf8) { STRLEN inclasslen = PL_regeol - locinput; - if (!reginclass(scan, (U8*)locinput, &inclasslen, do_utf8)) - sayNO_ANYOF; + if (!reginclass(rex, scan, (U8*)locinput, &inclasslen, do_utf8)) + goto anyof_fail; if (locinput >= PL_regeol) sayNO; locinput += inclasslen ? inclasslen : UTF8SKIP(locinput); @@ -2866,14 +3091,14 @@ S_regmatch(pTHX_ regnode *prog) else { if (nextchr < 0) nextchr = UCHARAT(locinput); - if (!REGINCLASS(scan, (U8*)locinput)) - sayNO_ANYOF; + if (!REGINCLASS(rex, scan, (U8*)locinput)) + goto anyof_fail; if (!nextchr && locinput >= PL_regeol) sayNO; nextchr = UCHARAT(++locinput); break; } - no_anyof: + anyof_fail: /* If we might have the case of the German sharp s * in a casefolding Unicode character class. */ @@ -2891,9 +3116,9 @@ S_regmatch(pTHX_ regnode *prog) if (!nextchr) sayNO; if (do_utf8) { - LOAD_UTF8_CHARCLASS(alnum,"a"); + LOAD_UTF8_CHARCLASS_ALNUM(); if (!(OP(scan) == ALNUM - ? swash_fetch(PL_utf8_alnum, (U8*)locinput, do_utf8) + ? (bool)swash_fetch(PL_utf8_alnum, (U8*)locinput, do_utf8) : isALNUM_LC_utf8((U8*)locinput))) { sayNO; @@ -2914,9 +3139,9 @@ S_regmatch(pTHX_ regnode *prog) if (!nextchr && locinput >= PL_regeol) sayNO; if (do_utf8) { - LOAD_UTF8_CHARCLASS(alnum,"a"); + LOAD_UTF8_CHARCLASS_ALNUM(); if (OP(scan) == NALNUM - ? swash_fetch(PL_utf8_alnum, (U8*)locinput, do_utf8) + ? (bool)swash_fetch(PL_utf8_alnum, (U8*)locinput, do_utf8) : isALNUM_LC_utf8((U8*)locinput)) { sayNO; @@ -2941,13 +3166,13 @@ S_regmatch(pTHX_ regnode *prog) if (locinput == PL_bostr) ln = '\n'; else { - U8 *r = reghop3((U8*)locinput, -1, (U8*)PL_bostr); + const U8 * const r = reghop3((U8*)locinput, -1, (U8*)PL_bostr); - ln = utf8n_to_uvchr(r, UTF8SKIP(r), 0, 0); + ln = utf8n_to_uvchr(r, UTF8SKIP(r), 0, uniflags); } if (OP(scan) == BOUND || OP(scan) == NBOUND) { ln = isALNUM_uni(ln); - LOAD_UTF8_CHARCLASS(alnum,"a"); + LOAD_UTF8_CHARCLASS_ALNUM(); n = swash_fetch(PL_utf8_alnum, (U8*)locinput, do_utf8); } else { @@ -2979,9 +3204,9 @@ S_regmatch(pTHX_ regnode *prog) sayNO; if (do_utf8) { if (UTF8_IS_CONTINUED(nextchr)) { - LOAD_UTF8_CHARCLASS(space," "); + LOAD_UTF8_CHARCLASS_SPACE(); if (!(OP(scan) == SPACE - ? swash_fetch(PL_utf8_space, (U8*)locinput, do_utf8) + ? (bool)swash_fetch(PL_utf8_space, (U8*)locinput, do_utf8) : isSPACE_LC_utf8((U8*)locinput))) { sayNO; @@ -3009,9 +3234,9 @@ S_regmatch(pTHX_ regnode *prog) if (!nextchr && locinput >= PL_regeol) sayNO; if (do_utf8) { - LOAD_UTF8_CHARCLASS(space," "); + LOAD_UTF8_CHARCLASS_SPACE(); if (OP(scan) == NSPACE - ? swash_fetch(PL_utf8_space, (U8*)locinput, do_utf8) + ? (bool)swash_fetch(PL_utf8_space, (U8*)locinput, do_utf8) : isSPACE_LC_utf8((U8*)locinput)) { sayNO; @@ -3032,9 +3257,9 @@ S_regmatch(pTHX_ regnode *prog) if (!nextchr) sayNO; if (do_utf8) { - LOAD_UTF8_CHARCLASS(digit,"0"); + LOAD_UTF8_CHARCLASS_DIGIT(); if (!(OP(scan) == DIGIT - ? swash_fetch(PL_utf8_digit, (U8*)locinput, do_utf8) + ? (bool)swash_fetch(PL_utf8_digit, (U8*)locinput, do_utf8) : isDIGIT_LC_utf8((U8*)locinput))) { sayNO; @@ -3055,9 +3280,9 @@ S_regmatch(pTHX_ regnode *prog) if (!nextchr && locinput >= PL_regeol) sayNO; if (do_utf8) { - LOAD_UTF8_CHARCLASS(digit,"0"); + LOAD_UTF8_CHARCLASS_DIGIT(); if (OP(scan) == NDIGIT - ? swash_fetch(PL_utf8_digit, (U8*)locinput, do_utf8) + ? (bool)swash_fetch(PL_utf8_digit, (U8*)locinput, do_utf8) : isDIGIT_LC_utf8((U8*)locinput)) { sayNO; @@ -3075,7 +3300,7 @@ S_regmatch(pTHX_ regnode *prog) if (locinput >= PL_regeol) sayNO; if (do_utf8) { - LOAD_UTF8_CHARCLASS(mark,"~"); + LOAD_UTF8_CHARCLASS_MARK(); if (swash_fetch(PL_utf8_mark,(U8*)locinput, do_utf8)) sayNO; locinput += PL_utf8skip[nextchr]; @@ -3089,12 +3314,33 @@ S_regmatch(pTHX_ 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: + case REFF: n = ARG(scan); /* which paren pair */ + type = OP(scan); + do_ref: ln = PL_regstartp[n]; PL_reg_leftiter = PL_reg_maxiter; /* Void cache */ if ((I32)*PL_reglastparen < n || ln == -1) @@ -3103,19 +3349,20 @@ S_regmatch(pTHX_ regnode *prog) break; s = PL_bostr + ln; - if (do_utf8 && OP(scan) != REF) { /* REF can do byte comparison */ + if (do_utf8 && type != REF) { /* REF can do byte comparison */ char *l = locinput; - char *e = PL_bostr + PL_regendp[n]; + const char *e = PL_bostr + PL_regendp[n]; /* * Note that we can't do the "other character" lookup trick as * 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) { - STRLEN ulen1, ulen2; - U8 tmpbuf1[UTF8_MAXBYTES_CASE+1]; - U8 tmpbuf2[UTF8_MAXBYTES_CASE+1]; + if (type == REFF) { while (s < e) { + STRLEN ulen1, ulen2; + U8 tmpbuf1[UTF8_MAXBYTES_CASE+1]; + U8 tmpbuf2[UTF8_MAXBYTES_CASE+1]; + if (l >= PL_regeol) sayNO; toLOWER_utf8((U8*)s, tmpbuf1, &ulen1); @@ -3133,45 +3380,78 @@ S_regmatch(pTHX_ 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; ln = PL_regendp[n] - ln; if (locinput + ln > PL_regeol) sayNO; - if (ln > 1 && (OP(scan) == REF + if (ln > 1 && (type == REF ? memNE(s, locinput, ln) - : (OP(scan) == REFF + : (type == REFF ? ibcmp(s, locinput, ln) : ibcmp_locale(s, locinput, ln)))) sayNO; locinput += ln; nextchr = UCHARAT(locinput); break; - + } case NOTHING: case TAIL: break; case BACK: break; - case EVAL: + +#undef ST +#define ST st->u.eval { - dSP; - OP_4tree *oop = PL_op; - COP *ocurcop = PL_curcop; - PAD *old_comppad; SV *ret; - struct regexp *oreg = PL_reg_re; - - n = ARG(scan); - PL_op = (OP_4tree*)PL_regdata->data[n]; - DEBUG_EXECUTE_r( PerlIO_printf(Perl_debug_log, " re_eval 0x%"UVxf"\n", PTR2UV(PL_op)) ); - PAD_SAVE_LOCAL(old_comppad, (PAD*)PL_regdata->data[n + 2]); - PL_regendp[0] = PL_reg_magic->mg_len = locinput - PL_bostr; - + regexp *re; + regnode *startpoint; + + case SRECURSE: + case RECURSE: /* /(...(?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 ( ++nochange_depth > MAX_RECURSE_EVAL_NOCHANGE_DEPTH ) + Perl_croak(aTHX_ "RECURSE without pos change exceeded limit in regexp"); + } else { + nochange_depth = 0; + } + re = rex; + (void)ReREFCNT_inc(rex); + if (OP(scan)==RECURSE) { + startpoint = scan + ARG2L(scan); + ST.close_paren = ARG(scan); + } else { + startpoint = re->program+1; + ST.close_paren = 0; + } + goto eval_recurse_doit; + /* NOTREACHED */ + 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"); + } else { + nochange_depth = 0; + } { - SV **before = SP; + /* execute the code in the {...} */ + dSP; + SV ** const before = SP; + OP_4tree * const oop = PL_op; + COP * const ocurcop = PL_curcop; + 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_regendp[0] = PL_reg_magic->mg_len = locinput - PL_bostr; + CALLRUNOPS(aTHX); /* Scalar context. */ SPAGAIN; if (SP == before) @@ -3180,20 +3460,24 @@ S_regmatch(pTHX_ regnode *prog) ret = POPs; PUTBACK; } - } - PL_op = oop; - PAD_RESTORE_LOCAL(old_comppad); - PL_curcop = ocurcop; - if (logical) { - if (logical == 2) { /* Postponed subexpression. */ - regexp *re; - MAGIC *mg = Null(MAGIC*); - re_cc_state state; - CHECKPOINT cp, lastcp; - int toggleutf; - register SV *sv; + PL_op = oop; + PAD_RESTORE_LOCAL(old_comppad); + PL_curcop = ocurcop; + if (!logical) { + /* /(?{...})/ */ + sv_setsv(save_scalar(PL_replgv), ret); + break; + } + } + if (logical == 2) { /* Postponed subexpression: /(??{...})/ */ + logical = 0; + { + /* extract RE object from returned value; compiling if + * necessary */ + MAGIC *mg = NULL; + const SV *sv; if(SvROK(ret) && SvSMAGICAL(sv = SvRV(ret))) mg = mg_find(sv, PERL_MAGIC_qr); else if (SvSMAGICAL(ret)) { @@ -3209,98 +3493,103 @@ S_regmatch(pTHX_ regnode *prog) } else { STRLEN len; - char *t = SvPV(ret, len); + const char * const t = SvPV_const(ret, len); PMOP pm; - char *oprecomp = PL_regprecomp; - I32 osize = PL_regsize; - I32 onpar = PL_regnpar; + const I32 osize = PL_regsize; Zero(&pm, 1, PMOP); - if (DO_UTF8(ret)) pm.op_pmdynflags |= PMdf_DYN_UTF8; - re = CALLREGCOMP(aTHX_ t, t + len, &pm); + if (DO_UTF8(ret)) pm.op_pmdynflags |= PMdf_DYN_UTF8; + re = CALLREGCOMP((char*)t, (char*)t + len, &pm); if (!