* 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
*/
#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
/*
**** 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.
#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_utf8 8 /* Pattern contains multibyte chars? */
#define UTF ((PL_reg_flags & RF_utf8) != 0)
#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.
#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(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_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 \
)
/*
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)
{
+ dVAR;
const int retval = PL_savestack_ix;
#define REGCP_PAREN_ELEMS 4
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");
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. */
SSPUSHINT(PL_regsize);
}
/* 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;
GET_RE_DEBUG_FLAGS_DECL;
for (i -= (REGCP_OTHER_ELEMS - REGCP_FRAME_ELEMS);
i > 0; i -= REGCP_PAREN_ELEMS) {
I32 tmps;
- paren = (U32)SSPOPINT;
+ U32 paren = (U32)SSPOPINT;
PL_reg_start_tmp[paren] = (char *) SSPOPPTR;
PL_regstartp[paren] = SSPOPINT;
tmps = SSPOPINT;
);
}
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
* 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)
-{
- const 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
*/
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:
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;
register SV *check;
char *strbeg;
char *t;
- const 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 */
+ 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
const char * const i_strpos = strpos;
- SV * const dsv = PERL_DEBUG_PAD_ZERO(0);
#endif
GET_RE_DEBUG_FLAGS_DECL;
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({
- const char *s = PL_reg_match_utf8 ?
- sv_uni_display(dsv, sv, 60, UNI_DISPLAY_REGEX) :
- strpos;
- const 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) {
"String too short... [re_intuit_start]\n"));
goto fail;
}
+
strbeg = (sv && SvPOK(sv)) ? strend - SvCUR(sv) : strpos;
PL_regeol = strend;
if (do_utf8) {
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 */
I32 slen;
s = HOP3c(strpos, prog->check_offset_min, strend);
+
if (SvTAIL(check)) {
slen = SvCUR(check); /* >= 1 */
|| ((slen = SvCUR(check)) > 1
&& 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) {
const I32 end = prog->check_offset_max + CHR_SVLEN(check)
- (SvTAIL(check) != 0);
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_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 * const pp = data ? data->scream_pos : &p;
&& (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 (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_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_const(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
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)
{
char * const last = HOP3c(s, -start_shift, strbeg);
char *last1, *last2;
- char *s1 = s;
+ 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);
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(
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_const(must),
- PL_colors[1], (SvTAIL(must) ? "$" : "")));
+ quoted, RE_SV_TAIL(must));
+ });
+
+
if (!s) {
if (last1 >= last2) {
DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log,
}
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;
(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;
}
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);
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_const(must),
- PL_colors[1], (SvTAIL(must) ? "$" : "")));
+ quoted, RE_SV_TAIL(must));
+ });
if (!s) {
if (last1 == last) {
DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log,
}
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;
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;
}
}
- t = s - prog->check_offset_max;
+
+ t= (char*)HOP3( s, -prog->check_offset_max, (prog->check_offset_max<0) ? strend : strpos);
+
+ DEBUG_OPTIMISE_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:
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
/* 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;
/* If regstclass takes bytelength more than 1: If charlength==1, OK.
This leaves EXACTF only, which is dealt with in find_byclass(). */
const U8* const str = (U8*)STRING(prog->regstclass);
- const int cl_l = (PL_regkind[(U8)OP(prog->regstclass)] == EXACT
+ const int cl_l = (PL_regkind[OP(prog->regstclass)] == EXACT
? CHR_DIST(str+STR_LEN(prog->regstclass), str)
: 1);
- const char * const 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 * 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, 1);
+ s = find_byclass(prog, prog->regstclass, s, endpos, NULL);
if (!s) {
#ifdef DEBUGGING
- const char *what = 0;
+ const char *what = NULL;
#endif
if (endpos == strend) {
DEBUG_EXECUTE_r( PerlIO_printf(Perl_debug_log,
fail:
DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "%sMatch rejected by optimizer%s\n",
PL_colors[4], PL_colors[5]));
- return Nullch;
+ 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, const char *strend, I32 norun)
+S_find_byclass(pTHX_ regexp * prog, const regnode *c, char *s,
+ const char *strend, const regmatch_info *reginfo)
{
dVAR;
const I32 doevery = (prog->reganch & ROPT_SKIP) == 0;
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;
}
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);
U8 *sm = (U8 *) m;
U8 tmpbuf1[UTF8_MAXBYTES_CASE+1];
U8 tmpbuf2[UTF8_MAXBYTES_CASE+1];
- const U32 uniflags = ckWARN(WARN_UTF8) ? 0 : UTF8_ALLOW_ANY;
+ const U32 uniflags = UTF8_ALLOW_DEFAULT;
to_utf8_lower((U8*)m, tmpbuf1, &ulen1);
to_utf8_upper((U8*)m, tmpbuf2, &ulen2);
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
UV c, f;
U8 tmpbuf [UTF8_MAXBYTES+1];
STRLEN len, foldlen;
- const U32 uniflags = ckWARN(WARN_UTF8) ? 0 : UTF8_ALLOW_ANY;
+ 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,
uniflags);
- 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 {
- 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,
- (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);
}
}
else {
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 {
- 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,
- (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:
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();
- while (s + (uskip = UTF8SKIP(s)) <= strend) {
+ 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:
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();
- while (s + (uskip = UTF8SKIP(s)) <= strend) {
+ 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();
- 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();
- 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;
- case DIGIT:
- if (do_utf8) {
- LOAD_UTF8_CHARCLASS_DIGIT();
- 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;
- 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;
- case NDIGIT:
- if (do_utf8) {
- LOAD_UTF8_CHARCLASS_DIGIT();
- 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;
- 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(
+ !(*s == ' ' || isSPACE_LC_utf8((U8*)s)),
+ !isSPACE_LC(*s)
+ );
+ case DIGIT:
+ REXEC_FBC_CSCAN_PRELOAD(
+ LOAD_UTF8_CHARCLASS_DIGIT(),
+ swash_fetch(PL_utf8_digit,(U8*)s, do_utf8),
+ isDIGIT(*s)
+ );
+ case DIGITL:
+ REXEC_FBC_CSCAN_TAINT(
+ isDIGIT_LC_utf8((U8*)s),
+ isDIGIT_LC(*s)
+ );
+ case NDIGIT:
+ REXEC_FBC_CSCAN_PRELOAD(
+ LOAD_UTF8_CHARCLASS_DIGIT(),
+ !swash_fetch(PL_utf8_digit,(U8*)s, do_utf8),
+ !isDIGIT(*s)
+ );
+ case NDIGITL:
+ 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:
/* data: May be used for some additional optimizations. */
/* nosave: For optimizations. */
{
+ dVAR;
register char *s;
register regnode *c;
register char *startpos = stringarg;
I32 dontbother = 0; /* how many characters not to try at end */
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);
+ char *scream_olds = NULL;
+ SV* const oreplsv = GvSV(PL_replgv);
const bool do_utf8 = DO_UTF8(sv);
- const I32 multiline = prog->reganch & PMf_MULTILINE;
-#ifdef DEBUGGING
- SV *dsv0 = PERL_DEBUG_PAD_ZERO(0);
- SV *dsv1 = PERL_DEBUG_PAD_ZERO(1);
-#endif
+ I32 multiline;
+
+ regmatch_info reginfo; /* create some info to pass to regtry etc */
GET_RE_DEBUG_FLAGS_DECL;
PERL_UNUSED_ARG(data);
- RX_MATCH_UTF8_set(prog,do_utf8);
-
- PL_regcc = 0;
-
- cache_re(prog);
-#ifdef DEBUGGING
- PL_regnarrate = DEBUG_r_TEST;
-#endif
/* Be paranoid... */
if (prog == NULL || startpos == NULL) {
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");
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;
}
}
- DEBUG_EXECUTE_r({
- const char * const s0 = UTF
- ? pv_uni_display(dsv0, (U8*)prog->precomp, prog->prelen, 60,
- UNI_DISPLAY_REGEX)
- : prog->precomp;
- const int len0 = UTF ? SvCUR(dsv0) : prog->prelen;
- const char * const s1 = do_utf8 ? sv_uni_display(dsv1, sv, 60,
- UNI_DISPLAY_REGEX) : startpos;
- const 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? */
if (s == startpos)
goto after_try;
while (1) {
- if (regtry(prog, s))
+ if (regtry(®info, s))
goto got_it;
after_try:
if (s >= end)
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))
+ if (regtry(®info, reginfo.ganch))
goto got_it;
goto phooey;
}
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")
);
}
- 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;
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
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 */
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_const(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();
- const char *s0;
- const char *s1;
- int len0;
- int len1;
-
- regprop(prop, c);
- s0 = UTF ?
