**** 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 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) \
+#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*)(off < 0 ? PL_regeol : PL_bostr))) \
: (pos - off >= PL_bostr) \
? (U8*)(pos - off) \
- : (U8*)NULL \
+ : (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 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
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;
STATIC char *
S_regcppop(pTHX)
{
+ dVAR;
I32 i;
U32 paren = 0;
char *input;
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;
#define regcpblow(cp) LEAVE_SCOPE(cp) /* Ignores regcppush()ed data. */
-#define TRYPAREN(paren, n, input) { \
+#define TRYPAREN(paren, n, input, where) { \
if (paren) { \
if (n) { \
PL_regstartp[paren] = HOPc(input, -1) - PL_bostr; \
else \
PL_regendp[paren] = -1; \
} \
- if (regmatch(next)) \
+ REGMATCH(next, where); \
+ if (result) \
sayYES; \
if (paren && n) \
PL_regendp[paren] = -1; \
STATIC void
S_cache_re(pTHX_ regexp *prog)
{
+ dVAR;
PL_regprecomp = prog->precomp; /* Needed for FAIL. */
#ifdef DEBUGGING
PL_regprogram = prog->program;
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;
char *t;
const int do_utf8 = sv ? SvUTF8(sv) : 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
- char *i_strpos = strpos;
- SV *dsv = PERL_DEBUG_PAD_ZERO(0);
+ const char * const i_strpos = strpos;
+ SV * const dsv = PERL_DEBUG_PAD_ZERO(0);
#endif
GET_RE_DEBUG_FLAGS_DECL;
|| ((slen = SvCUR(check)) > 1
&& memNE(SvPVX_const(check), s, slen)))
goto report_neq;
+ check_at = s;
goto success_at_start;
}
}
if (check == (do_utf8 ? prog->float_utf8 : prog->float_substr)) {
do_other_anchored:
{
- char *last = HOP3c(s, -start_shift, strbeg), *last1, *last2;
+ char * const last = HOP3c(s, -start_shift, strbeg);
+ char *last1, *last2;
char *s1 = s;
SV* must;
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
regstclass does not come from lookahead... */
/* If regstclass takes bytelength more than 1: If charlength==1, OK.
This leaves EXACTF only, which is dealt with in find_byclass(). */
- const U8* str = (U8*)STRING(prog->regstclass);
+ const U8* const str = (U8*)STRING(prog->regstclass);
const int cl_l = (PL_regkind[(U8)OP(prog->regstclass)] == EXACT
? CHR_DIST(str+STR_LEN(prog->regstclass), str)
: 1);
s = find_byclass(prog, prog->regstclass, s, endpos, 1);
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;
}
/* We know what class REx starts with. Try to find this position... */
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;
to_utf8_lower((U8*)m, tmpbuf1, &ulen1);
to_utf8_upper((U8*)m, tmpbuf2, &ulen2);
c1 = utf8n_to_uvchr(tmpbuf1, UTF8_MAXBYTES_CASE,
- 0, ckWARN(WARN_UTF8) ? 0 : UTF8_ALLOW_ANY);
+ 0, uniflags);
c2 = utf8n_to_uvchr(tmpbuf2, UTF8_MAXBYTES_CASE,
- 0, ckWARN(WARN_UTF8) ? 0 : UTF8_ALLOW_ANY);
+ 0, uniflags);
lnc = 0;
while (sm < ((U8 *) m + ln)) {
lnc++;
UV c, f;
U8 tmpbuf [UTF8_MAXBYTES+1];
STRLEN len, foldlen;
-
+ const U32 uniflags = ckWARN(WARN_UTF8) ? 0 : UTF8_ALLOW_ANY;
if (c1 == c2) {
/* Upper and lower of 1st char are equal -
* probably not a "letter". */
while (s <= e) {
c = utf8n_to_uvchr((U8*)s, UTF8_MAXBYTES, &len,
- ckWARN(WARN_UTF8) ?
- 0 : UTF8_ALLOW_ANY);
+ uniflags);
if ( c == c1
&& (ln == len ||
ibcmp_utf8(s, (char **)0, 0, do_utf8,
else {
while (s <= e) {
c = utf8n_to_uvchr((U8*)s, UTF8_MAXBYTES, &len,
- ckWARN(WARN_UTF8) ?
- 0 : UTF8_ALLOW_ANY);
+ uniflags);
/* Handle some of the three Greek sigmas cases.
* Note that not all the possible combinations
if (s == PL_bostr)
tmp = '\n';
else {
- U8 *r = reghop3((U8*)s, -1, (U8*)PL_bostr);
-
+ U8 * const r = reghop3((U8*)s, -1, (U8*)PL_bostr);
tmp = utf8n_to_uvchr(r, UTF8SKIP(r), 0, 0);
}
tmp = ((OP(c) == BOUND ?
if (s == PL_bostr)
tmp = '\n';
else {
- U8 *r = reghop3((U8*)s, -1, (U8*)PL_bostr);
-
+ U8 * const r = reghop3((U8*)s, -1, (U8*)PL_bostr);
tmp = utf8n_to_uvchr(r, UTF8SKIP(r), 0, 0);
}
tmp = ((OP(c) == NBOUND ?