(SvFLAGS(ret) & (SVs_TEMP | SVs_PADTMP | SVf_READONLY | SVs_GMG))) sv_magic(ret,(SV*)ReREFCNT_inc(re), PERL_MAGIC_qr,0,0); - PL_regprecomp = oprecomp; PL_regsize = osize; - PL_regnpar = onpar; } - DEBUG_EXECUTE_r( - PerlIO_printf(Perl_debug_log, - "Entering embedded `%s%.60s%s%s'\n", - PL_colors[0], - re->precomp, - PL_colors[1], - (strlen(re->precomp) > 60 ? "..." : "")) - ); - state.node = next; - state.prev = PL_reg_call_cc; - state.cc = PL_regcc; - state.re = PL_reg_re; - - PL_regcc = 0; - - cp = regcppush(0); /* Save *all* the positions. */ - REGCP_SET(lastcp); - cache_re(re); - state.ss = PL_savestack_ix; - *PL_reglastparen = 0; - *PL_reglastcloseparen = 0; - PL_reg_call_cc = &state; - PL_reginput = locinput; - toggleutf = ((PL_reg_flags & RF_utf8) != 0) ^ - ((re->reganch & ROPT_UTF8) != 0); - if (toggleutf) PL_reg_flags ^= RF_utf8; - - /* XXXX This is too dramatic a measure... */ - PL_reg_maxiter = 0; - - if (regmatch(re->program + 1)) { - /* Even though we succeeded, we need to restore - global variables, since we may be wrapped inside - SUSPEND, thus the match may be not finished yet. */ - - /* XXXX Do this only if SUSPENDed? */ - PL_reg_call_cc = state.prev; - PL_regcc = state.cc; - PL_reg_re = state.re; - cache_re(PL_reg_re); - if (toggleutf) PL_reg_flags ^= RF_utf8; - - /* XXXX This is too dramatic a measure... */ - PL_reg_maxiter = 0; - - /* These are needed even if not SUSPEND. */ - ReREFCNT_dec(re); - regcpblow(cp); - sayYES; - } - ReREFCNT_dec(re); - REGCP_UNWIND(lastcp); - regcppop(); - PL_reg_call_cc = state.prev; - PL_regcc = state.cc; - PL_reg_re = state.re; - cache_re(PL_reg_re); - if (toggleutf) PL_reg_flags ^= RF_utf8; - - /* XXXX This is too dramatic a measure... */ - PL_reg_maxiter = 0; - - logical = 0; - sayNO; } - sw = SvTRUE(ret); - logical = 0; - } - else { - sv_setsv(save_scalar(PL_replgv), ret); - cache_re(oreg); + 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 */ + /* borrowed from regtry */ + if (PL_reg_start_tmpl <= re->nparens) { + PL_reg_start_tmpl = re->nparens*3/2 + 3; + if(PL_reg_start_tmp) + Renew(PL_reg_start_tmp, PL_reg_start_tmpl, char*); + else + Newx(PL_reg_start_tmp, PL_reg_start_tmpl, char*); + } + + eval_recurse_doit: /* Share code with RECURSE 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_reglastparen = 0; + *PL_reglastcloseparen = 0; + PL_reginput = locinput; + + /* XXXX This is too dramatic a measure... */ + PL_reg_maxiter = 0; + + ST.toggle_reg_flags = PL_reg_flags; + if (re->reganch & ROPT_UTF8) + PL_reg_flags |= RF_utf8; + else + PL_reg_flags &= ~RF_utf8; + ST.toggle_reg_flags ^= PL_reg_flags; /* diff of old and new */ + + ST.prev_rex = rex; + ST.prev_curlyx = cur_curlyx; + rex = re; + cur_curlyx = NULL; + ST.B = next; + ST.prev_eval = cur_eval; + cur_eval = st; + /* now continue from first node in postoned RE */ + PUSH_YES_STATE_GOTO(EVAL_AB, startpoint); + /* NOTREACHED */ } + /* logical is 1, /(?(?{...})X|Y)/ */ + sw = (bool)SvTRUE(ret); + logical = 0; break; } + + case EVAL_AB: /* cleanup after a successful (??{A})B */ + /* note: this is called twice; first after popping B, then A */ + PL_reg_flags ^= ST.toggle_reg_flags; + ReREFCNT_dec(rex); + rex = ST.prev_rex; + regcpblow(ST.cp); + cur_eval = ST.prev_eval; + cur_curlyx = ST.prev_curlyx; + /* XXXX This is too dramatic a measure... */ + PL_reg_maxiter = 0; + sayYES; + + + case EVAL_AB_fail: /* unsuccessfully ran A or B in (??{A})B */ + /* note: this is called twice; first after popping B, then A */ + PL_reg_flags ^= ST.toggle_reg_flags; + ReREFCNT_dec(rex); + rex = ST.prev_rex; + PL_reginput = locinput; + REGCP_UNWIND(ST.lastcp); + regcppop(rex); + cur_eval = ST.prev_eval; + cur_curlyx = ST.prev_curlyx; + /* XXXX This is too dramatic a measure... */ + PL_reg_maxiter = 0; + sayNO_SILENT; +#undef ST + case OPEN: n = ARG(scan); /* which paren pair */ PL_reg_start_tmp[n] = locinput; @@ -3314,11 +3603,25 @@ S_regmatch(pTHX_ regnode *prog) if (n > (I32)*PL_reglastparen) *PL_reglastparen = n; *PL_reglastcloseparen = n; + if (cur_eval && cur_eval->u.eval.close_paren == (U32)n) { + goto fake_end; + } break; case GROUPP: n = ARG(scan); /* which paren pair */ - sw = ((I32)*PL_reglastparen >= n && PL_regendp[n] != -1); + sw = (bool)((I32)*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 RECURSEP: + 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 (sw) @@ -3332,176 +3635,197 @@ S_regmatch(pTHX_ regnode *prog) case LOGICAL: logical = scan->flags; break; + /******************************************************************* - PL_regcc contains infoblock about the innermost (...)* loop, and - a pointer to the next outer infoblock. - - Here is how Y(A)*Z is processed (if it is compiled into CURLYX/WHILEM): - - 1) After matching X, regnode for CURLYX is processed; - - 2) This regnode creates infoblock on the stack, and calls - regmatch() recursively with the starting point at WHILEM node; - - 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 the infoblock, as is info on how many times "A" - was already matched, and greediness. - - 4) After A matches, the same WHILEM node is hit again. - - 5) Each time WHILEM is hit, PL_regcc is the infoblock created by CURLYX - of the same pair. Thus when WHILEM tries to match Z, it temporarily - resets PL_regcc, 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. - - Currently present infoblocks form a tree with a stem formed by PL_curcc - and whatever it mentions via ->next, and additional attached trees - corresponding to temporarily unset infoblocks as in "5" above. - - 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. - - 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.] - - 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 - - YAAZ Y(A)*?Z)*?T x <- O - I - - YAAZY (A)*?Z)*?T x <- O - I - - YAAZY A)*?Z)*?T x <- O <- I - I - - YAAZYA )*?Z)*?T x <- O <- I - I - - YAAZYA Z)*?T x <- O # Temporary unset I - I,I - - YAAZYAZ )*?T x <- O - I,I - - YAAZYAZ T x # Temporary unset O - O - I,I - - YAAZYAZT x - O - I,I - *******************************************************************/ - case CURLYX: { - CURCUR cc; - CHECKPOINT cp = PL_savestack_ix; - /* No need to save/restore up to this paren */ - I32 parenfloor = scan->flags; - - if (OP(PREVOPER(next)) == NOTHING) /* LONGJMP */ - next += ARG(next); - cc.oldcc = PL_regcc; - PL_regcc = &cc; - /* XXXX Probably it is better to teach regpush to support - parenfloor > PL_regsize... */ - if (parenfloor > (I32)*PL_reglastparen) - parenfloor = *PL_reglastparen; /* Pessimization... */ - cc.parenfloor = parenfloor; - cc.cur = -1; - cc.min = ARG1(scan); - cc.max = ARG2(scan); - cc.scan = NEXTOPER(scan) + EXTRA_STEP_2ARGS; - cc.next = next; - cc.minmod = minmod; - cc.lastloc = 0; - PL_reginput = locinput; - n = regmatch(PREVOPER(next)); /* start on the WHILEM */ - regcpblow(cp); - PL_regcc = cc.oldcc; - saySAME(n); - } - /* NOT REACHED */ - 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. - */ - CHECKPOINT cp, lastcp; - CURCUR* cc = PL_regcc; - char *lastloc = cc->lastloc; /* Detection of 0-len. */ - - n = cc->cur + 1; /* how many we know we matched */ - PL_reginput = locinput; +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.) - DEBUG_EXECUTE_r( - PerlIO_printf(Perl_debug_log, - "%*s %ld out of %ld..