- pv_uni_display(dsv0, (U8*)SvPVX_const(prop), SvCUR(prop), 60,
- UNI_DISPLAY_REGEX) :
- SvPVX_const(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, 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;
last = strend - len + 1;
else if (!multiline)
last = memEQ(strend - len, little, len)
- ? strend - len : Nullch;
+ ? strend - len : NULL;
else
goto find_last;
} else {
/* 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;
}
else {
do {
- if (regtry(prog, s))
+ if (regtry(®info, s))
goto got_it;
} while (s++ < strend);
}
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);
}
/* 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);
+ 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)) {
DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "%sMatch failed%s\n",
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;
/* 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) {
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);
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 */
}
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;
}
#endif
REGCP_SET(lastcp);
- if (regmatch(prog->program + 1)) {
+ if (regmatch(reginfo, prog->program + 1)) {
prog->endp[0] = PL_reginput - PL_bostr;
return 1;
}
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 POSCACHE_SUCCESS 0 /* caching success rather than failure */
-#define POSCACHE_SEEN 1 /* we know what we're caching */
-#define POSCACHE_START 2 /* the real cache: this bit maps to pos 0 */
-#define CACHEsayYES STMT_START { \
- if (cache_offset | cache_bit) { \
- if (!(PL_reg_poscache[0] & (1<<POSCACHE_SEEN))) \
- PL_reg_poscache[0] |= (1<<POSCACHE_SUCCESS) || (1<<POSCACHE_SEEN); \
- else if (!(PL_reg_poscache[0] & (1<<POSCACHE_SUCCESS))) { \
- /* cache records failure, but this is success */ \
- DEBUG_r( \
- PerlIO_printf(Perl_debug_log, \
- "%*s (remove success from failure cache)\n", \
- REPORT_CODE_OFF+PL_regindent*2, "") \
- ); \
- PL_reg_poscache[cache_offset] &= ~(1<<cache_bit); \
- } \
- } \
- sayYES; \
-} STMT_END
#define CACHEsayNO STMT_START { \
- if (cache_offset | cache_bit) { \
- if (!(PL_reg_poscache[0] & (1<<POSCACHE_SEEN))) \
- PL_reg_poscache[0] |= (1<<POSCACHE_SEEN); \
- else if ((PL_reg_poscache[0] & (1<<POSCACHE_SUCCESS))) { \
- /* cache records success, but this is failure */ \
- DEBUG_r( \
- PerlIO_printf(Perl_debug_log, \
- "%*s (remove failure from success cache)\n", \
- REPORT_CODE_OFF+PL_regindent*2, "") \
- ); \
- PL_reg_poscache[cache_offset] &= ~(1<<cache_bit); \
- } \
- } \
+ if (st->u.whilem.cache_offset | st->u.whilem.cache_bit) \
+ PL_reg_poscache[st->u.whilem.cache_offset] |= \
+ (1<<st->u.whilem.cache_bit); \
sayNO; \
} STMT_END
+
/* 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
/* 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 > 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; \
- } 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);
+}
+
+/* simulate a recursive call to regmatch */
+
+#define REGMATCH(ns, where) \
+ st->scan = scan; \
+ scan = (ns); \
+ st->resume_state = resume_##where; \
+ goto start_recurse; \
+ resume_point_##where:
+
+/* push a new state then goto it */
+
+#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
/* [lwall] I've hoisted the register declarations to the outer block in order to
* maybe save a little bit of pushing and popping on the stack. It also takes
* advantage of machines that use a register save mask on subroutine entry.
+ *
+ * This function used to be heavily recursive, but since this had the
+ * effect of blowing the CPU stack on complex regexes, it has been
+ * restructured to be iterative, and to save state onto the heap rather
+ * than the stack. Essentially whereever regmatch() used to be called, it
+ * pushes the current state, notes where to return, then jumps back into
+ * the main loop.
+ *
+ * Originally the structure of this function used to look something like
+
+ S_regmatch() {
+ int a = 1, b = 2;
+ ...
+ while (scan != NULL) {
+ a++; // do stuff with a and b
+ ...
+ switch (OP(scan)) {
+ case FOO: {
+ int local = 3;
+ ...
+ if (regmatch(...)) // recurse
+ goto yes;
+ }
+ ...
+ }
+ }
+ yes:
+ return 1;
+ }
+
+ * Now it looks something like this:
+
+ typedef struct {
+ int a, b, local;
+ int resume_state;
+ } regmatch_state;
+
+ S_regmatch() {
+ regmatch_state *st = new();
+ int depth=0;
+ st->a++; // do stuff with a and b
+ ...
+ while (scan != NULL) {
+ ...
+ switch (OP(scan)) {
+ case FOO: {
+ st->local = 3;
+ ...
+ st->scan = scan;
+ scan = ...;
+ st->resume_state = resume_FOO;
+ goto start_recurse; // recurse
+
+ resume_point_FOO:
+ if (result)
+ goto yes;
+ }
+ ...
+ }
+ start_recurse:
+ st = new(); push a new state
+ st->a = 1; st->b = 2;
+ depth++;
+ }
+ yes:
+ result = 1;
+ if (depth--) {
+ st = pop();
+ switch (resume_state) {
+ case resume_FOO:
+ goto resume_point_FOO;
+ ...
+ }
+ }
+ return result
+ }
+
+ * WARNING: this means that any line in this function that contains a
+ * REGMATCH() or TRYPAREN() is actually simulating a recursive call to
+ * regmatch() using gotos instead. Thus the values of any local variables
+ * not saved in the regmatch_state structure will have been lost when
+ * execution resumes on the next line .
+ *
+ * States (ie the st pointer) are allocated in slabs of about 4K in size.
+ * PL_regmatch_state always points to the currently active state, and
+ * PL_regmatch_slab points to the slab currently containing PL_regmatch_state.
+ * The first time regmatch is called, the first slab is allocated, and is
+ * never freed until interpreter desctruction. When the slab is full,
+ * a new one is allocated chained to the end. At exit from regmatch, slabs
+ * allocated since entry are freed.