/* 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;
+ char *scream_olds = NULL;
SV* 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);
+ SV * const dsv0 = PERL_DEBUG_PAD_ZERO(0);
+ SV * const dsv1 = PERL_DEBUG_PAD_ZERO(1);
#endif
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;
PL_reg_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;
"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;
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) {
}
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 {
STATIC I32 /* 0 failure, 1 success */
S_regtry(pTHX_ regexp *prog, char *startpos)
{
+ dVAR;
register I32 i;
register I32 *sp;
register I32 *ep;
if (!(SvTYPE(PL_reg_sv) >= SVt_PVMG && SvMAGIC(PL_reg_sv)
&& (mg = mg_find(PL_reg_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(PL_reg_sv, (SV*)0, PERL_MAGIC_regex_global,
+ &PL_vtbl_mglob, NULL, 0);
mg->mg_len = -1;
}
PL_reg_magic = mg;
SAVEDESTRUCTOR_X(restore_pos, 0);
}
if (!PL_reg_curpm) {
- Newz(22, PL_reg_curpm, 1, PMOP);
+ Newxz(PL_reg_curpm, 1, PMOP);
#ifdef USE_ITHREADS
{
SV* repointer = newSViv(0);
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 */
}
if(PL_reg_start_tmp)
Renew(PL_reg_start_tmp, PL_reg_start_tmpl, char*);
else
- New(22, PL_reg_start_tmp, PL_reg_start_tmpl, char*);
+ Newx(PL_reg_start_tmp, PL_reg_start_tmpl, char*);
}
/* XXXX What this code is doing here?!!! There should be no need
/* 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
+
+/* simulate a recursive call to regmatch */
+
+#define REGMATCH(ns, where) \
+ new_scan = (ns); \
+ resume_state = resume_##where; \
+ goto start_recurse; \
+ resume_point_##where:
+
+typedef enum {
+ resume_TRIE1,
+ resume_TRIE2,
+ resume_CURLYX,
+ resume_WHILEM1,
+ resume_WHILEM2,
+ resume_WHILEM3,
+ resume_WHILEM4,
+ resume_WHILEM5,
+ resume_WHILEM6,
+ resume_CURLYM1,
+ resume_CURLYM2,
+ resume_CURLYM3,
+ resume_CURLYM4,
+ resume_IFMATCH,
+ resume_PLUS1,
+ resume_PLUS2,
+ resume_PLUS3,
+ resume_PLUS4,
+ resume_END
+} resume_states;
+
+
+struct regmatch_state {
+ struct regmatch_state *prev_state;
+ resume_states resume_state;
+ regnode *scan;
+ regnode *next;
+ bool minmod;
+ bool sw;
+ int logical;
+ I32 unwind;
+ CURCUR *cc;
+ char *locinput;
+
+ I32 n;
+ I32 ln;
+ I32 c1, c2, paren;
+ CHECKPOINT cp, lastcp;
+ CURCUR *oldcc;
+ char *lastloc;
+ I32 cache_offset, cache_bit;
+ I32 curlym_l;
+ I32 matches;
+ I32 maxwanted;
+ char *e;
+ char *old;
+ int count;
+ re_cc_state *cur_call_cc;
+ regexp *end_re;
+ reg_trie_accepted *accept_buff;
+ U32 accepted;
+
+ re_cc_state *reg_call_cc; /* saved value of PL_reg_call_cc */
+};
/*
- 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) {
+ ...
+ switch (OP(scan)) {
+ case FOO: {
+ int local = 3;
+ ...
+ if (regmatch(...)) // recurse
+ goto yes;
+ }
+ ...
+ }
+ }
+ yes:
+ return 1;
+ }
+
+ * Now it looks something like this:
+
+ struct regmatch_state {
+ int a, b, local;
+ int resume_state;
+ };
+
+ S_regmatch() {
+ int a = 1, b = 2;
+ int local;
+ ...
+ while (scan != NULL) {
+ ...
+ switch (OP(scan)) {
+ case FOO: {
+ local = 3;
+ ...
+ resume_state = resume_FOO;
+ new_scan = ...;
+ goto start_recurse;
+ resume_point_FOO:
+
+ if (result) // recurse
+ goto yes;
+ }
+ ...
+ }
+ start_recurse:
+ ...push a, b, local onto the heap
+ a = 1; b = 2;
+ scan = new_scan;
+ }
+ yes:
+ result = 1;
+ if (states pushed on heap) {
+ ... restore a, b, local from heap
+ 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 .
*/
+
+
STATIC I32 /* 0 failure, 1 success */
S_regmatch(pTHX_ regnode *prog)
{
dVAR;
+ register const bool do_utf8 = PL_reg_match_utf8;
+ const U32 uniflags = ckWARN(WARN_UTF8) ? 0 : UTF8_ALLOW_ANY;
+
+ /************************************************************
+ * the following variabes are saved and restored on each fake
+ * recursive call to regmatch:
+ *
+ * The first ones contain state that needs to be maintained
+ * across the main while loop: */
+
+ struct regmatch_state *prev_state = NULL; /* stack of pushed states */
+ resume_states resume_state; /* where to jump to on return */
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 */
+ bool minmod = 0; /* the next {n.m} is a {n,m}? */
+ bool sw = 0; /* the condition value in (?(cond)a|b) */
+ int logical = 0;
+ I32 unwind = 0; /* savestack index of current unwind block */
+ CURCUR *cc = NULL; /* current innermost curly struct */
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;
-#endif
- const register bool do_utf8 = PL_reg_match_utf8;
-#ifdef DEBUGGING
- SV *dsv0 = PERL_DEBUG_PAD_ZERO(0);
- SV *dsv1 = PERL_DEBUG_PAD_ZERO(1);
- SV *dsv2 = PERL_DEBUG_PAD_ZERO(2);
- SV *re_debug_flags = NULL;
-#endif
-
- GET_RE_DEBUG_FLAGS;
+ /* while the rest of these are local to an individual branch, and
+ * have only been hoisted into this outer scope to allow for saving and
+ * restoration - thus they can be safely reused in other branches.