%ld cc=%"UVxf"\n", - REPORT_CODE_OFF+PL_regindent*2, "", - (long)n, (long)cc->min, - (long)cc->max, PTR2UV(cc)) - ); +A*B is compiled as - /* If degenerate scan matches "", assume scan done. */ +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. - if (locinput == cc->lastloc && n >= cc->min) { - PL_regcc = cc->oldcc; - if (PL_regcc) - ln = PL_regcc->cur; - DEBUG_EXECUTE_r( - PerlIO_printf(Perl_debug_log, - "%*s empty match detected, try continuation...\n", - REPORT_CODE_OFF+PL_regindent*2, "") - ); - if (regmatch(cc->next)) - sayYES; - if (PL_regcc) - PL_regcc->cur = ln; - PL_regcc = cc; - sayNO; - } +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). - /* First just match a string of min scans. */ +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. - if (n < cc->min) { - cc->cur = n; - cc->lastloc = locinput; - if (regmatch(cc->scan)) - sayYES; - cc->cur = n - 1; - cc->lastloc = lastloc; - sayNO; - } +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): + + + + a1 + a1 a2 + a1 a2 a3 + a1 a2 a3 b + +(Contrast this with something like CURLYM, which maintains only a single +backtrack state: + + a1 + a1 a2 + a1 a2 a3 + a1 a2 a3 b +) + +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. + +This is complicated slightly by nested CURLYX/WHILEM's. Since cur_curlyx +must always point to the *current* CURLYX block, the rules are: + +When executing CURLYX, save the old cur_curlyx in the CURLYX state block, +and set cur_curlyx to point the new block. + +When popping the CURLYX block after a successful or unsuccessful match, +restore the previous cur_curlyx. + +When WHILEM is about to execute B, save the current cur_curlyx, and set it +to the outer one saved in the CURLYX block. + +When popping the WHILEM block after a successful or unsuccessful B match, +restore the previous cur_curlyx. + +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: + +cur_ +curlyx backtrack stack +------ --------------- +NULL +CO +CI ai +CO ai bi +NULL ai bi bo + +At this point the pattern succeeds, and we work back down the stack to +clean up, restoring as we go: + +CO ai bi +CI ai +CO +NULL + +*******************************************************************/ + +#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 */ + regcpblow(ST.cp); + cur_curlyx = ST.prev_curlyx; + sayYES; + /* NOTREACHED */ + + case CURLYX_end_fail: /* just failed to match all of A*B */ + regcpblow(ST.cp); + cur_curlyx = ST.prev_curlyx; + sayNO; + /* NOTREACHED */ + + +#undef ST +#define ST st->u.whilem + + 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; + + 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) { - I32 size = (PL_reg_maxiter + 7)/8; + /* initialise cache */ + const I32 size = (PL_reg_maxiter + 7)/8; if (PL_reg_poscache) { if ((I32)PL_reg_poscache_size < size) { Renew(PL_reg_poscache, size, char); @@ -3511,378 +3835,421 @@ S_regmatch(pTHX_ regnode *prog) } else { PL_reg_poscache_size = size; - Newz(29, PL_reg_poscache, size, char); + 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) { - I32 o = locinput - PL_bostr, b; - - o = (scan->flags & 0xf) - 1 + o * (scan->flags>>4); - b = o % 8; - o /= 8; - if (PL_reg_poscache[o] & (1<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, "") ); - if (PL_reg_flags & RF_false) - sayYES; - else - sayNO_SILENT; + sayNO; /* cache records failure */ } - PL_reg_poscache[o] |= (1<minmod) { - PL_regcc = cc->oldcc; - if (PL_regcc) - ln = PL_regcc->cur; - cp = regcppush(cc->parenfloor); - REGCP_SET(lastcp); - if (regmatch(cc->next)) { - regcpblow(cp); - sayYES; /* All done. */ - } - REGCP_UNWIND(lastcp); - regcppop(); - if (PL_regcc) - PL_regcc->cur = ln; - PL_regcc = cc; - - if (n >= cc->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); - } - sayNO; - } + /* Prefer B over A for minimal matching. */ - DEBUG_EXECUTE_r( - PerlIO_printf(Perl_debug_log, - "%*s trying longer...\n", - REPORT_CODE_OFF+PL_regindent*2, "") - ); - /* Try scanning more and see if it helps. */ - PL_reginput = locinput; - cc->cur = n; - cc->lastloc = locinput; - cp = regcppush(cc->parenfloor); - REGCP_SET(lastcp); - if (regmatch(cc->scan)) { - regcpblow(cp); - sayYES; - } - REGCP_UNWIND(lastcp); - regcppop(); - cc->cur = n - 1; - cc->lastloc = lastloc; - sayNO; - } + 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. */ + /* Prefer A over B for maximal matching. */ - if (n < cc->max) { /* More greed allowed? */ - cp = regcppush(cc->parenfloor); - cc->cur = n; - cc->lastloc = locinput; - REGCP_SET(lastcp); - if (regmatch(cc->scan)) { - regcpblow(cp); - sayYES; - } - REGCP_UNWIND(lastcp); - regcppop(); /* Restore some previous $s? */ - PL_reginput = locinput; - DEBUG_EXECUTE_r( - PerlIO_printf(Perl_debug_log, - "%*s failed, try continuation...\n", - REPORT_CODE_OFF+PL_regindent*2, "") - ); - } - if (ckWARN(WARN_REGEXP) && n >= REG_INFTY - && !(PL_reg_flags & RF_warned)) { + 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. */ - PL_regcc = cc->oldcc; - if (PL_regcc) - ln = PL_regcc->cur; - if (regmatch(cc->next)) - sayYES; - if (PL_regcc) - PL_regcc->cur = ln; - PL_regcc = cc; - cc->cur = n - 1; - cc->lastloc = lastloc; - sayNO; + cur_curlyx->u.curlyx.count--; + CACHEsayNO; } - /* NOT REACHED */ - case BRANCHJ: + + 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 +#define ST st->u.branch + + case BRANCHJ: /* /(...|A|...)/ with long next pointer */ next = scan + ARG(scan); if (next == scan) next = NULL; - inner = NEXTOPER(NEXTOPER(scan)); - goto do_branch; - case BRANCH: - inner = NEXTOPER(scan); - do_branch: - { - c1 = OP(scan); - if (OP(next) != c1) /* No choice. */ - next = inner; /* Avoid recursion. */ - else { - I32 lastparen = *PL_reglastparen; - I32 unwind1; - re_unwind_branch_t *uw; - - /* Put unwinding data on stack */ - unwind1 = SSNEWt(1,re_unwind_branch_t); - uw = SSPTRt(unwind1,re_unwind_branch_t); - uw->prev = unwind; - unwind = unwind1; - uw->type = ((c1 == BRANCH) - ? RE_UNWIND_BRANCH - : RE_UNWIND_BRANCHJ); - uw->lastparen = lastparen; - uw->next = next; - uw->locinput = locinput; - uw->nextchr = nextchr; -#ifdef DEBUGGING - uw->regindent = ++PL_regindent; -#endif + scan = NEXTOPER(scan); + /* FALL THROUGH */ - REGCP_SET(uw->lastcp); + case BRANCH: /* /(...|A|...)/ */ + scan = NEXTOPER(scan); /* scan now points to inner node */ + if (!next || (OP(next) != BRANCH && OP(next) != BRANCHJ)) + /* 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 first branch */ - next = inner; - } - } - break; + /* Now go into the branch */ + PUSH_STATE_GOTO(BRANCH_next, scan); + /* NOTREACHED */ + + case BRANCH_next_fail: /* that branch failed; try the next, if any */ + 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; + continue; /* execute next BRANCH[J] op */ + /* NOTREACHED */ + case MINMOD: minmod = 1; break; - case CURLYM: - { - I32 l = 0; - CHECKPOINT lastcp; - - /* We suppose that the next guy does not need - backtracking: in particular, it is of constant non-zero length, - and has no parenths to influence future backrefs. */ - ln = ARG1(scan); /* min to match */ - n = ARG2(scan); /* max to match */ - paren = scan->flags; - if (paren) { + +#undef ST +#define ST st->u.curlym + + case CURLYM: /* /A{m,n}B/ where A is fixed-length */ + + /* This is an optimisation of CURLYX that enables us to push + * only a single backtracking state, no matter now many matches + * there are in {m,n}. It relies on the pattern being constant + * length, with no parens to influence future backrefs + */ + + ST.me = scan; + scan = NEXTOPER(scan) + NODE_STEP_REGNODE; + + /* if paren positive, emulate an OPEN/CLOSE around A */ + if (ST.me->flags) { + I32 paren = ST.me->flags; if (paren > PL_regsize) PL_regsize = paren; if (paren > (I32)*PL_reglastparen) *PL_reglastparen = paren; - } - scan = NEXTOPER(scan) + NODE_STEP_REGNODE; - if (paren) scan += NEXT_OFF(scan); /* Skip former OPEN. */ - PL_reginput = locinput; - if (minmod) { - minmod = 0; - if (ln && regrepeat_hard(scan, ln, &l) < ln) - sayNO; - locinput = PL_reginput; - if (HAS_TEXT(next) || JUMPABLE(next)) { - regnode *text_node = next; - - if (! HAS_TEXT(text_node)) FIND_NEXT_IMPT(text_node); - - if (! HAS_TEXT(text_node)) c1 = c2 = -1000; - else { - if (PL_regkind[(U8)OP(text_node)] == REF) { - c1 = c2 = -1000; - goto assume_ok_MM; - } - else { c1 = (U8)*STRING(text_node); } - if (OP(text_node) == EXACTF || OP(text_node) == REFF) - c2 = PL_fold[c1]; - else if (OP(text_node) == EXACTFL || OP(text_node) == REFFL) - c2 = PL_fold_locale[c1]; - else - c2 = c1; + } + ST.A = scan; + ST.B = next; + ST.alen = 0; + ST.count = 0; + ST.minmod = minmod; + minmod = 0; + ST.c1 = CHRTEST_UNINIT; + REGCP_SET(ST.cp); + + if (!