*/
+
+/* *** every FOO_fail should = FOO+1 */
+#define TRIE_next (REGNODE_MAX+1)
+#define TRIE_next_fail (REGNODE_MAX+2)
+#define EVAL_AB (REGNODE_MAX+3)
+#define EVAL_AB_fail (REGNODE_MAX+4)
+#define resume_CURLYX (REGNODE_MAX+5)
+#define resume_WHILEM1 (REGNODE_MAX+6)
+#define resume_WHILEM2 (REGNODE_MAX+7)
+#define resume_WHILEM3 (REGNODE_MAX+8)
+#define resume_WHILEM4 (REGNODE_MAX+9)
+#define resume_WHILEM5 (REGNODE_MAX+10)
+#define resume_WHILEM6 (REGNODE_MAX+11)
+#define BRANCH_next (REGNODE_MAX+12)
+#define BRANCH_next_fail (REGNODE_MAX+13)
+#define CURLYM_A (REGNODE_MAX+14)
+#define CURLYM_A_fail (REGNODE_MAX+15)
+#define CURLYM_B (REGNODE_MAX+16)
+#define CURLYM_B_fail (REGNODE_MAX+17)
+#define IFMATCH_A (REGNODE_MAX+18)
+#define IFMATCH_A_fail (REGNODE_MAX+19)
+#define CURLY_B_min_known (REGNODE_MAX+20)
+#define CURLY_B_min_known_fail (REGNODE_MAX+21)
+#define CURLY_B_min (REGNODE_MAX+22)
+#define CURLY_B_min_fail (REGNODE_MAX+23)
+#define CURLY_B_max (REGNODE_MAX+24)
+#define CURLY_B_max_fail (REGNODE_MAX+25)
+
+#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, "", \
+ state_names[st->resume_state-REGNODE_MAX-1] ) \
+ );
+
+
+#define REG_NODE_NUM(x) ((x) ? (int)((x)-prog) : -1)
+
+#ifdef DEBUGGING
+static const char * const state_names[] = {
+ "TRIE_next",
+ "TRIE_next_fail",
+ "EVAL_AB",
+ "EVAL_AB_fail",
+ "resume_CURLYX",
+ "resume_WHILEM1",
+ "resume_WHILEM2",
+ "resume_WHILEM3",
+ "resume_WHILEM4",
+ "resume_WHILEM5",
+ "resume_WHILEM6",
+ "BRANCH_next",
+ "BRANCH_next_fail",
+ "CURLYM_A",
+ "CURLYM_A_fail",
+ "CURLYM_B",
+ "CURLYM_B_fail",
+ "IFMATCH_A",
+ "IFMATCH_A_fail",
+ "CURLY_B_min_known",
+ "CURLY_B_min_known_fail",
+ "CURLY_B_min",
+ "CURLY_B_min_fail",
+ "CURLY_B_max",
+ "CURLY_B_max_fail"
+};
+
+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,
+ tlen > 19 ? 0 : 19 - tlen,
+ "");
+ }
+}
+
+#endif
+
STATIC I32 /* 0 failure, 1 success */
-S_regmatch(pTHX_ regnode *prog)
+S_regmatch(pTHX_ const regmatch_info *reginfo, regnode *prog)
{
- dVAR;
- 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 = 0; /* accepting states we have traversed */
- reg_trie_accepted *accept_buff = 0; /* "" */
- 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
+ dVAR;
register const 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);
+ const U32 uniflags = UTF8_ALLOW_DEFAULT;
- SV *re_debug_flags = NULL;
-#endif
- U32 uniflags = ckWARN(WARN_UTF8) ? 0 : UTF8_ALLOW_ANY;
+ regexp *rex = reginfo->prog;
+
+ regmatch_slab *orig_slab;
+ regmatch_state *orig_state;
+
+ /* 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; /* initialize to shut up compiler warning */
+ register char *locinput = PL_reginput;
- GET_RE_DEBUG_FLAGS;
+ /* these variables are NOT saved during a recusive RFEGMATCH: */
+ register I32 nextchr; /* is always set to UCHARAT(locinput) */
+ bool result = 0; /* return value of S_regmatch */
+ int depth = 0; /* depth of recursion */
+ 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;
+
+ I32 parenfloor = 0;
#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);
+
+ st->minmod = 0;
+ st->sw = 0;
+ st->logical = 0;
+ cur_curlyx = NULL;
/* Note that nextchr is a byte even in UTF */
nextchr = UCHARAT(locinput);
while (scan != NULL) {
DEBUG_EXECUTE_r( {
- SV *prop = sv_newmortal();
- const int docolor = *PL_colors[0];
- const 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);
- {
- const char * const s0 =
- do_utf8 && OP(scan) != CANY ?
- pv_uni_display(dsv0, (U8*)(locinput - pref_len),
- pref0_len, 60, UNI_DISPLAY_REGEX) :
- locinput - pref_len;
- const int len0 = do_utf8 ? strlen(s0) : pref0_len;
- const char * const 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;
- const int len1 = do_utf8 ? strlen(s1) : pref_len - pref0_len;
- const char * const s2 = do_utf8 && OP(scan) != CANY ?
- pv_uni_display(dsv2, (U8*)locinput,
- PL_regeol - locinput, 60, UNI_DISPLAY_REGEX) :
- locinput;
- const 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_const(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;
break;
sayNO;
case GPOS:
- if (locinput == PL_reg_ganch)
+ if (locinput == reginfo->ganch)
break;
sayNO;
case EOL:
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:
{
+ /* 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 *uscan = (U8*)NULL;
STRLEN bufflen=0;
- accepted = 0;
+ SV *sv_accept_buff = NULL;
+ U8 foldbuf[ UTF8_MAXBYTES_CASE + 1 ];
+
+ ST.accepted = 0; /* how many accepting states we have seen */
+ ST.B = next;
+ ST.jump = trie->jump;
+
+#ifdef DEBUGGING
+ ST.me = scan;
+#endif
+
+
- trie = (reg_trie_data*)PL_regdata->data[ ARG( scan ) ];
+ /*
+ 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;
+ 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: %4"UVxf", Base: %4"UVxf", Accepted: %4"UVxf" ",
- REPORT_CODE_OFF + PL_regindent * 2, "", PL_colors[4],
- (UV)state, (UV)base, (UV)accepted );
- );
+ "%*s %sState: %4"UVxf" Accepted: %4"UVxf" ",
+ 2+depth * 2, "", PL_colors[4],
+ (UV)state, (UV)ST.accepted );
+ });
if ( base ) {
-
- if ( do_utf8 ) {
- if ( foldlen>0 ) {
- uvc = utf8n_to_uvuni( uscan, UTF8_MAXLEN, &len, uniflags );
- foldlen -= len;
- uscan += len;
- len=0;
- } else {
- U8 foldbuf[ UTF8_MAXBYTES_CASE + 1 ];
- 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 = (UV)*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:%4"UVxf" After State: %4"UVxf"%s\n",
+ "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:
- {