+ * Note that they are only initialized here to silence compiler
+ * warnings :-( */
+ register I32 n = 0; /* no or next */
+ register I32 ln = 0; /* len or last */
+ register I32 c1 = 0, c2 = 0, paren = 0; /* case fold search, parenth */
+ CHECKPOINT cp = 0; /* remember current savestack indexes */
+ CHECKPOINT lastcp = 0;
+ CURCUR *oldcc = NULL; /* tmp copy of cc */
+ char *lastloc = NULL; /* Detection of 0-len. */
+ I32 cache_offset = 0;
+ I32 cache_bit = 0;
+ I32 curlym_l = 0;
+ I32 matches = 0;
+ I32 maxwanted = 0;
+ char *e = NULL;
+ char *old = NULL;
+ int count = 0;
+ re_cc_state *cur_call_cc = NULL;
+ regexp *end_re = NULL;
+ reg_trie_accepted *accept_buff = NULL;
+ U32 accepted = 0; /* how many accepting states we have seen */
+
+ /************************************************************
+ * these variables are NOT saved: */
+
+ register I32 nextchr; /* is always set to UCHARAT(locinput) */
+ regnode *new_scan; /* node to begin exedcution at when recursing */
+ bool result; /* return value of S_regmatch */
+
+ regnode *inner; /* Next node in internal branch. */
+
#ifdef DEBUGGING
+ SV *re_debug_flags = NULL;
+ GET_RE_DEBUG_FLAGS;
PL_regindent++;
#endif
-
/* Note that nextchr is a byte even in UTF */
nextchr = UCHARAT(locinput);
scan = prog;
while (scan != NULL) {
DEBUG_EXECUTE_r( {
- SV *prop = sv_newmortal();
+ SV * const 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);
{
const char * const s0 =
do_utf8 && OP(scan) != CANY ?
- pv_uni_display(dsv0, (U8*)(locinput - pref_len),
+ pv_uni_display(PERL_DEBUG_PAD(0), (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),
+ pv_uni_display(PERL_DEBUG_PAD(1),
+ (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,
+ pv_uni_display(PERL_DEBUG_PAD(2), (U8*)locinput,
PL_regeol - locinput, 60, UNI_DISPLAY_REGEX) :
locinput;
const int len2 = do_utf8 ? strlen(s2) : l;
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 TRIE:
case TRIEF:
case TRIEFL:
{
-
- const U32 uniflags = ckWARN( WARN_UTF8 ) ? 0 : UTF8_ALLOW_ANY;
U8 *uc = ( U8* )locinput;
U32 state = 1;
U16 charid = 0;
STRLEN foldlen = 0;
U8 *uscan = (U8*)NULL;
STRLEN bufflen=0;
- accepted = 0;
-
- trie = (reg_trie_data*)PL_regdata->data[ ARG( scan ) ];
+ SV *sv_accept_buff = NULL;
+ const enum { trie_plain, trie_utf8, trie_uft8_fold }
+ trie_type = do_utf8 ?
+ (OP(scan) == TRIE ? trie_utf8 : trie_uft8_fold)
+ : trie_plain;
+
+ /* what trie are we using right now */
+ reg_trie_data *trie
+ = (reg_trie_data*)PL_regdata->data[ ARG( scan ) ];
+ accepted = 0; /* how many accepting states we have seen */
+ result = 0;
while ( state && uc <= (U8*)PL_regeol ) {
- TRIE_CHECK_STATE_IS_ACCEPTING;
+ if (trie->states[ state ].wordnum) {
+ if (!accepted ) {
+ ENTER;
+ SAVETMPS;
+ bufflen = TRIE_INITAL_ACCEPT_BUFFLEN;
+ sv_accept_buff=newSV(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;
+ }
base = trie->states[ state ].trans.base;
);
if ( base ) {
-
- if ( do_utf8 || UTF ) {
+ switch (trie_type) {
+ case trie_uft8_fold:
if ( foldlen>0 ) {
uvc = utf8n_to_uvuni( uscan, UTF8_MAXLEN, &len, uniflags );
foldlen -= len;
foldlen -= UNISKIP( uvc );
uscan = foldbuf + UNISKIP( uvc );
}
- } else {
+ break;
+ case trie_utf8:
+ uvc = utf8n_to_uvuni( (U8*)uc, UTF8_MAXLEN,
+ &len, uniflags );
+ break;
+ case trie_plain:
uvc = (UV)*uc;
len = 1;
}
- TRIE_HANDLE_CHAR;
+ 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;
- } else {
+ }
+ else {
state = 0;
}
DEBUG_TRIE_EXECUTE_r(
charid, uvc, (UV)state, PL_colors[5] );
);
}
- if ( !accepted ) {
- sayNO;
- } else {
- goto TrieAccept;
- }
- }
- /* unreached codepoint: we jump into the middle of the next case
- from previous if blocks */
- case TRIE:
- {
- const U32 uniflags = ckWARN( WARN_UTF8 ) ? 0 : UTF8_ALLOW_ANY;
- U8 *uc = (U8*)locinput;
- U32 state = 1;
- U16 charid = 0;
- U32 base = 0;
- UV uvc = 0;
- STRLEN len = 0;
- STRLEN bufflen = 0;
- accepted = 0;
-
- trie = (reg_trie_data*)PL_regdata->data[ ARG( scan ) ];
-
- while ( state && uc <= (U8*)PL_regeol ) {
-
- TRIE_CHECK_STATE_IS_ACCEPTING;
-
- base = trie->states[ state ].trans.base;
-
- DEBUG_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 );
- );
-
- if ( base ) {
-
- if ( do_utf8 || UTF ) {
- uvc = utf8n_to_uvuni( (U8*)uc, UTF8_MAXLEN, &len, uniflags );
- } else {
- uvc = (U32)*uc;
- len = 1;
- }
-
- TRIE_HANDLE_CHAR;
-
- } else {
- state = 0;
- }
- 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] );
- );
- }
- if ( !accepted ) {
+ if (!accepted )
sayNO;
- }
- }
-
/*
There was at least one accepting state that we
eventually terminates once all of the accepting states
have been tried.