(ST.minmod ? ARG1(ST.me) : ARG2(ST.me))) /* min/max */ + goto curlym_do_B; + + curlym_do_A: /* execute the A in /A{m,n}B/ */ + PL_reginput = locinput; + PUSH_YES_STATE_GOTO(CURLYM_A, ST.A); /* match A */ + /* NOTREACHED */ + + case CURLYM_A: /* we've just matched an A */ + locinput = st->locinput; + nextchr = UCHARAT(locinput); + + ST.count++; + /* after first match, determine A's length: u.curlym.alen */ + if (ST.count == 1) { + if (PL_reg_match_utf8) { + char *s = locinput; + while (s < PL_reginput) { + ST.alen++; + s += UTF8SKIP(s); } } - else - c1 = c2 = -1000; - assume_ok_MM: - REGCP_SET(lastcp); - while (n >= ln || (n == REG_INFTY && ln > 0)) { /* ln overflow ? */ - /* If it could work, try it. */ - if (c1 == -1000 || - UCHARAT(PL_reginput) == c1 || - UCHARAT(PL_reginput) == c2) + else { + ST.alen = PL_reginput - locinput; + } + if (ST.alen == 0) + ST.count = ST.minmod ? ARG1(ST.me) : ARG2(ST.me); + } + DEBUG_EXECUTE_r( + PerlIO_printf(Perl_debug_log, + "%*s CURLYM now matched %"IVdf" times, len=%"IVdf"...\n", + (int)(REPORT_CODE_OFF+(depth*2)), "", + (IV) ST.count, (IV)ST.alen) + ); + + locinput = PL_reginput; + + if (cur_eval && cur_eval->u.eval.close_paren && + cur_eval->u.eval.close_paren == 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*/ + || (cur_eval && cur_eval->u.eval.close_paren && + cur_eval->u.eval.close_paren == ST.me->flags)) + sayNO; + + curlym_do_B: /* execute the B in /A{m,n}B/ */ + PL_reginput = locinput; + if (ST.c1 == CHRTEST_UNINIT) { + /* calculate c1 and c2 for possible match of 1st char + * following curly */ + ST.c1 = ST.c2 = CHRTEST_VOID; + if (HAS_TEXT(ST.B) || JUMPABLE(ST.B)) { + regnode *text_node = ST.B; + if (! HAS_TEXT(text_node)) + FIND_NEXT_IMPT(text_node); + if (HAS_TEXT(text_node) + && PL_regkind[OP(text_node)] != REF) { - if (paren) { - if (ln) { - PL_regstartp[paren] = - HOPc(PL_reginput, -l) - PL_bostr; - PL_regendp[paren] = PL_reginput - PL_bostr; - } - else - PL_regendp[paren] = -1; - } - if (regmatch(next)) - sayYES; - REGCP_UNWIND(lastcp); - } - /* Couldn't or didn't -- move forward. */ - PL_reginput = locinput; - if (regrepeat_hard(scan, 1, &l)) { - ln++; - locinput = PL_reginput; + ST.c1 = (U8)*STRING(text_node); + ST.c2 = + (OP(text_node) == EXACTF || OP(text_node) == REFF) + ? PL_fold[ST.c1] + : (OP(text_node) == EXACTFL || OP(text_node) == REFFL) + ? PL_fold_locale[ST.c1] + : ST.c1; } - else - sayNO; } } - else { - n = regrepeat_hard(scan, n, &l); - locinput = PL_reginput; - DEBUG_EXECUTE_r( - PerlIO_printf(Perl_debug_log, - "%*s matched %"IVdf" times, len=%"IVdf"...\n", - (int)(REPORT_CODE_OFF+PL_regindent*2), "", - (IV) n, (IV)l) - ); - if (n >= ln) { - if (HAS_TEXT(next) || JUMPABLE(next)) { - regnode *text_node = next; - if (! HAS_TEXT(text_node)) FIND_NEXT_IMPT(text_node); + DEBUG_EXECUTE_r( + PerlIO_printf(Perl_debug_log, + "%*s CURLYM trying tail with matches=%"IVdf"...\n", + (int)(REPORT_CODE_OFF+(depth*2)), + "", (IV)ST.count) + ); + if (ST.c1 != CHRTEST_VOID + && UCHARAT(PL_reginput) != ST.c1 + && UCHARAT(PL_reginput) != ST.c2) + { + /* simulate B failing */ + state_num = CURLYM_B_fail; + goto reenter_switch; + } - if (! HAS_TEXT(text_node)) c1 = c2 = -1000; - else { - if (PL_regkind[(U8)OP(text_node)] == REF) { - c1 = c2 = -1000; - goto assume_ok_REG; - } - else { c1 = (U8)*STRING(text_node); } - - if (OP(text_node) == EXACTF || OP(text_node) == REFF) - c2 = PL_fold[c1]; - else if (OP(text_node) == EXACTFL || OP(text_node) == REFFL) - c2 = PL_fold_locale[c1]; - else - c2 = c1; - } - } - else - c1 = c2 = -1000; - } - assume_ok_REG: - REGCP_SET(lastcp); - while (n >= ln) { - /* If it could work, try it. */ - if (c1 == -1000 || - UCHARAT(PL_reginput) == c1 || - UCHARAT(PL_reginput) == c2) - { - DEBUG_EXECUTE_r( - PerlIO_printf(Perl_debug_log, - "%*s trying tail with n=%"IVdf"...\n", - (int)(REPORT_CODE_OFF+PL_regindent*2), "", (IV)n) - ); - if (paren) { - if (n) { - PL_regstartp[paren] = HOPc(PL_reginput, -l) - PL_bostr; - PL_regendp[paren] = PL_reginput - PL_bostr; - } - else - PL_regendp[paren] = -1; - } - if (regmatch(next)) - sayYES; - REGCP_UNWIND(lastcp); - } - /* Couldn't or didn't -- back up. */ - n--; - locinput = HOPc(locinput, -l); - PL_reginput = locinput; + if (ST.me->flags) { + /* mark current A as captured */ + I32 paren = ST.me->flags; + if (ST.count) { + 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 == ST.me->flags) + { + if (ST.count) + goto fake_end; + else + sayNO; + } } - sayNO; - break; - } - case CURLYN: - paren = scan->flags; /* Which paren to set */ - if (paren > PL_regsize) - PL_regsize = paren; - if (paren > (I32)*PL_reglastparen) - *PL_reglastparen = paren; - ln = ARG1(scan); /* min to match */ - n = ARG2(scan); /* max to match */ - scan = regnext(NEXTOPER(scan) + NODE_STEP_REGNODE); - goto repeat; - case CURLY: - paren = 0; - ln = ARG1(scan); /* min to match */ - n = ARG2(scan); /* max to match */ - scan = NEXTOPER(scan) + NODE_STEP_REGNODE; - goto repeat; - case STAR: - ln = 0; - n = REG_INFTY; + + PUSH_STATE_GOTO(CURLYM_B, ST.B); /* match B */ + /* NOTREACHED */ + + case CURLYM_B_fail: /* just failed to match a B */ + REGCP_UNWIND(ST.cp); + if (ST.minmod) { + if (ST.count == ARG2(ST.me) /* max */) + sayNO; + goto curlym_do_A; /* try to match a further A */ + } + /* backtrack one A */ + if (ST.count == ARG1(ST.me) /* min */) + sayNO; + ST.count--; + locinput = HOPc(locinput, -ST.alen); + goto curlym_do_B; /* try to match B */ + +#undef ST +#define ST st->u.curly + +#define CURLY_SETPAREN(paren, success) \ + if (paren) { \ + if (success) { \ + PL_regstartp[paren] = HOPc(locinput, -1) - PL_bostr; \ + PL_regendp[paren] = locinput - PL_bostr; \ + *PL_reglastcloseparen = paren; \ + } \ + else \ + PL_regendp[paren] = -1; \ + } + + case STAR: /* /A*B/ where A is width 1 */ + ST.paren = 0; + ST.min = 0; + ST.max = REG_INFTY; scan = NEXTOPER(scan); - paren = 0; goto repeat; - case PLUS: - ln = 1; - n = REG_INFTY; + case PLUS: /* /A+B/ where A is width 1 */ + ST.paren = 0; + ST.min = 1; + ST.max = REG_INFTY; scan = NEXTOPER(scan); - paren = 0; + goto repeat; + case CURLYN: /* /(A){m,n}B/ where A is width 1 */ + ST.paren = scan->flags; /* Which paren to set */ + if (ST.paren > PL_regsize) + PL_regsize = ST.paren; + if (ST.paren > (I32)*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 == 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 */ + ST.paren = 0; + ST.min = ARG1(scan); /* min to match */ + ST.max = ARG2(scan); /* max to match */ + scan = NEXTOPER(scan) + NODE_STEP_REGNODE; repeat: /* * Lookahead to avoid useless match attempts * when we know what character comes next. - */ - - /* + * * Used to only do .*x and .*?x, but now it allows * for )'s, ('s and (?