- 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: %4"UVxf", Base: %4"UVxf", Accepted: %4"UVxf" ",
- REPORT_CODE_OFF + PL_regindent * 2, "", PL_colors[4],
- (UV)state, (UV)base, (UV)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 ) {
- 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:%4"UVxf" After State: %4"UVxf"%s\n",
- charid, uvc, (UV)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 (ST.accept_buff[cur].wordnum <
+ ST.accept_buff[best].wordnum)
+ best = cur;
}
- if ( !accepted ) {
- sayNO;
+
+ DEBUG_EXECUTE_r({
+ reg_trie_data * const trie
+ = (reg_trie_data*)rex->data->data[ ARG(ST.me) ];
+ SV ** const tmp = RX_DEBUG(reginfo->prog)
+ ? av_fetch( trie->words, ST.accept_buff[ best ].wordnum - 1, 0 )
+ : NULL;
+ 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(scan),
+ PL_colors[5] );
+ });
+
+ if ( best<ST.accepted ) {
+ reg_trie_accepted tmp = ST.accept_buff[ best ];
+ ST.accept_buff[ best ] = ST.accept_buff[ ST.accepted ];
+ ST.accept_buff[ ST.accepted ] = tmp;
+ best = ST.accepted;
}
+ PL_reginput = (char *)ST.accept_buff[ best ].endpos;
+ if ( !ST.jump ) {
+ PUSH_STATE_GOTO(TRIE_next, ST.B);
+ /* NOTREACHED */
+ } else {
+ PUSH_STATE_GOTO(TRIE_next, ST.B - ST.jump[ST.accept_buff[best].wordnum]);
+ /* NOTREACHED */
+ }
+ /* NOTREACHED */
}
+ /* NOTREACHED */
+#undef ST
- /*
- 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_const( *tmp ) : "not compiled under -Dr",
- PL_colors[5] );
- });
- PL_reginput = (char *)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 %"IVdf" possible matches%s\n",
- REPORT_CODE_OFF + PL_regindent * 2, "", PL_colors[4], (IV)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 %"IVdf" (%d) as best, looking at %"IVdf" (%d)%s\n",
- REPORT_CODE_OFF + PL_regindent * 2, "", PL_colors[4],
- (IV)best, accept_buff[ best ].wordnum, (IV)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_const( *tmp ) : "not compiled under -Dr",scan,
- PL_colors[5] );
- });
- if ( best<accepted ) {
- reg_trie_accepted tmp = accept_buff[ best ];
- accept_buff[ best ] = accept_buff[ accepted ];
- accept_buff[ accepted ] = tmp;
- best = accepted;
- }
- PL_reginput = (char *)accept_buff[ best ].endpos;
-
- /*
- as far as I can tell we only need the SAVETMPS/FREETMPS
- for re's with EVAL in them but I'm leaving them in for
- all until I can be sure.
- */
- SAVETMPS;
- gotit = regmatch( scan + NEXT_OFF( scan ) ) ;
- FREETMPS;
- }
- FREETMPS;
- LEAVE;
- }
-
- if ( gotit ) {
- sayYES;
- } else {
- sayNO;
- }
- }
- /* unreached codepoint */
- case EXACT:
- s = STRING(scan);
- ln = STR_LEN(scan);
+ case EXACT: {
+ char *s = STRING(scan);
+ st->ln = STR_LEN(scan);
if (do_utf8 != UTF) {
/* The target and the pattern have differing utf8ness. */
char *l = locinput;
- const char *e = s + ln;
+ const char * const e = s + st->ln;
if (do_utf8) {
/* The target is utf8, the pattern is not utf8. */
/* Inline the first character, for speed. */
if (UCHARAT(s) != nextchr)
sayNO;
- if (PL_regeol - locinput < ln)
+ if (PL_regeol - locinput < st->ln)
sayNO;
- if (ln > 1 && memNE(s, locinput, ln))
+ if (st->ln > 1 && memNE(s, locinput, st->ln))
sayNO;
- locinput += ln;
+ locinput += st->ln;
nextchr = UCHARAT(locinput);
break;
+ }
case EXACTFL:
PL_reg_flags |= RF_tainted;
/* FALL THROUGH */
- case EXACTF:
- s = STRING(scan);
- ln = STR_LEN(scan);
+ case EXACTF: {
+ char * const s = STRING(scan);
+ st->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,
+ if (ibcmp_utf8(s, 0, st->ln, (bool)UTF,
l, &e, 0, do_utf8)) {
/* One more case for the sharp s:
* pack("U0U*", 0xDF) =~ /ss/i,
* byte sequence for the U+00DF. */
if (!(do_utf8 &&
toLOWER(s[0]) == 's' &&
- ln >= 2 &&
+ st->ln >= 2 &&
toLOWER(s[1]) == 's' &&
(U8)l[0] == 0xC3 &&
e - l >= 2 &&
UCHARAT(s) != ((OP(scan) == EXACTF)
? PL_fold : PL_fold_locale)[nextchr])
sayNO;
- if (PL_regeol - locinput < ln)
+ if (PL_regeol - locinput < st->ln)
sayNO;
- if (ln > 1 && (OP(scan) == EXACTF
- ? ibcmp(s, locinput, ln)
- : ibcmp_locale(s, locinput, ln)))
+ if (st->ln > 1 && (OP(scan) == EXACTF
+ ? ibcmp(s, locinput, st->ln)
+ : ibcmp_locale(s, locinput, st->ln)))
sayNO;
- locinput += ln;
+ locinput += st->ln;
nextchr = UCHARAT(locinput);
break;
+ }
case ANYOF:
if (do_utf8) {
STRLEN inclasslen = PL_regeol - locinput;
- if (!reginclass(scan, (U8*)locinput, &inclasslen, do_utf8))
+ if (!reginclass(rex, scan, (U8*)locinput, &inclasslen, do_utf8))
sayNO_ANYOF;
if (locinput >= PL_regeol)
sayNO;
else {
if (nextchr < 0)
nextchr = UCHARAT(locinput);
- if (!REGINCLASS(scan, (U8*)locinput))
+ if (!REGINCLASS(rex, scan, (U8*)locinput))
sayNO_ANYOF;
if (!nextchr && locinput >= PL_regeol)
sayNO;
if (do_utf8) {
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;
if (do_utf8) {
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;
/* was last char in word? */
if (do_utf8) {
if (locinput == PL_bostr)
- ln = '\n';
+ st->ln = '\n';
else {
const U8 * const r = reghop3((U8*)locinput, -1, (U8*)PL_bostr);
- ln = utf8n_to_uvchr(r, UTF8SKIP(r), 0, 0);
+ st->ln = utf8n_to_uvchr(r, UTF8SKIP(r), 0, uniflags);
}
if (OP(scan) == BOUND || OP(scan) == NBOUND) {
- ln = isALNUM_uni(ln);
+ st->ln = isALNUM_uni(st->ln);
LOAD_UTF8_CHARCLASS_ALNUM();
n = swash_fetch(PL_utf8_alnum, (U8*)locinput, do_utf8);
}
else {
- ln = isALNUM_LC_uvchr(UNI_TO_NATIVE(ln));
+ st->ln = isALNUM_LC_uvchr(UNI_TO_NATIVE(st->ln));
n = isALNUM_LC_utf8((U8*)locinput);
}
}
else {
- ln = (locinput != PL_bostr) ?
+ st->ln = (locinput != PL_bostr) ?