*/
- TrieAccept:
- {
- int gotit = 0;
if ( accepted == 1 ) {
DEBUG_EXECUTE_r({
as we wont be using accept_buff again. */
FREETMPS;
LEAVE;
- gotit = regmatch( scan + NEXT_OFF( scan ) );
+ REGMATCH(scan + NEXT_OFF(scan), TRIE1);
+ /*** all unsaved local vars undefined at this point */
} 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-- ) {
+ while ( !result && accepted-- ) {
U32 best = 0;
U32 cur;
for( cur = 1 ; cur <= accepted ; cur++ ) {
best = cur;
}
DEBUG_EXECUTE_r({
- SV **tmp = av_fetch( trie->words, accept_buff[ best ].wordnum - 1, 0 );
+ SV ** const 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,
all until I can be sure.
*/
SAVETMPS;
- gotit = regmatch( scan + NEXT_OFF( scan ) ) ;
+ REGMATCH(scan + NEXT_OFF(scan), TRIE2);
+ /*** all unsaved local vars undefined at this point */
FREETMPS;
}
FREETMPS;
LEAVE;
}
- if ( gotit ) {
+ if (result) {
sayYES;
} else {
sayNO;
}
}
/* unreached codepoint */
- case EXACT:
- s = STRING(scan);
+ case EXACT: {
+ char *s = STRING(scan);
ln = STR_LEN(scan);
if (do_utf8 != UTF) {
/* The target and the pattern have differing utf8ness. */
sayNO;
if (NATIVE_TO_UNI(*(U8*)s) !=
utf8n_to_uvuni((U8*)l, UTF8_MAXBYTES, &ulen,
- ckWARN(WARN_UTF8) ?
- 0 : UTF8_ALLOW_ANY))
+ uniflags))
sayNO;
l += ulen;
s ++;
sayNO;
if (NATIVE_TO_UNI(*((U8*)l)) !=
utf8n_to_uvuni((U8*)s, UTF8_MAXBYTES, &ulen,
- ckWARN(WARN_UTF8) ?
- 0 : UTF8_ALLOW_ANY))
+ uniflags))
sayNO;
s += ulen;
l ++;
locinput += ln;
nextchr = UCHARAT(locinput);
break;
+ }
case EXACTFL:
PL_reg_flags |= RF_tainted;
/* FALL THROUGH */
- case EXACTF:
- s = STRING(scan);
+ case EXACTF: {
+ char *s = STRING(scan);
ln = STR_LEN(scan);
if (do_utf8 || UTF) {
locinput += ln;
nextchr = UCHARAT(locinput);
break;
+ }
case ANYOF:
if (do_utf8) {
STRLEN inclasslen = PL_regeol - locinput;
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];
PL_reg_leftiter = PL_reg_maxiter; /* Void cache */
locinput += ln;
nextchr = UCHARAT(locinput);
break;
+ }
case NOTHING:
case TAIL:
case EVAL:
{
dSP;
- OP_4tree *oop = PL_op;
- COP *ocurcop = PL_curcop;
+ OP_4tree * const oop = PL_op;
+ COP * const ocurcop = PL_curcop;
PAD *old_comppad;
SV *ret;
- struct regexp *oreg = PL_reg_re;
+ struct regexp * const oreg = PL_reg_re;
n = ARG(scan);
PL_op = (OP_4tree*)PL_regdata->data[n];
PL_regendp[0] = PL_reg_magic->mg_len = locinput - PL_bostr;
{
- SV **before = SP;
+ SV ** const before = SP;
CALLRUNOPS(aTHX); /* Scalar context. */
SPAGAIN;
if (SP == before)
if (logical) {
if (logical == 2) { /* Postponed subexpression. */
regexp *re;
- MAGIC *mg = Null(MAGIC*);
+ MAGIC *mg = NULL;
re_cc_state state;
- CHECKPOINT cp, lastcp;
int toggleutf;
register SV *sv;
}
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;
);
state.node = next;
state.prev = PL_reg_call_cc;
- state.cc = PL_regcc;
+ state.cc = cc;
state.re = PL_reg_re;
- PL_regcc = 0;
+ cc = 0;
cp = regcppush(0); /* Save *all* the positions. */
REGCP_SET(lastcp);
/* XXXX This is too dramatic a measure... */
PL_reg_maxiter = 0;
+ /* XXX the only recursion left in regmatch() */
if (regmatch(re->program + 1)) {
/* Even though we succeeded, we need to restore
global variables, since we may be wrapped inside
/* XXXX Do this only if SUSPENDed? */
PL_reg_call_cc = state.prev;
- PL_regcc = state.cc;
+ cc = state.cc;
PL_reg_re = state.re;
cache_re(PL_reg_re);
if (toggleutf) PL_reg_flags ^= RF_utf8;
REGCP_UNWIND(lastcp);
regcppop();
PL_reg_call_cc = state.prev;
- PL_regcc = state.cc;
+ cc = state.cc;
PL_reg_re = state.re;
cache_re(PL_reg_re);
if (toggleutf) PL_reg_flags ^= RF_utf8;
logical = scan->flags;
break;
/*******************************************************************
- PL_regcc contains infoblock about the innermost (...)* loop, and
+ cc contains infoblock about the innermost (...)* loop, and
a pointer to the next outer infoblock.
Here is how Y(A)*Z is processed (if it is compiled into CURLYX/WHILEM):
- 1) After matching X, regnode for CURLYX is processed;
+ 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 mallocs an infoblock, 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
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, cc 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 cc, 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.