{ ... })'s to be in the way * of the quantifier and the EXACT-like node. -- japhy */ + if (ST.min > ST.max) /* XXX make this a compile-time check? */ + sayNO; if (HAS_TEXT(next) || JUMPABLE(next)) { U8 *s; regnode *text_node = next; - if (! HAS_TEXT(text_node)) FIND_NEXT_IMPT(text_node); + if (! HAS_TEXT(text_node)) + FIND_NEXT_IMPT(text_node); - if (! HAS_TEXT(text_node)) c1 = c2 = -1000; + if (! HAS_TEXT(text_node)) + ST.c1 = ST.c2 = CHRTEST_VOID; else { - if (PL_regkind[(U8)OP(text_node)] == REF) { - c1 = c2 = -1000; + if (PL_regkind[OP(text_node)] == REF) { + ST.c1 = ST.c2 = CHRTEST_VOID; goto assume_ok_easy; } - else { s = (U8*)STRING(text_node); } + else + s = (U8*)STRING(text_node); if (!UTF) { - c2 = c1 = *s; + ST.c2 = ST.c1 = *s; if (OP(text_node) == EXACTF || OP(text_node) == REFF) - c2 = PL_fold[c1]; + ST.c2 = PL_fold[ST.c1]; else if (OP(text_node) == EXACTFL || OP(text_node) == REFFL) - c2 = PL_fold_locale[c1]; + ST.c2 = PL_fold_locale[ST.c1]; } else { /* UTF */ if (OP(text_node) == EXACTF || OP(text_node) == REFF) { @@ -3892,322 +4259,353 @@ S_regmatch(pTHX_ regnode *prog) to_utf8_lower((U8*)s, tmpbuf1, &ulen1); to_utf8_upper((U8*)s, tmpbuf2, &ulen2); - - c1 = utf8n_to_uvuni(tmpbuf1, UTF8_MAXBYTES, 0, - ckWARN(WARN_UTF8) ? - 0 : UTF8_ALLOW_ANY); - c2 = utf8n_to_uvuni(tmpbuf2, UTF8_MAXBYTES, 0, - ckWARN(WARN_UTF8) ? - 0 : UTF8_ALLOW_ANY); +#ifdef EBCDIC + ST.c1 = utf8n_to_uvchr(tmpbuf1, UTF8_MAXLEN, 0, + ckWARN(WARN_UTF8) ? + 0 : UTF8_ALLOW_ANY); + ST.c2 = utf8n_to_uvchr(tmpbuf2, UTF8_MAXLEN, 0, + ckWARN(WARN_UTF8) ? + 0 : UTF8_ALLOW_ANY); +#else + ST.c1 = utf8n_to_uvuni(tmpbuf1, UTF8_MAXBYTES, 0, + uniflags); + ST.c2 = utf8n_to_uvuni(tmpbuf2, UTF8_MAXBYTES, 0, + uniflags); +#endif } else { - c2 = c1 = utf8n_to_uvchr(s, UTF8_MAXBYTES, 0, - ckWARN(WARN_UTF8) ? - 0 : UTF8_ALLOW_ANY); + ST.c2 = ST.c1 = utf8n_to_uvchr(s, UTF8_MAXBYTES, 0, + uniflags); } } } } else - c1 = c2 = -1000; + ST.c1 = ST.c2 = CHRTEST_VOID; assume_ok_easy: + + ST.A = scan; + ST.B = next; PL_reginput = locinput; if (minmod) { - CHECKPOINT lastcp; minmod = 0; - if (ln && regrepeat(scan, ln) < ln) + if (ST.min && regrepeat(rex, ST.A, ST.min) < ST.min) sayNO; + ST.count = ST.min; locinput = PL_reginput; - REGCP_SET(lastcp); - if (c1 != -1000) { - char *e; /* Should not check after this */ - char *old = locinput; - int count = 0; - - if (n == REG_INFTY) { - e = PL_regeol - 1; - if (do_utf8) - while (UTF8_IS_CONTINUATION(*(U8*)e)) - e--; - } - else if (do_utf8) { - int m = n - ln; - for (e = locinput; - m >0 && e + UTF8SKIP(e) <= PL_regeol; m--) - e += UTF8SKIP(e); - } - else { - e = locinput + n - ln; - if (e >= PL_regeol) - e = PL_regeol - 1; - } - while (1) { - /* Find place 'next' could work */ - if (!do_utf8) { - if (c1 == c2) { - while (locinput <= e && - UCHARAT(locinput) != c1) - locinput++; - } else { - while (locinput <= e - && UCHARAT(locinput) != c1 - && UCHARAT(locinput) != c2) - locinput++; - } - count = locinput - old; - } - else { - STRLEN len; - if (c1 == c2) { - /* count initialised to - * utf8_distance(old, locinput) */ - while (locinput <= e && - utf8n_to_uvchr((U8*)locinput, - UTF8_MAXBYTES, &len, - ckWARN(WARN_UTF8) ? - 0 : UTF8_ALLOW_ANY) != (UV)c1) { - locinput += len; - count++; - } - } else { - /* count initialised to - * utf8_distance(old, locinput) */ - while (locinput <= e) { - UV c = utf8n_to_uvchr((U8*)locinput, - UTF8_MAXBYTES, &len, - ckWARN(WARN_UTF8) ? - 0 : UTF8_ALLOW_ANY); - if (c == (UV)c1 || c == (UV)c2) - break; - locinput += len; - count++; - } - } - } - if (locinput > e) - sayNO; - /* PL_reginput == old now */ - if (locinput != old) { - ln = 1; /* Did some */ - if (regrepeat(scan, count) < count) - sayNO; - } - /* PL_reginput == locinput now */ - TRYPAREN(paren, ln, locinput); - PL_reginput = locinput; /* Could be reset... */ - REGCP_UNWIND(lastcp); - /* Couldn't or didn't -- move forward. */ - old = locinput; - if (do_utf8) - locinput += UTF8SKIP(locinput); - else - locinput++; - count = 1; - } + REGCP_SET(ST.cp); + if (ST.c1 == CHRTEST_VOID) + goto curly_try_B_min; + + ST.oldloc = locinput; + + /* set ST.maxpos to the furthest point along the + * string that could possibly match */ + if (ST.max == REG_INFTY) { + ST.maxpos = PL_regeol - 1; + if (do_utf8) + while (UTF8_IS_CONTINUATION(*(U8*)ST.maxpos)) + ST.maxpos--; } - else - while (n >= ln || (n == REG_INFTY && ln > 0)) { /* ln overflow ? */ - UV c; - if (c1 != -1000) { - if (do_utf8) - c = utf8n_to_uvchr((U8*)PL_reginput, - UTF8_MAXBYTES, 0, - ckWARN(WARN_UTF8) ? - 0 : UTF8_ALLOW_ANY); - else - c = UCHARAT(PL_reginput); - /* If it could work, try it. */ - if (c == (UV)c1 || c == (UV)c2) - { - TRYPAREN(paren, ln, PL_reginput); - REGCP_UNWIND(lastcp); - } - } - /* If it could work, try it. */ - else if (c1 == -1000) - { - TRYPAREN(paren, ln, PL_reginput); - REGCP_UNWIND(lastcp); - } - /* Couldn't or didn't -- move forward. */ - PL_reginput = locinput; - if (regrepeat(scan, 1)) { - ln++; - locinput = PL_reginput; - } - else - sayNO; + else if (do_utf8) { + int m = ST.max - ST.min; + for (ST.maxpos = locinput; + m >0 && ST.maxpos + UTF8SKIP(ST.maxpos) <= PL_regeol; m--) + ST.maxpos += UTF8SKIP(ST.maxpos); + } + else { + ST.maxpos = locinput + ST.max - ST.min; + if (ST.maxpos >= PL_regeol) + ST.maxpos = PL_regeol - 1; } + goto curly_try_B_min_known; + } else { - CHECKPOINT lastcp; - n = regrepeat(scan, n); + ST.count = regrepeat(rex, ST.A, ST.max); locinput = PL_reginput; - if (ln < n && PL_regkind[(U8)OP(next)] == EOL && - (OP(next) != MEOL || - OP(next) == SEOL || OP(next) == EOS)) + if (ST.count < ST.min) + sayNO; + if ((ST.count > ST.min) + && (PL_regkind[OP(ST.B)] == EOL) && (OP(ST.B) != MEOL)) { - ln = n; /* why back off? */ - /* ...because $ and \Z can match before *and* after + /* A{m,n} must come at the end of the string, there's + * no point in backing off ... */ + ST.min = ST.count; + /* ...except that $ and \Z can match before *and* after newline at the end. Consider "\n\n" =~ /\n+\Z\n/. - We should back off by one in this case. */ - if (UCHARAT(PL_reginput - 1) == '\n' && OP(next) != EOS) - ln--; + We may back off by one in this case. */ + if (UCHARAT(PL_reginput - 1) == '\n' && OP(ST.B) != EOS) + ST.min--; } - REGCP_SET(lastcp); - if (paren) { - UV c = 0; - while (n >= ln) { - if (c1 != -1000) { - if (do_utf8) - c = utf8n_to_uvchr((U8*)PL_reginput, - UTF8_MAXBYTES, 0, - ckWARN(WARN_UTF8) ? - 0 : UTF8_ALLOW_ANY); - else - c = UCHARAT(PL_reginput); + REGCP_SET(ST.cp); + goto curly_try_B_max; + } + /* NOTREACHED */ + + + case CURLY_B_min_known_fail: + /* failed to find B in a non-greedy match where c1,c2 valid */ + if (ST.paren && ST.count) + PL_regendp[ST.paren] = -1; + + PL_reginput = locinput; /* Could be reset... */ + REGCP_UNWIND(ST.cp); + /* Couldn't or didn't -- move forward. */ + ST.oldloc = locinput; + if (do_utf8) + locinput += UTF8SKIP(locinput); + else + locinput++; + ST.count++; + curly_try_B_min_known: + /* find the next place where 'B' could work, then call B */ + { + int n; + if (do_utf8) { + n = (ST.oldloc == locinput) ? 