UCHARAT(locinput - 1) : '\n';
if (OP(scan) == BOUND || OP(scan) == NBOUND) {
- ln = isALNUM(ln);
+ st->ln = isALNUM(st->ln);
n = isALNUM(nextchr);
}
else {
- ln = isALNUM_LC(ln);
+ st->ln = isALNUM_LC(st->ln);
n = isALNUM_LC(nextchr);
}
}
- if (((!ln) == (!n)) == (OP(scan) == BOUND ||
+ if (((!st->ln) == (!n)) == (OP(scan) == BOUND ||
OP(scan) == BOUNDL))
sayNO;
break;
if (UTF8_IS_CONTINUED(nextchr)) {
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;
if (do_utf8) {
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;
if (do_utf8) {
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;
if (do_utf8) {
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;
PL_reg_flags |= RF_tainted;
/* FALL THROUGH */
case REF:
- case REFF:
+ case REFF: {
+ char *s;
n = ARG(scan); /* which paren pair */
- ln = PL_regstartp[n];
+ st->ln = PL_regstartp[n];
PL_reg_leftiter = PL_reg_maxiter; /* Void cache */
- if ((I32)*PL_reglastparen < n || ln == -1)
+ if ((I32)*PL_reglastparen < n || st->ln == -1)
sayNO; /* Do not match unless seen CLOSEn. */
- if (ln == PL_regendp[n])
+ if (st->ln == PL_regendp[n])
break;
- s = PL_bostr + ln;
+ s = PL_bostr + st->ln;
if (do_utf8 && OP(scan) != REF) { /* REF can do byte comparison */
char *l = locinput;
const char *e = PL_bostr + PL_regendp[n];
(UCHARAT(s) != ((OP(scan) == REFF
? PL_fold : PL_fold_locale)[nextchr]))))
sayNO;
- ln = PL_regendp[n] - ln;
- if (locinput + ln > PL_regeol)
+ st->ln = PL_regendp[n] - st->ln;
+ if (locinput + st->ln > PL_regeol)
sayNO;
- if (ln > 1 && (OP(scan) == REF
- ? memNE(s, locinput, ln)
+ if (st->ln > 1 && (OP(scan) == REF
+ ? memNE(s, locinput, st->ln)
: (OP(scan) == REFF
- ? ibcmp(s, locinput, ln)
- : ibcmp_locale(s, locinput, ln))))
+ ? ibcmp(s, locinput, st->ln)
+ : ibcmp_locale(s, locinput, st->ln))))
sayNO;
- locinput += ln;
+ locinput += st->ln;
nextchr = UCHARAT(locinput);
break;
+ }
case NOTHING:
case TAIL:
break;
case BACK:
break;
- case EVAL:
+
+#undef ST
+#define ST st->u.eval
+
+ case EVAL: /* /(?{A})B/ /(??{A})B/ and /(?(?{A})X|Y)B/ */
{
- 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;
-
{
- 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)
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 (!st->logical) {
+ /* /(?{...})/ */
+ sv_setsv(save_scalar(PL_replgv), ret);
+ break;
+ }
+ }
+ if (st->logical == 2) { /* Postponed subexpression: /(??{...})/ */
+ regexp *re;
+ {
+ /* 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)) {
}
else {
STRLEN len;
- const char *t = SvPV_const(ret, len);
+ const char * const t = SvPV_const(ret, len);
PMOP pm;
- char * const oprecomp = PL_regprecomp;
const I32 osize = PL_regsize;
- const I32 onpar = PL_regnpar;
Zero(&pm, 1, PMOP);
- if (DO_UTF8(ret)) pm.op_pmdynflags |= PMdf_DYN_UTF8;
+ if (DO_UTF8(ret)) pm.op_pmdynflags |= PMdf_DYN_UTF8;
re = CALLREGCOMP(aTHX_ (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);
+
+ /* run the pattern returned from (??{...}) */
+ ST.cp = regcppush(0); /* Save *all* the positions. */
+ REGCP_SET(ST.lastcp);
+ *PL_reglastparen = 0;
+ *PL_reglastcloseparen = 0;
+ PL_reginput = locinput;
+
+ /* XXXX This is too dramatic a measure... */
+ PL_reg_maxiter = 0;
+
+ st->logical = 0;
+ ST.toggle_reg_flags = PL_reg_flags;
+ if (re->reganch & ROPT_UTF8)
+ 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;
+
+ DEBUG_EXECUTE_r(
+ debug_start_match(re, do_utf8, locinput, PL_regeol,
+ "Matching embedded");
+ );
+ /* now continue from first node in postoned RE */
+ PUSH_YES_STATE_GOTO(EVAL_AB, re->program + 1);
+ /* NOTREACHED */
}
+ /* /(?(?{...})X|Y)/ */
+ st->sw = SvTRUE(ret);
+ st->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_FINAL;
+
+
+ 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;
break;
case GROUPP:
n = ARG(scan); /* which paren pair */
- sw = ((I32)*PL_reglastparen >= n && PL_regendp[n] != -1);
+ st->sw = ((I32)*PL_reglastparen >= n && PL_regendp[n] != -1);
break;
case IFTHEN:
PL_reg_leftiter = PL_reg_maxiter; /* Void cache */
- if (sw)
+ if (st->sw)
next = NEXTOPER(NEXTOPER(scan));
else {
next = scan + ARG(scan);
}
break;
case LOGICAL:
- logical = scan->flags;
+ st->logical = scan->flags;
break;
/*******************************************************************
- PL_regcc contains infoblock about the innermost (...)* loop, and
- a pointer to the next outer infoblock.
+ cur_curlyx points to the regmatch_state associated with the most recent CURLYX.
+ This struct contains info about the innermost (...)* loop (an
+ "infoblock"), and a pointer to the next outer cur_curlyx.
Here is how Y(A)*Z is processed (if it is compiled into CURLYX/WHILEM):
- 1) After matching X, regnode for CURLYX is processed;
+ 1) After matching Y, regnode for CURLYX is processed;
- 2) This regnode creates infoblock on the stack, and calls
- regmatch() recursively with the starting point at WHILEM node;
+ 2) This regnode populates cur_curlyx, 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"
+ and "Z" is read from cur_curlyx, 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
+ 5) Each time WHILEM is hit, cur_curlyx is the infoblock created by CURLYX
of the same pair. Thus when WHILEM tries to match Z, it temporarily
- resets PL_regcc, since this Y(A)*Z can be a part of some other loop:
+ resets cur_curlyx, since this Y(A)*Z can be a part of some other loop:
as in (Y(A)*Z)*. If Z matches, the automaton will hit the WHILEM node
of the external loop.
- Currently present infoblocks form a tree with a stem formed by PL_curcc
+ Currently present infoblocks form a tree with a stem formed by cur_curlyx
and whatever it mentions via ->next, and additional attached trees
corresponding to temporarily unset infoblocks as in "5" above.
- In the following picture infoblocks for outer loop of
+ 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.
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;
-
+ parenfloor = scan->flags;
+
+ /* Dave says:
+
+ CURLYX and WHILEM are always paired: they're the moral
+ equivalent of pp_enteriter anbd pp_iter.
+
+ The only time next could be null is if the node tree is
+ corrupt. This was mentioned on p5p a few days ago.
+
+ See http://www.xray.mpe.mpg.de/mailing-lists/perl5-porters/2006-04/msg00556.html
+ So we'll assert that this is true:
+ */
+ assert(next);
if (OP(PREVOPER(next)) == NOTHING) /* LONGJMP */
next += ARG(next);
- 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;
+
+ st->u.curlyx.cp = PL_savestack_ix;
+ st->u.curlyx.outercc = cur_curlyx;
+ cur_curlyx = st;
+ /* these fields contain the state of the current curly.