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;
+ {
+ CURCUR *newcc;
+ Newx(newcc, 1, CURCUR);
+ oldcc = cc;
+ newcc->oldcc = cc;
+ cc = newcc;
+ }
+ cp = PL_savestack_ix;
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;
+ cc->parenfloor = parenfloor;
+ cc->cur = -1;
+ cc->min = ARG1(scan);
+ cc->max = ARG2(scan);
+ cc->scan = NEXTOPER(scan) + EXTRA_STEP_2ARGS;
+ cc->next = next;
+ cc->minmod = minmod;
+ cc->lastloc = 0;
PL_reginput = locinput;
- n = regmatch(PREVOPER(next)); /* start on the WHILEM */
+ REGMATCH(PREVOPER(next), CURLYX); /* start on the WHILEM */
+ /*** all unsaved local vars undefined at this point */
regcpblow(cp);
- PL_regcc = cc.oldcc;
- saySAME(n);
+ Safefree(cc);
+ cc = oldcc;
+ 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;
+ lastloc = cc->lastloc; /* Detection of 0-len. */
+ cache_offset = 0;
+ cache_bit = 0;
n = cc->cur + 1; /* how many we know we matched */
PL_reginput = locinput;
/* 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;
+ oldcc = cc;
+ cc = cc->oldcc;
+ if (cc)
+ ln = cc->cur;
DEBUG_EXECUTE_r(
PerlIO_printf(Perl_debug_log,
"%*s empty match detected, try continuation...\n",
REPORT_CODE_OFF+PL_regindent*2, "")
);
- if (regmatch(cc->next))
+ REGMATCH(oldcc->next, WHILEM1);
+ /*** all unsaved local vars undefined at this point */
+ cc = oldcc;
+ if (result)
sayYES;
- if (PL_regcc)
- PL_regcc->cur = ln;
- PL_regcc = cc;
+ if (cc->oldcc)
+ cc->oldcc->cur = ln;
sayNO;
}
if (n < cc->min) {
cc->cur = n;
cc->lastloc = locinput;
- if (regmatch(cc->scan))
+ REGMATCH(cc->scan, WHILEM2);
+ /*** all unsaved local vars undefined at this point */
+ if (result)
sayYES;
cc->cur = n - 1;
cc->lastloc = lastloc;
}
else {
PL_reg_poscache_size = size;
- Newz(29, PL_reg_poscache, size, char);
+ Newxz(PL_reg_poscache, size, char);
}
DEBUG_EXECUTE_r(
PerlIO_printf(Perl_debug_log,
/* 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);
+ oldcc = cc;
+ cc = cc->oldcc;
+ if (cc)
+ ln = cc->cur;
+ cp = regcppush(oldcc->parenfloor);
REGCP_SET(lastcp);
- if (regmatch(cc->next)) {
+ REGMATCH(oldcc->next, WHILEM3);
+ /*** all unsaved local vars undefined at this point */
+ cc = oldcc;
+ if (result) {
regcpblow(cp);
CACHEsayYES; /* All done. */
}
REGCP_UNWIND(lastcp);
regcppop();
- if (PL_regcc)
- PL_regcc->cur = ln;
- PL_regcc = cc;
+ if (cc->oldcc)
+ cc->oldcc->cur = ln;
if (n >= cc->max) { /* Maximum greed exceeded? */
if (ckWARN(WARN_REGEXP) && n >= REG_INFTY
cc->lastloc = locinput;
cp = regcppush(cc->parenfloor);
REGCP_SET(lastcp);
- if (regmatch(cc->scan)) {
+ REGMATCH(cc->scan, WHILEM4);
+ /*** all unsaved local vars undefined at this point */
+ if (result) {
regcpblow(cp);
CACHEsayYES;
}
cc->cur = n;
cc->lastloc = locinput;
REGCP_SET(lastcp);
- if (regmatch(cc->scan)) {
+ REGMATCH(cc->scan, WHILEM5);
+ /*** all unsaved local vars undefined at this point */
+ if (result) {
regcpblow(cp);
CACHEsayYES;
}
}
/* 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))
+ oldcc = cc;
+ cc = cc->oldcc;
+ if (cc)
+ ln = cc->cur;
+ REGMATCH(oldcc->next, WHILEM6);
+ /*** all unsaved local vars undefined at this point */
+ cc = oldcc;
+ if (result)
CACHEsayYES;
- if (PL_regcc)
- PL_regcc->cur = ln;
- PL_regcc = cc;
+ if (cc->oldcc)
+ cc->oldcc->cur = ln;
cc->cur = n - 1;
cc->lastloc = lastloc;
CACHEsayNO;
}
- /* NOT REACHED */
+ /* NOTREACHED */
case BRANCHJ:
next = scan + ARG(scan);
if (next == scan)
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);
+ const I32 unwind1 = SSNEWt(1,re_unwind_branch_t);
+ re_unwind_branch_t * const uw = SSPTRt(unwind1,re_unwind_branch_t);
+
uw->prev = unwind;
unwind = unwind1;
uw->type = ((c1 == BRANCH)
break;
case CURLYM:
{
- I32 l = 0;
- CHECKPOINT lastcp;
+ curlym_l = matches = 0;
/* We suppose that the next guy does not need
backtracking: in particular, it is of constant non-zero length,
if (paren)
scan += NEXT_OFF(scan); /* Skip former OPEN. */
PL_reginput = locinput;
+ maxwanted = minmod ? ln : n;
+ if (maxwanted) {