0 : 1; + if (ST.c1 == ST.c2) { + STRLEN len; + /* set n to utf8_distance(oldloc, locinput) */ + while (locinput <= ST.maxpos && + utf8n_to_uvchr((U8*)locinput, + UTF8_MAXBYTES, &len, + uniflags) != (UV)ST.c1) { + locinput += len; + n++; + } + } + else { + /* set n to utf8_distance(oldloc, locinput) */ + while (locinput <= ST.maxpos) { + STRLEN len; + const UV c = utf8n_to_uvchr((U8*)locinput, + UTF8_MAXBYTES, &len, + uniflags); + if (c == (UV)ST.c1 || c == (UV)ST.c2) + break; + locinput += len; + n++; } - /* If it could work, try it. */ - if (c1 == -1000 || c == (UV)c1 || c == (UV)c2) - { - TRYPAREN(paren, n, PL_reginput); - REGCP_UNWIND(lastcp); - } - /* Couldn't or didn't -- back up. */ - n--; - PL_reginput = locinput = HOPc(locinput, -1); } } else { - UV c = 0; - while (n >= ln) { - if (c1 != -1000) { - if (do_utf8) - c = utf8n_to_uvchr((U8*)PL_reginput, - UTF8_MAXBYTES, 0, - ckWARN(WARN_UTF8) ? - 0 : UTF8_ALLOW_ANY); - else - c = UCHARAT(PL_reginput); - } - /* If it could work, try it. */ - if (c1 == -1000 || c == (UV)c1 || c == (UV)c2) - { - TRYPAREN(paren, n, PL_reginput); - REGCP_UNWIND(lastcp); - } - /* Couldn't or didn't -- back up. */ - n--; - PL_reginput = locinput = HOPc(locinput, -1); + if (ST.c1 == ST.c2) { + while (locinput <= ST.maxpos && + UCHARAT(locinput) != ST.c1) + locinput++; + } + else { + while (locinput <= ST.maxpos + && UCHARAT(locinput) != ST.c1 + && UCHARAT(locinput) != ST.c2) + locinput++; } + n = locinput - ST.oldloc; + } + if (locinput > ST.maxpos) + sayNO; + /* PL_reginput == oldloc now */ + if (n) { + ST.count += n; + if (regrepeat(rex, ST.A, n) < 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 == ST.paren) { + goto fake_end; + } + PUSH_STATE_GOTO(CURLY_B_min_known, ST.B); + } + /* NOTREACHED */ + + + case CURLY_B_min_fail: + /* failed to find B in a non-greedy match where c1,c2 invalid */ + if (ST.paren && ST.count) + PL_regendp[ST.paren] = -1; + + REGCP_UNWIND(ST.cp); + /* failed -- move forward one */ + PL_reginput = locinput; + if (regrepeat(rex, ST.A, 1)) { + ST.count++; + locinput = PL_reginput; + if (ST.count <= ST.max || (ST.max == REG_INFTY && + ST.count > 0)) /* count overflow ? */ + { + 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 == ST.paren) { + goto fake_end; + } + PUSH_STATE_GOTO(CURLY_B_min, ST.B); } } sayNO; - break; - case END: - if (PL_reg_call_cc) { - re_cc_state *cur_call_cc = PL_reg_call_cc; - CURCUR *cctmp = PL_regcc; - regexp *re = PL_reg_re; - CHECKPOINT cp, lastcp; - - cp = regcppush(0); /* Save *all* the positions. */ - REGCP_SET(lastcp); - regcp_set_to(PL_reg_call_cc->ss); /* Restore parens of - the caller. */ - PL_reginput = locinput; /* Make position available to - the callcc. */ - cache_re(PL_reg_call_cc->re); - PL_regcc = PL_reg_call_cc->cc; - PL_reg_call_cc = PL_reg_call_cc->prev; - if (regmatch(cur_call_cc->node)) { - PL_reg_call_cc = cur_call_cc; - regcpblow(cp); - sayYES; + /* NOTREACHED */ + + + curly_try_B_max: + /* a successful greedy match: now try to match B */ + { + UV c = 0; + if (ST.c1 != CHRTEST_VOID) + c = do_utf8 ? utf8n_to_uvchr((U8*)PL_reginput, + UTF8_MAXBYTES, 0, uniflags) + : (UV) UCHARAT(PL_reginput); + /* If it could work, try it. */ + if (ST.c1 == CHRTEST_VOID || c == (UV)ST.c1 || c == (UV)ST.c2) { + CURLY_SETPAREN(ST.paren, ST.count); + if (cur_eval && cur_eval->u.eval.close_paren && + cur_eval->u.eval.close_paren == ST.paren) { + goto fake_end; + } + PUSH_STATE_GOTO(CURLY_B_max, ST.B); + /* NOTREACHED */ } - REGCP_UNWIND(lastcp); - regcppop(); - PL_reg_call_cc = cur_call_cc; - PL_regcc = cctmp; - PL_reg_re = re; - cache_re(re); + } + /* FALL THROUGH */ + case CURLY_B_max_fail: + /* failed to find B in a greedy match */ + if (ST.paren && ST.count) + PL_regendp[ST.paren] = -1; + + REGCP_UNWIND(ST.cp); + /* back up. */ + if (--ST.count < ST.min) + sayNO; + PL_reginput = locinput = HOPc(locinput, -1); + goto curly_try_B_max; + +#undef ST + + + case END: + fake_end: + if (cur_eval) { + /* we've just finished A in /(??{A})B/; now continue with B */ + I32 tmpix; + + + st->u.eval.toggle_reg_flags + = cur_eval->u.eval.toggle_reg_flags; + PL_reg_flags ^= st->u.eval.toggle_reg_flags; + + st->u.eval.prev_rex = rex; /* inner */ + rex = cur_eval->u.eval.prev_rex; /* outer */ + cur_curlyx = cur_eval->u.eval.prev_curlyx; + ReREFCNT_inc(rex); + st->u.eval.cp = regcppush(0); /* Save *all* the positions. */ + REGCP_SET(st->u.eval.lastcp); + PL_reginput = locinput; + /* Restore parens of the outer rex without popping the + * savestack */ + tmpix = PL_savestack_ix; + PL_savestack_ix = cur_eval->u.eval.lastcp; + regcppop(rex); + PL_savestack_ix = tmpix; + + st->u.eval.prev_eval = cur_eval; + cur_eval = cur_eval->u.eval.prev_eval; DEBUG_EXECUTE_r( - PerlIO_printf(Perl_debug_log, - "%*s continuation failed...\n", - REPORT_CODE_OFF+PL_regindent*2, "") - ); - sayNO_SILENT; + PerlIO_printf(Perl_debug_log, "%*s EVAL trying tail ... %x\n", + REPORT_CODE_OFF+depth*2, "",(int)cur_eval);); + PUSH_YES_STATE_GOTO(EVAL_AB, + st->u.eval.prev_eval->u.eval.B); /* match B */ } - if (locinput < PL_regtill) { + + if (locinput < reginfo->till) { DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "%sMatch possible, but length=%ld is smaller than requested=%ld, failing!%s\n", PL_colors[4], (long)(locinput - PL_reg_starttry), - (long)(PL_regtill - PL_reg_starttry), + (long)(reginfo->till - PL_reg_starttry), PL_colors[5])); - sayNO_FINAL; /* Cannot match: too short. */ + sayNO_SILENT; /* Cannot match: too short. */ } PL_reginput = locinput; /* put where regtry can find it */ - sayYES_FINAL; /* Success! */ - case SUCCEED: + sayYES; /* Success! */ + + case SUCCEED: /* successful SUSPEND/UNLESSM/IFMATCH/CURLYM */ + DEBUG_EXECUTE_r( + PerlIO_printf(Perl_debug_log, + "%*s %ssubpattern success...%s\n", + REPORT_CODE_OFF+depth*2, "", PL_colors[4], PL_colors[5])); PL_reginput = locinput; /* put where regtry can find it */ - sayYES_LOUD; /* Success! */ - case SUSPEND: - n = 1; + sayYES; /* Success! */ + +#undef ST +#define ST st->u.ifmatch + + case SUSPEND: /* (?>A) */ + ST.wanted = 1; PL_reginput = locinput; goto do_ifmatch; - case UNLESSM: - n = 0; - if (scan->flags) { - s = HOPBACKc(locinput, scan->flags); - if (!s) - goto say_yes; - PL_reginput = s; - } - else - PL_reginput = locinput; - PL_reg_flags ^= RF_false; - goto do_ifmatch; - case IFMATCH: - n = 1; + + case UNLESSM: /* -ve lookaround: (?!A), or with flags, (?flags) { - s = HOPBACKc(locinput, scan->flags); - if (!s) - goto say_no; + char * const s = HOPBACKc(locinput, scan->flags); + if (!s) { + /* trivial fail */ + if (logical) { + logical = 0; + sw = 1 - (bool)ST.wanted; + } + else if (ST.wanted) + sayNO; + next = scan + ARG(scan); + if (next == scan) + next = NULL; + break; + } PL_reginput = s; } else PL_reginput = locinput; do_ifmatch: - inner = NEXTOPER(NEXTOPER(scan)); - if (regmatch(inner) != n) { - if (n == 0) - PL_reg_flags ^= RF_false; - say_no: - if (logical) { - logical = 0; - sw = 0; - goto do_longjump; - } - else - sayNO; - } - if (n == 0) - PL_reg_flags ^= RF_false; - say_yes: - if (logical) { - logical = 0; - sw = 1; + ST.me = scan; + ST.logical = logical; + /* execute body of (?...A) */ + PUSH_YES_STATE_GOTO(IFMATCH_A, NEXTOPER(NEXTOPER(scan))); + /* NOTREACHED */ + + case IFMATCH_A_fail: /* body of (?...A) failed */ + ST.wanted = !ST.wanted; + /* FALL THROUGH */ + + case IFMATCH_A: /* body of (?...A) succeeded */ + if (ST.logical) { + sw = (bool)ST.wanted; } - if (OP(scan) == SUSPEND) { + else if (!ST.wanted) + sayNO; + + if (OP(ST.me) == SUSPEND) locinput = PL_reginput; + else { + locinput = PL_reginput = st->locinput; nextchr = UCHARAT(locinput); } - /* FALL THROUGH. */ + scan = ST.me + ARG(ST.me); + if (scan == ST.me) + scan = NULL; + continue; /* execute B */ + +#undef ST + case LONGJMP: - do_longjump: next = scan + ARG(scan); if (next == scan) next = NULL; @@ -4217,8 +4615,35 @@ S_regmatch(pTHX_ regnode *prog) PTR2UV(scan), OP(scan)); Perl_croak(aTHX_ "regexp memory corruption"); } - reenter: + scan = next; + continue; + /* NOTREACHED */ + + push_yes_state: + /* push a state that backtracks on success */ + st->u.yes.prev_yes_state = yes_state; + yes_state = st; + /* FALL THROUGH */ + push_state: + /* push a new regex state, then continue at scan */ + { + regmatch_state *newst; + + 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; + + locinput = PL_reginput; + nextchr = UCHARAT(locinput); + st = newst; + continue; + /* NOTREACHED */ + } } /* @@ -4229,89 +4654,94 @@ S_regmatch(pTHX_ regnode *prog) /*NOTREACHED*/ sayNO; -yes_loud: - DEBUG_EXECUTE_r( - PerlIO_printf(Perl_debug_log, - "%*s %scould match...