+ * they are accessed by subsequent WHILEMs;
+ * cur and lastloc are also updated by WHILEM */
+ st->u.curlyx.parenfloor = parenfloor;
+ st->u.curlyx.cur = -1; /* this will be updated by WHILEM */
+ st->u.curlyx.min = ARG1(scan);
+ st->u.curlyx.max = ARG2(scan);
+ st->u.curlyx.scan = NEXTOPER(scan) + EXTRA_STEP_2ARGS;
+ st->u.curlyx.lastloc = 0;
+ /* st->next and st->minmod are also read by WHILEM */
+
PL_reginput = locinput;
- n = regmatch(PREVOPER(next)); /* start on the WHILEM */
- regcpblow(cp);
- PL_regcc = cc.oldcc;
- saySAME(n);
+ REGMATCH(PREVOPER(next), CURLYX); /* start on the WHILEM */
+ /*** all unsaved local vars undefined at this point */
+ regcpblow(st->u.curlyx.cp);
+ cur_curlyx = st->u.curlyx.outercc;
+ saySAME(result);
}
- /* NOT REACHED */
+ /* NOTREACHED */
case WHILEM: {
/*
* This is really hard to understand, because after we match
* that we can try again after backing off.
*/
- CHECKPOINT cp, lastcp;
- CURCUR* cc = PL_regcc;
- char *lastloc = cc->lastloc; /* Detection of 0-len. */
- I32 cache_offset = 0, cache_bit = 0;
+ /* Dave says:
+
+ cur_curlyx gets initialised by CURLYX ready for use by WHILEM.
+ So again, unless somethings been corrupted, cur_curlyx cannot
+ be null at that point in WHILEM.
+
+ See http://www.xray.mpe.mpg.de/mailing-lists/perl5-porters/2006-04/msg00556.html
+ So we'll assert that this is true:
+ */
+ assert(cur_curlyx);
+ st->u.whilem.lastloc = cur_curlyx->u.curlyx.lastloc; /* Detection of 0-len. */
+ st->u.whilem.cache_offset = 0;
+ st->u.whilem.cache_bit = 0;
- n = cc->cur + 1; /* how many we know we matched */
+ n = cur_curlyx->u.curlyx.cur + 1; /* how many we know we matched */
PL_reginput = locinput;
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))
+ REPORT_CODE_OFF+depth*2, "",
+ (long)n, (long)cur_curlyx->u.curlyx.min,
+ (long)cur_curlyx->u.curlyx.max,
+ PTR2UV(cur_curlyx))
);
/* If degenerate scan matches "", assume scan done. */
- if (locinput == cc->lastloc && n >= cc->min) {
- PL_regcc = cc->oldcc;
- if (PL_regcc)
- ln = PL_regcc->cur;
+ if (locinput == cur_curlyx->u.curlyx.lastloc && n >=
+ cur_curlyx->u.curlyx.min)
+ {
+ st->u.whilem.savecc = cur_curlyx;
+ cur_curlyx = cur_curlyx->u.curlyx.outercc;
+ if (cur_curlyx)
+ st->ln = cur_curlyx->u.curlyx.cur;
DEBUG_EXECUTE_r(
PerlIO_printf(Perl_debug_log,
"%*s empty match detected, try continuation...\n",
- REPORT_CODE_OFF+PL_regindent*2, "")
+ REPORT_CODE_OFF+depth*2, "")
);
- if (regmatch(cc->next))
+ REGMATCH(st->u.whilem.savecc->next, WHILEM1);
+ /*** all unsaved local vars undefined at this point */
+ cur_curlyx = st->u.whilem.savecc;
+ if (result)
sayYES;
- if (PL_regcc)
- PL_regcc->cur = ln;
- PL_regcc = cc;
+ if (cur_curlyx->u.curlyx.outercc)
+ cur_curlyx->u.curlyx.outercc->u.curlyx.cur = st->ln;
sayNO;
}
/* First just match a string of min scans. */
- if (n < cc->min) {
- cc->cur = n;
- cc->lastloc = locinput;
- if (regmatch(cc->scan))
+ if (n < cur_curlyx->u.curlyx.min) {
+ cur_curlyx->u.curlyx.cur = n;
+ cur_curlyx->u.curlyx.lastloc = locinput;
+ REGMATCH(cur_curlyx->u.curlyx.scan, WHILEM2);
+ /*** all unsaved local vars undefined at this point */
+ if (result)
sayYES;
- cc->cur = n - 1;
- cc->lastloc = lastloc;
+ cur_curlyx->u.curlyx.cur = n - 1;
+ cur_curlyx->u.curlyx.lastloc = st->u.whilem.lastloc;
sayNO;
}
*that* much linear. */
if (!PL_reg_maxiter) {
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) {
- const I32 size = (PL_reg_maxiter + 7 + POSCACHE_START)/8;
+ 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);
);
}
if (PL_reg_leftiter < 0) {
- cache_offset = locinput - PL_bostr;
+ st->u.whilem.cache_offset = locinput - PL_bostr;
- cache_offset = (scan->flags & 0xf) - 1 + POSCACHE_START
- + cache_offset * (scan->flags>>4);
- cache_bit = cache_offset % 8;
- cache_offset /= 8;
- if (PL_reg_poscache[cache_offset] & (1<<cache_bit)) {
+ st->u.whilem.cache_offset = (scan->flags & 0xf) - 1
+ + st->u.whilem.cache_offset * (scan->flags>>4);
+ st->u.whilem.cache_bit = st->u.whilem.cache_offset % 8;
+ st->u.whilem.cache_offset /= 8;
+ if (PL_reg_poscache[st->u.whilem.cache_offset] & (1<<st->u.whilem.cache_bit)) {
DEBUG_EXECUTE_r(
PerlIO_printf(Perl_debug_log,
"%*s already tried at this position...\n",
- REPORT_CODE_OFF+PL_regindent*2, "")
+ REPORT_CODE_OFF+depth*2, "")
);
- if (PL_reg_poscache[0] & (1<<POSCACHE_SUCCESS))
- /* cache records success */
- sayYES;
- else
- /* cache records failure */
- sayNO_SILENT;
+ sayNO; /* cache records failure */
}
- PL_reg_poscache[cache_offset] |= (1<<cache_bit);
}
}
/* Prefer next over scan for minimal matching. */
- if (cc->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);
- CACHEsayYES; /* All done. */
+ if (cur_curlyx->minmod) {
+ st->u.whilem.savecc = cur_curlyx;
+ cur_curlyx = cur_curlyx->u.curlyx.outercc;
+ if (cur_curlyx)
+ st->ln = cur_curlyx->u.curlyx.cur;
+ st->u.whilem.cp = regcppush(st->u.whilem.savecc->u.curlyx.parenfloor);
+ REGCP_SET(st->u.whilem.lastcp);
+ REGMATCH(st->u.whilem.savecc->next, WHILEM3);
+ /*** all unsaved local vars undefined at this point */
+ cur_curlyx = st->u.whilem.savecc;
+ if (result) {
+ regcpblow(st->u.whilem.cp);
+ sayYES; /* All done. */
}
- REGCP_UNWIND(lastcp);
- regcppop();
- if (PL_regcc)
- PL_regcc->cur = ln;
- PL_regcc = cc;
+ REGCP_UNWIND(st->u.whilem.lastcp);
+ regcppop(rex);
+ if (cur_curlyx->u.curlyx.outercc)
+ cur_curlyx->u.curlyx.outercc->u.curlyx.cur = st->ln;
- if (n >= cc->max) { /* Maximum greed exceeded? */
+ if (n >= cur_curlyx->u.curlyx.max) { /* Maximum greed exceeded? */
if (ckWARN(WARN_REGEXP) && n >= REG_INFTY
&& !(PL_reg_flags & RF_warned)) {
PL_reg_flags |= RF_warned;
DEBUG_EXECUTE_r(
PerlIO_printf(Perl_debug_log,
"%*s trying longer...\n",
- REPORT_CODE_OFF+PL_regindent*2, "")
+ REPORT_CODE_OFF+depth*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);
- CACHEsayYES;
+ cur_curlyx->u.curlyx.cur = n;
+ cur_curlyx->u.curlyx.lastloc = locinput;
+ st->u.whilem.cp = regcppush(cur_curlyx->u.curlyx.parenfloor);
+ REGCP_SET(st->u.whilem.lastcp);
+ REGMATCH(cur_curlyx->u.curlyx.scan, WHILEM4);
+ /*** all unsaved local vars undefined at this point */
+ if (result) {
+ regcpblow(st->u.whilem.cp);
+ sayYES;
}
- REGCP_UNWIND(lastcp);
- regcppop();
- cc->cur = n - 1;
- cc->lastloc = lastloc;
+ REGCP_UNWIND(st->u.whilem.lastcp);
+ regcppop(rex);
+ cur_curlyx->u.curlyx.cur = n - 1;