+ while (PL_reginput < PL_regeol && matches < maxwanted) {
+ REGMATCH(scan, CURLYM1);
+ /*** all unsaved local vars undefined at this point */
+ if (!result)
+ break;
+ /* on first match, determine length, curlym_l */
+ if (!matches++) {
+ if (PL_reg_match_utf8) {
+ char *s = locinput;
+ while (s < PL_reginput) {
+ curlym_l++;
+ s += UTF8SKIP(s);
+ }
+ }
+ else {
+ curlym_l = PL_reginput - locinput;
+ }
+ if (curlym_l == 0) {
+ matches = maxwanted;
+ break;
+ }
+ }
+ locinput = PL_reginput;
+ }
+ }
+
+ PL_reginput = locinput;
+
if (minmod) {
minmod = 0;
- if (ln && regrepeat_hard(scan, ln, &l) < ln)
+ if (ln && matches < ln)
sayNO;
- locinput = PL_reginput;
if (HAS_TEXT(next) || JUMPABLE(next)) {
regnode *text_node = next;
if (paren) {
if (ln) {
PL_regstartp[paren] =
- HOPc(PL_reginput, -l) - PL_bostr;
+ HOPc(PL_reginput, -curlym_l) - PL_bostr;
PL_regendp[paren] = PL_reginput - PL_bostr;
}
else
PL_regendp[paren] = -1;
}
- if (regmatch(next))
+ REGMATCH(next, CURLYM2);
+ /*** all unsaved local vars undefined at this point */
+ if (result)
sayYES;
REGCP_UNWIND(lastcp);
}
/* Couldn't or didn't -- move forward. */
PL_reginput = locinput;
- if (regrepeat_hard(scan, 1, &l)) {
+ REGMATCH(scan, CURLYM3);
+ /*** all unsaved local vars undefined at this point */
+ if (result) {
ln++;
locinput = PL_reginput;
}
}
}
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)
+ "%*s matched %"IVdf" times, len=%"IVdf"...\n",
+ (int)(REPORT_CODE_OFF+PL_regindent*2), "",
+ (IV) matches, (IV)curlym_l)
);
- if (n >= ln) {
+ if (matches >= ln) {
if (HAS_TEXT(next) || JUMPABLE(next)) {
regnode *text_node = next;
}
assume_ok_REG:
REGCP_SET(lastcp);
- while (n >= ln) {
+ while (matches >= 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)
+ PerlIO_printf(Perl_debug_log,
+ "%*s trying tail with matches=%"IVdf"...\n",
+ (int)(REPORT_CODE_OFF+PL_regindent*2),
+ "", (IV)matches)
);
if (paren) {
- if (n) {
- PL_regstartp[paren] = HOPc(PL_reginput, -l) - PL_bostr;
+ if (matches) {
+ PL_regstartp[paren]
+ = HOPc(PL_reginput, -curlym_l) - PL_bostr;
PL_regendp[paren] = PL_reginput - PL_bostr;
}
else
PL_regendp[paren] = -1;
}
- if (regmatch(next))
+ REGMATCH(next, CURLYM4);
+ /*** all unsaved local vars undefined at this point */
+ if (result)
sayYES;
REGCP_UNWIND(lastcp);
}
/* Couldn't or didn't -- back up. */
- n--;
- locinput = HOPc(locinput, -l);
+ matches--;
+ locinput = HOPc(locinput, -curlym_l);
PL_reginput = locinput;
}
}
sayNO;
+ /* NOTREACHED */
break;
}
case CURLYN:
to_utf8_upper((U8*)s, tmpbuf2, &ulen2);
c1 = utf8n_to_uvuni(tmpbuf1, UTF8_MAXBYTES, 0,
- ckWARN(WARN_UTF8) ?
- 0 : UTF8_ALLOW_ANY);
+ uniflags);
c2 = utf8n_to_uvuni(tmpbuf2, UTF8_MAXBYTES, 0,
- ckWARN(WARN_UTF8) ?
- 0 : UTF8_ALLOW_ANY);
+ uniflags);
}
else {
c2 = c1 = utf8n_to_uvchr(s, UTF8_MAXBYTES, 0,
- ckWARN(WARN_UTF8) ?
- 0 : UTF8_ALLOW_ANY);
+ uniflags);
}
}
}
assume_ok_easy:
PL_reginput = locinput;
if (minmod) {
- CHECKPOINT lastcp;
minmod = 0;
if (ln && regrepeat(scan, ln) < ln)
sayNO;
locinput = PL_reginput;
REGCP_SET(lastcp);
if (c1 != -1000) {
- char *e; /* Should not check after this */
- char *old = locinput;
- int count = 0;
+ old = locinput;
+ count = 0;
if (n == REG_INFTY) {
e = PL_regeol - 1;
while (locinput <= e &&
utf8n_to_uvchr((U8*)locinput,
UTF8_MAXBYTES, &len,
- ckWARN(WARN_UTF8) ?
- 0 : UTF8_ALLOW_ANY) != (UV)c1) {
+ uniflags) != (UV)c1) {
locinput += len;
count++;
}
while (locinput <= e) {
UV c = utf8n_to_uvchr((U8*)locinput,
UTF8_MAXBYTES, &len,
- ckWARN(WARN_UTF8) ?
- 0 : UTF8_ALLOW_ANY);
+ uniflags);
if (c == (UV)c1 || c == (UV)c2)
break;
locinput += len;
sayNO;
}
/* PL_reginput == locinput now */
- TRYPAREN(paren, ln, locinput);
+ TRYPAREN(paren, ln, locinput, PLUS1);
+ /*** all unsaved local vars undefined at this point */
PL_reginput = locinput; /* Could be reset... */
REGCP_UNWIND(lastcp);
/* Couldn't or didn't -- move forward. */
if (do_utf8)
c = utf8n_to_uvchr((U8*)PL_reginput,
UTF8_MAXBYTES, 0,
- ckWARN(WARN_UTF8) ?