%s\n", - REPORT_CODE_OFF+PL_regindent*2, "", PL_colors[4],PL_colors[5]) - ); - goto yes; -yes_final: - DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "%sMatch successful!%s\n", - PL_colors[4],PL_colors[5])); yes: + if (yes_state) { + /* we have successfully completed a subexpression, but we must now + * pop to the state marked by yes_state and continue from there */ + assert(st != yes_state); #ifdef DEBUGGING - PL_regindent--; + while (st != yes_state) { + st--; + if (st < SLAB_FIRST(PL_regmatch_slab)) { + PL_regmatch_slab = PL_regmatch_slab->prev; + st = SLAB_LAST(PL_regmatch_slab); + } + DEBUG_STATE_pp("pop (yes)"); + depth--; + } +#else + while (yes_state < SLAB_FIRST(PL_regmatch_slab) + || yes_state > SLAB_LAST(PL_regmatch_slab)) + { + /* not in this slab, pop slab */ + depth -= (st - SLAB_FIRST(PL_regmatch_slab) + 1); + PL_regmatch_slab = PL_regmatch_slab->prev; + st = SLAB_LAST(PL_regmatch_slab); + } + depth -= (st - yes_state); #endif + st = yes_state; + yes_state = st->u.yes.prev_yes_state; + PL_regmatch_state = st; -#if 0 /* Breaks $^R */ - if (unwind) - regcpblow(firstcp); -#endif - return 1; + state_num = st->resume_state; + goto reenter_switch; + } + + DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "%sMatch successful!%s\n", + PL_colors[4], PL_colors[5])); + + result = 1; + goto final_exit; no: DEBUG_EXECUTE_r( PerlIO_printf(Perl_debug_log, - "%*s %sfailed...%s\n", - REPORT_CODE_OFF+PL_regindent*2, "",PL_colors[4],PL_colors[5]) + "%*s %sfailed...%s\n", + REPORT_CODE_OFF+depth*2, "", + PL_colors[4], PL_colors[5]) ); - goto do_no; -no_final: -do_no: - if (unwind) { - re_unwind_t *uw = SSPTRt(unwind,re_unwind_t); - - switch (uw->type) { - case RE_UNWIND_BRANCH: - case RE_UNWIND_BRANCHJ: - { - re_unwind_branch_t *uwb = &(uw->branch); - I32 lastparen = uwb->lastparen; - - REGCP_UNWIND(uwb->lastcp); - for (n = *PL_reglastparen; n > lastparen; n--) - PL_regendp[n] = -1; - *PL_reglastparen = n; - scan = next = uwb->next; - if ( !scan || - OP(scan) != (uwb->type == RE_UNWIND_BRANCH - ? BRANCH : BRANCHJ) ) { /* Failure */ - unwind = uwb->prev; -#ifdef DEBUGGING - PL_regindent--; -#endif - goto do_no; - } - /* Have more choice yet. Reuse the same uwb. */ - /*SUPPRESS 560*/ - if ((n = (uwb->type == RE_UNWIND_BRANCH - ? NEXT_OFF(next) : ARG(next)))) - next += n; - else - next = NULL; /* XXXX Needn't unwinding in this case... */ - uwb->next = next; - next = NEXTOPER(scan); - if (uwb->type == RE_UNWIND_BRANCHJ) - next = NEXTOPER(next); - locinput = uwb->locinput; - nextchr = uwb->nextchr; -#ifdef DEBUGGING - PL_regindent = uwb->regindent; -#endif - goto reenter; +no_silent: + if (depth) { + /* there's a previous state to backtrack to */ + st--; + if (st < SLAB_FIRST(PL_regmatch_slab)) { + PL_regmatch_slab = PL_regmatch_slab->prev; + st = SLAB_LAST(PL_regmatch_slab); } - /* NOT REACHED */ - default: - Perl_croak(aTHX_ "regexp unwind memory corruption"); + PL_regmatch_state = st; + locinput= st->locinput; + nextchr = UCHARAT(locinput); + + DEBUG_STATE_pp("pop"); + depth--; + if (yes_state == st) + yes_state = st->u.yes.prev_yes_state; + + state_num = st->resume_state + 1; /* failure = success + 1 */ + goto reenter_switch; + } + result = 0; + + final_exit: + + /* restore original high-water mark */ + PL_regmatch_slab = orig_slab; + PL_regmatch_state = orig_state; + + /* free all slabs above current one */ + if (orig_slab->next) { + regmatch_slab *sl = orig_slab->next; + orig_slab->next = NULL; + while (sl) { + regmatch_slab * const osl = sl; + sl = sl->next; + Safefree(osl); } - /* NOT REACHED */ } -#ifdef DEBUGGING - PL_regindent--; -#endif - return 0; + + return result; } /* @@ -4323,8 +4753,9 @@ do_no: * rather than incrementing count on every character. [Er, except utf8.]] */ STATIC I32 -S_regrepeat(pTHX_ regnode *p, I32 max) +S_regrepeat(pTHX_ const regexp *prog, const regnode *p, I32 max) { + dVAR; register char *scan; register I32 c; register char *loceol = PL_regeol; @@ -4335,7 +4766,7 @@ S_regrepeat(pTHX_ regnode *p, I32 max) if (max == REG_INFTY) max = I32_MAX; else if (max < loceol - scan) - loceol = scan + max; + loceol = scan + max; switch (OP(p)) { case REG_ANY: if (do_utf8) { @@ -4385,19 +4816,19 @@ S_regrepeat(pTHX_ regnode *p, I32 max) if (do_utf8) { loceol = PL_regeol; while (hardcount < max && scan < loceol && - reginclass(p, (U8*)scan, 0, do_utf8)) { + reginclass(prog, p, (U8*)scan, 0, do_utf8)) { scan += UTF8SKIP(scan); hardcount++; } } else { - while (scan < loceol && REGINCLASS(p, (U8*)scan)) + while (scan < loceol && REGINCLASS(prog, p, (U8*)scan)) scan++; } break; case ALNUM: if (do_utf8) { loceol = PL_regeol; - LOAD_UTF8_CHARCLASS(alnum,"a"); + LOAD_UTF8_CHARCLASS_ALNUM(); while (hardcount < max && scan < loceol && swash_fetch(PL_utf8_alnum, (U8*)scan, do_utf8)) { scan += UTF8SKIP(scan); @@ -4425,7 +4856,7 @@ S_regrepeat(pTHX_ regnode *p, I32 max) case NALNUM: if (do_utf8) { loceol = PL_regeol; - LOAD_UTF8_CHARCLASS(alnum,"a"); + LOAD_UTF8_CHARCLASS_ALNUM(); while (hardcount < max && scan < loceol && !swash_fetch(PL_utf8_alnum, (U8*)scan, do_utf8)) { scan += UTF8SKIP(scan); @@ -4453,7 +4884,7 @@ S_regrepeat(pTHX_ regnode *p, I32 max) case SPACE: if (do_utf8) { loceol = PL_regeol; - LOAD_UTF8_CHARCLASS(space," "); + LOAD_UTF8_CHARCLASS_SPACE(); while (hardcount < max && scan < loceol && (*scan == ' ' || swash_fetch(PL_utf8_space,(U8*)scan, do_utf8))) { @@ -4482,7 +4913,7 @@ S_regrepeat(pTHX_ regnode *p, I32 max) case NSPACE: if (do_utf8) { loceol = PL_regeol; - LOAD_UTF8_CHARCLASS(space," "); + LOAD_UTF8_CHARCLASS_SPACE(); while (hardcount < max && scan < loceol && !(*scan == ' ' || swash_fetch(PL_utf8_space,(U8*)scan, do_utf8))) { @@ -4511,7 +4942,7 @@ S_regrepeat(pTHX_ regnode *p, I32 max) case DIGIT: if (do_utf8) { loceol = PL_regeol; - LOAD_UTF8_CHARCLASS(digit,"0"); + LOAD_UTF8_CHARCLASS_DIGIT(); while (hardcount < max && scan < loceol && swash_fetch(PL_utf8_digit, (U8*)scan, do_utf8)) { scan += UTF8SKIP(scan); @@ -4525,7 +4956,7 @@ S_regrepeat(pTHX_ regnode *p, I32 max) case NDIGIT: if (do_utf8) { loceol = PL_regeol; - LOAD_UTF8_CHARCLASS(digit,"0"); + LOAD_UTF8_CHARCLASS_DIGIT(); while (hardcount < max && scan < loceol && !swash_fetch(PL_utf8_digit, (U8*)scan, do_utf8)) { scan += UTF8SKIP(scan); @@ -4547,97 +4978,48 @@ S_regrepeat(pTHX_ regnode *p, I32 max) PL_reginput = scan; DEBUG_r({ - SV *re_debug_flags; - SV *prop = sv_newmortal(); - GET_RE_DEBUG_FLAGS; - DEBUG_EXECUTE_r({ - regprop(prop, p); - PerlIO_printf(Perl_debug_log, - "%*s %s can match %"IVdf" times out of %"IVdf"...\n", - REPORT_CODE_OFF+1, "", SvPVX(prop),(IV)c,(IV)max); - }); + GET_RE_DEBUG_FLAGS_DECL; + DEBUG_EXECUTE_r({ + SV * const prop = sv_newmortal(); + 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); }); + }); return(c); } -/* - - regrepeat_hard - repeatedly match something, report total lenth and length - * - * The repeater is supposed to have constant non-zero length. - */ - -STATIC I32 -S_regrepeat_hard(pTHX_ regnode *p, I32 max, I32 *lp) -{ - register char *scan = Nullch; - register char *start; - register char *loceol = PL_regeol; - I32 l = 0; - I32 count = 0, res = 1; - - if (!max) - return 0; - - start = PL_reginput; - if (PL_reg_match_utf8) { - while (PL_reginput < loceol && (scan = PL_reginput, res = regmatch(p))) { - if (!count++) { - l = 0; - while (start < PL_reginput) { - l++; - start += UTF8SKIP(start); - } - *lp = l; - if (l == 0) - return max; - } - if (count == max) - return count; - } - } - else { - while (PL_reginput < loceol && (scan = PL_reginput, res = regmatch(p))) { - if (!count++) { - *lp = l = PL_reginput - start; - if (max != REG_INFTY && l*max < loceol - scan) - loceol = scan + l*max; - if (l == 0) - return