+ cur_curlyx->u.curlyx.lastloc = st->u.whilem.lastloc;
CACHEsayNO;
}
/* Prefer scan over next 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);
- CACHEsayYES;
+ if (n < cur_curlyx->u.curlyx.max) { /* More greed allowed? */
+ st->u.whilem.cp = regcppush(cur_curlyx->u.curlyx.parenfloor);
+ cur_curlyx->u.curlyx.cur = n;
+ cur_curlyx->u.curlyx.lastloc = locinput;
+ REGCP_SET(st->u.whilem.lastcp);
+ REGMATCH(cur_curlyx->u.curlyx.scan, WHILEM5);
+ /*** all unsaved local vars undefined at this point */
+ if (result) {
+ regcpblow(st->u.whilem.cp);
+ sayYES;
}
- REGCP_UNWIND(lastcp);
- regcppop(); /* Restore some previous $<digit>s? */
+ REGCP_UNWIND(st->u.whilem.lastcp);
+ regcppop(rex); /* Restore some previous $<digit>s? */
PL_reginput = locinput;
DEBUG_EXECUTE_r(
PerlIO_printf(Perl_debug_log,
"%*s failed, try continuation...\n",
- REPORT_CODE_OFF+PL_regindent*2, "")
+ REPORT_CODE_OFF+depth*2, "")
);
}
if (ckWARN(WARN_REGEXP) && n >= REG_INFTY
}
/* 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))
- CACHEsayYES;
- if (PL_regcc)
- PL_regcc->cur = ln;
- PL_regcc = cc;
- cc->cur = n - 1;
- cc->lastloc = lastloc;
+ st->u.whilem.savecc = cur_curlyx;
+ cur_curlyx = cur_curlyx->u.curlyx.outercc;
+ if (cur_curlyx)
+ st->ln = cur_curlyx->u.curlyx.cur;
+ REGMATCH(st->u.whilem.savecc->next, WHILEM6);
+ /*** all unsaved local vars undefined at this point */
+ cur_curlyx = st->u.whilem.savecc;
+ if (result)
+ sayYES;
+ if (cur_curlyx->u.curlyx.outercc)
+ cur_curlyx->u.curlyx.outercc->u.curlyx.cur = st->ln;
+ cur_curlyx->u.curlyx.cur = n - 1;
+ cur_curlyx->u.curlyx.lastloc = st->u.whilem.lastloc;
CACHEsayNO;
}
- /* NOT REACHED */
- case BRANCHJ:
+ /* 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 {
- const 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;
+ 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;
+ st->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. */
+ }
+ ST.A = scan;
+ ST.B = next;
+ ST.alen = 0;
+ ST.count = 0;
+ ST.minmod = st->minmod;
+ st->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;
- 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;
+ PUSH_YES_STATE_GOTO(CURLYM_A, ST.A); /* match A */
+ /* NOTREACHED */
- if (! HAS_TEXT(text_node)) FIND_NEXT_IMPT(text_node);
+ case CURLYM_A: /* we've just matched an A */
+ locinput = st->locinput;
+ nextchr = UCHARAT(locinput);
- 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.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 (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*/ )
+ 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;
}
+ else
+ PL_regendp[paren] = -1;
}
- 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; \
+ } \
+ 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 */
+ 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) {
to_utf8_lower((U8*)s, tmpbuf1, &ulen1);
to_utf8_upper((U8*)s, tmpbuf2, &ulen2);
-
- c1 = utf8n_to_uvuni(tmpbuf1, UTF8_MAXBYTES, 0,
- uniflags);
- c2 = utf8n_to_uvuni(tmpbuf2, UTF8_MAXBYTES, 0,
- uniflags);
+#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,
+ 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->minmod) {
+ st->minmod = 0;
+ 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 {
- if (c1 == c2) {
- STRLEN len;
- /* count initialised to
- * utf8_distance(old, locinput) */
- while (locinput <= e &&
- utf8n_to_uvchr((U8*)locinput,
- UTF8_MAXBYTES, &len,
- uniflags) != (UV)c1) {
- locinput += len;
- count++;
- }
- } else {
- STRLEN len;
- /* count initialised to
- * utf8_distance(old, locinput) */
- while (locinput <= e) {
- UV c = utf8n_to_uvchr((U8*)locinput,
- UTF8_MAXBYTES, &len,
- uniflags);
- 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,
- uniflags);
- 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,
- uniflags);
- 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,
- uniflags);
- 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);
+ 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);
+ 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);
+ 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:
+ 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 ...\n",
+ REPORT_CODE_OFF+depth*2, ""););
+ 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. */
}
PL_reginput = locinput; /* put where regtry can find it */
sayYES_FINAL; /* Success! */
- case SUCCEED:
+
+ 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_FINAL; /* 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;
- goto do_ifmatch;
- case IFMATCH:
- n = 1;
+
+ case UNLESSM: /* -ve lookaround: (?!A), or with flags, (?<!A) */
+ ST.wanted = 0;
+ goto ifmatch_trivial_fail_test;
+
+ case IFMATCH: /* +ve lookaround: (?=A), or with flags, (?<=A) */
+ ST.wanted = 1;
+ ifmatch_trivial_fail_test:
if (scan->flags) {
- s = HOPBACKc(locinput, scan->flags);
- if (!s)
- goto say_no;
+ char * const s = HOPBACKc(locinput, scan->flags);
+ if (!s) {
+ /* trivial fail */
+ if (st->logical) {
+ st->logical = 0;
+ st->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) {
- say_no:
- if (logical) {
- logical = 0;
- sw = 0;
- goto do_longjump;
- }
- else
- sayNO;
- }
- say_yes:
- if (logical) {
- logical = 0;
- sw = 1;
+ ST.me = scan;
+ /* 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) {
+ st->logical = 0;
+ st->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;
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;
+ /* XXX probably don't need to initialise these */
+ newst->minmod = 0;
+ newst->sw = 0;
+ newst->logical = 0;
+
+
+ locinput = PL_reginput;
+ nextchr = UCHARAT(locinput);
+ st = newst;
+ continue;
+ /* NOTREACHED */
+ }
+
+ /* simulate recursively calling regmatch(), but without actually
+ * recursing - ie save the current state on the heap rather than on
+ * the stack, then re-enter the loop. This avoids complex regexes
+ * blowing the processor stack */
+
+ start_recurse:
+ {
+ /* push new state */
+ regmatch_state *oldst = st;
+
+ DEBUG_STATE_pp("push");
+ depth++;
+
+ /* grab the next free state slot */
+ st++;
+ if (st > SLAB_LAST(PL_regmatch_slab))
+ st = S_push_slab(aTHX);
+ PL_regmatch_state = st;
+
+ oldst->next = next;
+ oldst->n = n;
+ oldst->locinput = locinput;
+
+ locinput = PL_reginput;
+ nextchr = UCHARAT(locinput);
+ st->minmod = 0;
+ st->sw = 0;
+ st->logical = 0;
+
+ }
}
+
+
/*
* We get here only if there's trouble -- normally "case END" is
* the terminating point.