- 0 : UTF8_ALLOW_ANY);
+ uniflags);
else
c = UCHARAT(PL_reginput);
/* If it could work, try it. */
if (c == (UV)c1 || c == (UV)c2)
{
- TRYPAREN(paren, ln, PL_reginput);
+ TRYPAREN(paren, ln, PL_reginput, PLUS2);
+ /*** all unsaved local vars undefined at this point */
REGCP_UNWIND(lastcp);
}
}
/* If it could work, try it. */
else if (c1 == -1000)
{
- TRYPAREN(paren, ln, PL_reginput);
+ TRYPAREN(paren, ln, PL_reginput, PLUS3);
+ /*** all unsaved local vars undefined at this point */
REGCP_UNWIND(lastcp);
}
/* Couldn't or didn't -- move forward. */
}
}
else {
- CHECKPOINT lastcp;
n = regrepeat(scan, n);
locinput = PL_reginput;
if (ln < n && PL_regkind[(U8)OP(next)] == EOL &&
ln--;
}
REGCP_SET(lastcp);
- if (paren) {
- UV c = 0;
- while (n >= ln) {
- if (c1 != -1000) {
- if (do_utf8)
- c = utf8n_to_uvchr((U8*)PL_reginput,
- UTF8_MAXBYTES, 0,
- ckWARN(WARN_UTF8) ?
- 0 : UTF8_ALLOW_ANY);
- else
- c = UCHARAT(PL_reginput);
- }
- /* If it could work, try it. */
- if (c1 == -1000 || c == (UV)c1 || c == (UV)c2)
- {
- TRYPAREN(paren, n, PL_reginput);
- REGCP_UNWIND(lastcp);
- }
- /* Couldn't or didn't -- back up. */
- n--;
- PL_reginput = locinput = HOPc(locinput, -1);
- }
- }
- else {
+ {
UV c = 0;
while (n >= ln) {
if (c1 != -1000) {
if (do_utf8)
c = utf8n_to_uvchr((U8*)PL_reginput,
UTF8_MAXBYTES, 0,
- ckWARN(WARN_UTF8) ?
- 0 : UTF8_ALLOW_ANY);
+ 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);
+ TRYPAREN(paren, n, PL_reginput, PLUS4);
+ /*** all unsaved local vars undefined at this point */
REGCP_UNWIND(lastcp);
}
/* Couldn't or didn't -- back up. */
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. */
+ cur_call_cc = PL_reg_call_cc;
+ end_re = PL_reg_re;
+
+ /* Save *all* the positions. */
+ cp = regcppush(0);
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. */
+
+ /* Restore parens of the caller. */
+ {
+ I32 tmp = PL_savestack_ix;
+ PL_savestack_ix = PL_reg_call_cc->ss;
+ regcppop();
+ PL_savestack_ix = tmp;
+ }
+
+ /* Make position available to the callcc. */
+ PL_reginput = locinput;
+
cache_re(PL_reg_call_cc->re);
- PL_regcc = PL_reg_call_cc->cc;
+ oldcc = cc;
+ cc = PL_reg_call_cc->cc;
PL_reg_call_cc = PL_reg_call_cc->prev;
- if (regmatch(cur_call_cc->node)) {
+ REGMATCH(cur_call_cc->node, END);
+ /*** all unsaved local vars undefined at this point */
+ if (result) {
PL_reg_call_cc = cur_call_cc;
regcpblow(cp);
sayYES;
REGCP_UNWIND(lastcp);
regcppop();
PL_reg_call_cc = cur_call_cc;
- PL_regcc = cctmp;
- PL_reg_re = re;
- cache_re(re);
+ cc = oldcc;
+ PL_reg_re = end_re;
+ cache_re(end_re);
DEBUG_EXECUTE_r(
PerlIO_printf(Perl_debug_log,
case UNLESSM:
n = 0;
if (scan->flags) {
- s = HOPBACKc(locinput, scan->flags);
+ char * const s = HOPBACKc(locinput, scan->flags);
if (!s)
goto say_yes;
PL_reginput = s;
case IFMATCH:
n = 1;
if (scan->flags) {
- s = HOPBACKc(locinput, scan->flags);
+ char * const s = HOPBACKc(locinput, scan->flags);
if (!s)
goto say_no;
PL_reginput = s;
PL_reginput = locinput;
do_ifmatch:
- inner = NEXTOPER(NEXTOPER(scan));
- if (regmatch(inner) != n) {
+ REGMATCH(NEXTOPER(NEXTOPER(scan)), IFMATCH);
+ /*** all unsaved local vars undefined at this point */
+ if (result != n) {
say_no:
if (logical) {
logical = 0;
PTR2UV(scan), OP(scan));
Perl_croak(aTHX_ "regexp memory corruption");
}
+
reenter:
scan = next;
+ 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:
+ {
+ /* save existing local variables */
+ struct regmatch_state *p;
+
+ Newx(p, 1, struct regmatch_state);
+ p->prev_state = prev_state;
+ prev_state = p;
+ p->resume_state = resume_state;
+ p->scan = scan;
+ p->next = next;
+ p->minmod = minmod;
+ p->sw = sw;
+ p->logical = logical;
+ p->unwind = unwind;
+ p->cc = cc;
+ p->locinput = locinput;
+ p->n = n;
+ p->ln = ln;
+ p->c1 = c1;
+ p->c2 = c2;
+ p->paren = paren;
+ p->cp = cp;
+ p->lastcp = lastcp;
+ p->oldcc = oldcc;
+ p->lastloc = lastloc;
+ p->cache_offset = cache_offset;
+ p->cache_bit = cache_bit;
+ p->curlym_l = curlym_l;
+ p->matches = matches;
+ p->maxwanted = maxwanted;
+ p->e = e;
+ p->old = old;
+ p->count = count;
+ p->cur_call_cc = cur_call_cc;
+ p->end_re = end_re;
+ p->accept_buff = accept_buff;
+ p->accepted = accepted;
+ p->reg_call_cc = PL_reg_call_cc;
+
+ scan = new_scan;
+ locinput = PL_reginput;
+ nextchr = UCHARAT(locinput);