max; - } - } - } - if (!res) - PL_reginput = scan; - - return count; -} +#if !defined(PERL_IN_XSUB_RE) || defined(PLUGGABLE_RE_EXTENSION) /* - regclass_swash - prepare the utf8 swash */ SV * -Perl_regclass_swash(pTHX_ register regnode* node, bool doinit, SV** listsvp, SV **altsvp) +Perl_regclass_swash(pTHX_ const regexp *prog, register const regnode* node, bool doinit, SV** listsvp, SV **altsvp) { + dVAR; SV *sw = NULL; SV *si = NULL; SV *alt = NULL; + const struct reg_data * const data = prog ? prog->data : NULL; - if (PL_regdata && PL_regdata->count) { - U32 n = ARG(node); + if (data && data->count) { + const U32 n = ARG(node); - if (PL_regdata->what[n] == 's') { - SV *rv = (SV*)PL_regdata->data[n]; - AV *av = (AV*)SvRV((SV*)rv); - SV **ary = AvARRAY(av); + if (data->what[n] == 's') { + SV * const rv = (SV*)data->data[n]; + AV * const av = (AV*)SvRV((SV*)rv); + SV **const ary = AvARRAY(av); SV **a, **b; - /* See the end of regcomp.c:S_reglass() for + /* See the end of regcomp.c:S_regclass() for * documentation of these array elements. */ si = *ary; - a = SvTYPE(ary[1]) == SVt_RV ? &ary[1] : 0; + a = SvROK(ary[1]) ? &ary[1] : 0; b = SvTYPE(ary[2]) == SVt_PVAV ? &ary[2] : 0; if (a) @@ -4658,6 +5040,7 @@ Perl_regclass_swash(pTHX_ register regnode* node, bool doinit, SV** listsvp, SV return sw; } +#endif /* - reginclass - determine if a character falls into a character class @@ -4670,17 +5053,22 @@ Perl_regclass_swash(pTHX_ register regnode* node, bool doinit, SV** listsvp, SV */ STATIC bool -S_reginclass(pTHX_ register regnode *n, register U8* p, STRLEN* lenp, register bool do_utf8) +S_reginclass(pTHX_ const regexp *prog, register const regnode *n, register const U8* p, STRLEN* lenp, register bool do_utf8) { - char flags = ANYOF_FLAGS(n); + dVAR; + const char flags = ANYOF_FLAGS(n); bool match = FALSE; UV c = *p; STRLEN len = 0; STRLEN plen; - if (do_utf8 && !UTF8_IS_INVARIANT(c)) - c = utf8n_to_uvchr(p, UTF8_MAXBYTES, &len, - ckWARN(WARN_UTF8) ? 0 : UTF8_ALLOW_ANY); + if (do_utf8 && !UTF8_IS_INVARIANT(c)) { + c = utf8n_to_uvchr(p, UTF8_MAXBYTES, &len, + (UTF8_ALLOW_DEFAULT & UTF8_ALLOW_ANYUV) | UTF8_CHECK_ONLY); + /* see [perl #37836] for UTF8_ALLOW_ANYUV */ + if (len == (STRLEN)-1) + Perl_croak(aTHX_ "Malformed UTF-8 character (fatal)"); + } plen = lenp ? *lenp : UNISKIP(NATIVE_TO_UNI(c)); if (do_utf8 || (flags & ANYOF_UNICODE)) { @@ -4694,7 +5082,7 @@ S_reginclass(pTHX_ register regnode *n, register U8* p, STRLEN* lenp, register b match = TRUE; if (!match) { AV *av; - SV *sw = regclass_swash(n, TRUE, 0, (SV**)&av); + SV * const sw = regclass_swash(prog, n, TRUE, 0, (SV**)&av); if (sw) { if (swash_fetch(sw, p, do_utf8)) @@ -4702,11 +5090,10 @@ S_reginclass(pTHX_ register regnode *n, register U8* p, STRLEN* lenp, register b else if (flags & ANYOF_FOLD) { if (!match && lenp && av) { I32 i; - for (i = 0; i <= av_len(av); i++) { - SV* sv = *av_fetch(av, i, FALSE); + SV* const sv = *av_fetch(av, i, FALSE); STRLEN len; - char *s = SvPV(sv, len); + const char * const s = SvPV_const(sv, len); if (len <= plen && memEQ(s, (char*)p, len)) { *lenp = len; @@ -4789,14 +5176,9 @@ S_reginclass(pTHX_ register regnode *n, register U8* p, STRLEN* lenp, register b } STATIC U8 * -S_reghop(pTHX_ U8 *s, I32 off) -{ - return S_reghop3(aTHX_ s, off, (U8*)(off >= 0 ? PL_regeol : PL_bostr)); -} - -STATIC U8 * -S_reghop3(pTHX_ U8 *s, I32 off, U8* lim) +S_reghop3(U8 *s, I32 off, const U8* lim) { + dVAR; if (off >= 0) { while (off-- && s < lim) { /* XXX could check well-formedness here */ @@ -4804,52 +5186,70 @@ S_reghop3(pTHX_ U8 *s, I32 off, U8* lim) } } else { - while (off++) { - if (s > lim) { - s--; - if (UTF8_IS_CONTINUED(*s)) { - while (s > (U8*)lim && UTF8_IS_CONTINUATION(*s)) - s--; - } - /* XXX could check well-formedness here */ + while (off++ && s > lim) { + s--; + if (UTF8_IS_CONTINUED(*s)) { + while (s > lim && UTF8_IS_CONTINUATION(*s)) + s--; } + /* XXX could check well-formedness here */ } } return s; } +#ifdef XXX_dmq +/* there are a bunch of places where we use two reghop3's that should + be replaced with this routine. but since thats not done yet + we ifdef it out - dmq +*/ STATIC U8 * -S_reghopmaybe(pTHX_ U8 *s, I32 off) +S_reghop4(U8 *s, I32 off, const U8* llim, const U8* rlim) { - return S_reghopmaybe3(aTHX_ s, off, (U8*)(off >= 0 ? PL_regeol : PL_bostr)); + dVAR; + if (off >= 0) { + while (off-- && s < rlim) { + /* XXX could check well-formedness here */ + s += UTF8SKIP(s); + } + } + else { + while (off++ && s > llim) { + s--; + if (UTF8_IS_CONTINUED(*s)) { + while (s > llim && UTF8_IS_CONTINUATION(*s)) + s--; + } + /* XXX could check well-formedness here */ + } + } + return s; } +#endif STATIC U8 * -S_reghopmaybe3(pTHX_ U8* s, I32 off, U8* lim) +S_reghopmaybe3(U8* s, I32 off, const U8* lim) { + dVAR; if (off >= 0) { while (off-- && s < lim) { /* XXX could check well-formedness here */ s += UTF8SKIP(s); } if (off >= 0) - return 0; + return NULL; } else { - while (off++) { - if (s > lim) { - s--; - if (UTF8_IS_CONTINUED(*s)) { - while (s > (U8*)lim && UTF8_IS_CONTINUATION(*s)) - s--; - } - /* XXX could check well-formedness here */ + while (off++ && s > lim) { + s--; + if (UTF8_IS_CONTINUED(*s)) { + while (s > lim && UTF8_IS_CONTINUATION(*s)) + s--; } - else - break; + /* XXX could check well-formedness here */ } if (off <= 0) - return 0; + return NULL; } return s; } @@ -4857,14 +5257,16 @@ S_reghopmaybe3(pTHX_ U8* s, I32 off, U8* lim) static void restore_pos(pTHX_ void *arg) { + dVAR; + regexp * const rex = (regexp *)arg; if (PL_reg_eval_set) { if (PL_reg_oldsaved) { - PL_reg_re->subbeg = PL_reg_oldsaved; - PL_reg_re->sublen = PL_reg_oldsavedlen; -#ifdef PERL_COPY_ON_WRITE - PL_reg_re->saved_copy = PL_nrs; + rex->subbeg = PL_reg_oldsaved; + rex->sublen = PL_reg_oldsavedlen; +#ifdef PERL_OLD_COPY_ON_WRITE + rex->saved_copy = PL_nrs; #endif - RX_MATCH_COPIED_on(PL_reg_re); + RX_MATCH_COPIED_on(rex); } PL_reg_magic->mg_len = PL_reg_oldpos; PL_reg_eval_set = 0; @@ -4875,9 +5277,9 @@ restore_pos(pTHX_ void *arg) STATIC void S_to_utf8_substr(pTHX_ register regexp *prog) { - SV* sv; if (prog->float_substr && !prog->float_utf8) { - prog->float_utf8 = sv = newSVsv(prog->float_substr); + SV* const sv = newSVsv(prog->float_substr); + prog->float_utf8 = sv; sv_utf8_upgrade(sv); if (SvTAIL(prog->float_substr)) SvTAIL_on(sv); @@ -4885,7 +5287,8 @@ S_to_utf8_substr(pTHX_ register regexp *prog) prog->check_utf8 = sv; } if (prog->anchored_substr && !prog->anchored_utf8) { - prog->anchored_utf8 = sv = newSVsv(prog->anchored_substr); + SV* const sv = newSVsv(prog->anchored_substr); + prog->anchored_utf8 = sv; sv_utf8_upgrade(sv); if (SvTAIL(prog->anchored_substr)) SvTAIL_on(sv); @@ -4897,9 +5300,10 @@ S_to_utf8_substr(pTHX_ register regexp *prog) STATIC void S_to_byte_substr(pTHX_ register regexp *prog) { - SV* sv; + dVAR; if (prog->float_utf8 && !prog->float_substr) { - prog->float_substr = sv = newSVsv(prog->float_utf8); + SV* sv = newSVsv(prog->float_utf8); + prog->float_substr = sv; if (sv_utf8_downgrade(sv, TRUE)) { if (SvTAIL(prog->float_utf8)) SvTAIL_on(sv); @@ -4911,7 +5315,8 @@ S_to_byte_substr(pTHX_ register regexp *prog) prog->check_substr = sv; } if (prog->anchored_utf8 && !prog->anchored_substr) { - prog->anchored_substr = sv = newSVsv(prog->anchored_utf8); + SV* sv = newSVsv(prog->anchored_utf8); + prog->anchored_substr = sv; if (sv_utf8_downgrade(sv, TRUE)) { if (SvTAIL(prog->anchored_utf8)) SvTAIL_on(sv); @@ -4923,3 +5328,13 @@ S_to_byte_substr(pTHX_ register regexp *prog) prog->check_substr = sv; } } + +/* + * Local variables: + * c-indentation-style: bsd + * c-basic-offset: 4 + * indent-tabs-mode: t + * End: + * + * ex: set ts=8 sts=4 sw=4 noet: + */