/*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:
+
+ 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
+ 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;
+
+ switch (st->resume_state) {
+ case EVAL_AB:
+ case IFMATCH_A:
+ case CURLYM_A:
+ state_num = st->resume_state;
+ goto reenter_switch;
+
+ case CURLYM_B:
+ case BRANCH_next:
+ case TRIE_next:
+ case CURLY_B_max:
+ default:
+ Perl_croak(aTHX_ "unexpected yes resume state");
+ }
+ }
+
DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "%sMatch successful!%s\n",
PL_colors[4], PL_colors[5]));
yes:
-#ifdef DEBUGGING
- PL_regindent--;
-#endif
-#if 0 /* Breaks $^R */
- if (unwind)
- regcpblow(firstcp);
-#endif
- return 1;
+ result = 1;
+ /* XXX this is duplicate(ish) code to that in the do_no section.
+ * will disappear when REGFMATCH goes */
+ if (depth) {
+ /* restore previous state and re-enter */
+ st--;
+ if (st < SLAB_FIRST(PL_regmatch_slab)) {
+ PL_regmatch_slab = PL_regmatch_slab->prev;
+ st = SLAB_LAST(PL_regmatch_slab);
+ }
+ PL_regmatch_state = st;
+ scan = st->scan;
+ next = st->next;
+ n = st->n;
+ locinput= st->locinput;
+ nextchr = UCHARAT(locinput);
+
+ DEBUG_STATE_pp("pop");
+ depth--;
+
+ switch (st->resume_state) {
+ case resume_CURLYX:
+ goto resume_point_CURLYX;
+ case resume_WHILEM1:
+ goto resume_point_WHILEM1;
+ case resume_WHILEM2:
+ goto resume_point_WHILEM2;
+ case resume_WHILEM3:
+ goto resume_point_WHILEM3;
+ case resume_WHILEM4:
+ goto resume_point_WHILEM4;
+ case resume_WHILEM5:
+ goto resume_point_WHILEM5;
+ case resume_WHILEM6:
+ goto resume_point_WHILEM6;
+
+ case TRIE_next:
+ case CURLYM_A:
+ case CURLYM_B:
+ case EVAL_AB:
+ case IFMATCH_A:
+ case BRANCH_next:
+ case CURLY_B_max:
+ case CURLY_B_min:
+ case CURLY_B_min_known:
+ break;
+
+ default:
+ Perl_croak(aTHX_ "regexp resume memory corruption");
+ }
+ }
+ 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);
- const 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. */
- 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
+ result = 0;
- goto reenter;
+ 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 */
+ PL_regmatch_state = st;
+ scan = st->scan;
+ next = st->next;
+ n = st->n;
+ locinput= st->locinput;
+ nextchr = UCHARAT(locinput);
+
+ DEBUG_STATE_pp("pop");
+ depth--;
+
+ switch (st->resume_state) {
+ case resume_CURLYX:
+ goto resume_point_CURLYX;
+ case resume_WHILEM1:
+ goto resume_point_WHILEM1;
+ case resume_WHILEM2:
+ goto resume_point_WHILEM2;
+ case resume_WHILEM3:
+ goto resume_point_WHILEM3;
+ case resume_WHILEM4:
+ goto resume_point_WHILEM4;
+ case resume_WHILEM5:
+ goto resume_point_WHILEM5;
+ case resume_WHILEM6:
+ goto resume_point_WHILEM6;
+
+ case TRIE_next:
+ case EVAL_AB:
+ case BRANCH_next:
+ case CURLYM_A:
+ case CURLYM_B:
+ case IFMATCH_A:
+ case CURLY_B_max:
+ case CURLY_B_min:
+ case CURLY_B_min_known:
+ if (yes_state == st)
+ yes_state = st->u.yes.prev_yes_state;
+ state_num = st->resume_state + 1; /* failure = success + 1 */
+ goto reenter_switch;
+
default:
- Perl_croak(aTHX_ "regexp unwind memory corruption");
+ Perl_croak(aTHX_ "regexp resume memory corruption");
}
- /* NOT REACHED */
}
-#ifdef DEBUGGING
- PL_regindent--;
-#endif
- return 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);
+ }
+ }
+
+ return result;
+
}
/*
* rather than incrementing count on every character. [Er, except utf8.]]
*/
STATIC I32
-S_regrepeat(pTHX_ const regnode *p, I32 max)
+S_regrepeat(pTHX_ const regexp *prog, const regnode *p, I32 max)
{
dVAR;
register char *scan;
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) {
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;
PL_reginput = scan;
DEBUG_r({
- SV *re_debug_flags = NULL;
- 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_const(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 const 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) {
+ if (data && data->count) {
const U32 n = ARG(node);
- if (PL_regdata->what[n] == 's') {
- SV * const rv = (SV*)PL_regdata->data[n];
+ 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)
return sw;
}
+#endif
/*
- reginclass - determine if a character falls into a character class
*/
STATIC bool
-S_reginclass(pTHX_ register const regnode *n, register const 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)
{
dVAR;
const char flags = ANYOF_FLAGS(n);
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)) {
match = TRUE;
if (!match) {
AV *av;
- SV * const 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))
}
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 */
}
}
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;
}
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;
}
+
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;
}
static void
restore_pos(pTHX_ void *arg)
{
- PERL_UNUSED_ARG(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;
+ rex->subbeg = PL_reg_oldsaved;
+ rex->sublen = PL_reg_oldsavedlen;
#ifdef PERL_OLD_COPY_ON_WRITE
- PL_reg_re->saved_copy = PL_nrs;
+ 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;
S_to_utf8_substr(pTHX_ register regexp *prog)
{
if (prog->float_substr && !prog->float_utf8) {
- SV* sv;
- 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);
prog->check_utf8 = sv;
}
if (prog->anchored_substr && !prog->anchored_utf8) {
- SV* sv;
- 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);
STATIC void
S_to_byte_substr(pTHX_ register regexp *prog)
{
+ dVAR;
if (prog->float_utf8 && !prog->float_substr) {
- SV* sv;
- 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);
prog->check_substr = sv;
}
if (prog->anchored_utf8 && !prog->anchored_substr) {
- SV* sv;
- 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);
https://perl5.git.perl.org / perl5.git / blobdiff