+ minmod = 0;
+ sw = 0;
+ logical = 0;
+ unwind = 0;
+#ifdef DEBUGGING
+ PL_regindent++;
+#endif
+ }
}
/*
PL_regindent--;
#endif
-#if 0 /* Breaks $^R */
- if (unwind)
- regcpblow(firstcp);
-#endif
- return 1;
+ result = 1;
+ goto exit_level;
no:
DEBUG_EXECUTE_r(
no_final:
do_no:
if (unwind) {
- re_unwind_t *uw = SSPTRt(unwind,re_unwind_t);
+ re_unwind_t * const uw = SSPTRt(unwind,re_unwind_t);
switch (uw->type) {
case RE_UNWIND_BRANCH:
case RE_UNWIND_BRANCHJ:
{
- re_unwind_branch_t *uwb = &(uw->branch);
+ re_unwind_branch_t * const uwb = &(uw->branch);
const I32 lastparen = uwb->lastparen;
REGCP_UNWIND(uwb->lastcp);
goto reenter;
}
- /* NOT REACHED */
+ /* NOTREACHED */
default:
Perl_croak(aTHX_ "regexp unwind memory corruption");
}
- /* NOT REACHED */
+ /* NOTREACHED */
}
#ifdef DEBUGGING
PL_regindent--;
#endif
- return 0;
+ result = 0;
+exit_level:
+ if (prev_state) {
+ /* restore previous state and re-enter */
+
+ struct regmatch_state *p = prev_state;
+ resume_state = p->resume_state;
+ scan = p->scan;
+ next = p->next;
+ minmod = p->minmod;
+ sw = p->sw;
+ logical = p->logical;
+ unwind = p->unwind;
+ cc = p->cc;
+ locinput = p->locinput;
+ nextchr = UCHARAT(locinput);
+ n = p->n;
+ ln = p->ln;
+ c1 = p->c1;
+ c2 = p->c2;
+ paren = p->paren;
+ cp = p->cp;
+ lastcp = p->lastcp;
+ oldcc = p->oldcc;
+ lastloc = p->lastloc;
+ cache_offset = p->cache_offset;
+ cache_bit = p->cache_bit;
+ curlym_l = p->curlym_l;
+ matches = p->matches;
+ maxwanted = p->maxwanted;
+ e = p->e;
+ old = p->old;
+ count = p->count;
+ cur_call_cc = p->cur_call_cc;
+ end_re = p->end_re;
+ accept_buff = p->accept_buff;
+ accepted = p->accepted;
+ PL_reg_call_cc = p->reg_call_cc;
+ prev_state = p->prev_state;
+ Safefree(p);
+
+ switch (resume_state) {
+ case resume_TRIE1:
+ goto resume_point_TRIE1;
+ case resume_TRIE2:
+ goto resume_point_TRIE2;
+ 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 resume_CURLYM1:
+ goto resume_point_CURLYM1;
+ case resume_CURLYM2:
+ goto resume_point_CURLYM2;
+ case resume_CURLYM3:
+ goto resume_point_CURLYM3;
+ case resume_CURLYM4:
+ goto resume_point_CURLYM4;
+ case resume_IFMATCH:
+ goto resume_point_IFMATCH;
+ case resume_PLUS1:
+ goto resume_point_PLUS1;
+ case resume_PLUS2:
+ goto resume_point_PLUS2;
+ case resume_PLUS3:
+ goto resume_point_PLUS3;
+ case resume_PLUS4:
+ goto resume_point_PLUS4;
+ case resume_END:
+ goto resume_point_END;
+ default:
+ Perl_croak(aTHX_ "regexp resume memory corruption");
+ }
+ /* NOTREACHED */
+ }
+ return result;
+
}
/*
DEBUG_r({
SV *re_debug_flags = NULL;
- SV *prop = sv_newmortal();
+ SV * const prop = sv_newmortal();
GET_RE_DEBUG_FLAGS;
DEBUG_EXECUTE_r({
regprop(prop, p);
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;
-}
/*
- regclass_swash - prepare the utf8 swash
SV *
Perl_regclass_swash(pTHX_ register const regnode* node, bool doinit, SV** listsvp, SV **altsvp)
{
+ dVAR;
SV *sw = NULL;
SV *si = NULL;
SV *alt = NULL;
const U32 n = ARG(node);
if (PL_regdata->what[n] == 's') {
- SV *rv = (SV*)PL_regdata->data[n];
- AV *av = (AV*)SvRV((SV*)rv);
- SV **ary = AvARRAY(av);
+ SV * const rv = (SV*)PL_regdata->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)
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,
+ ckWARN(WARN_UTF8) ? UTF8_CHECK_ONLY :
+ UTF8_ALLOW_ANYUV|UTF8_CHECK_ONLY);
+ 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 *sw = regclass_swash(n, TRUE, 0, (SV**)&av);
+ SV * const sw = regclass_swash(n, TRUE, 0, (SV**)&av);
if (sw) {
if (swash_fetch(sw, p, do_utf8))
else if (flags & ANYOF_FOLD) {
if (!match && lenp && av) {
I32 i;
-
for (i = 0; i <= av_len(av); i++) {
- SV* sv = *av_fetch(av, i, FALSE);
+ SV* const sv = *av_fetch(av, i, FALSE);
STRLEN len;
- const char *s = SvPV_const(sv, len);
+ const char * const s = SvPV_const(sv, len);
if (len <= plen && memEQ(s, (char*)p, len)) {
*lenp = len;
}
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, U8* lim)
{
+ dVAR;
if (off >= 0) {
while (off-- && s < lim) {
/* XXX could check well-formedness here */
}
STATIC U8 *
-S_reghopmaybe(pTHX_ U8 *s, I32 off)
-{
- return S_reghopmaybe3(aTHX_ s, off, (U8*)(off >= 0 ? PL_regeol : PL_bostr));
-}
-
-STATIC U8 *
-S_reghopmaybe3(pTHX_ U8* s, I32 off, U8* lim)
+S_reghopmaybe3(U8* s, I32 off, U8* lim)
{
+ dVAR;
if (off >= 0) {
while (off-- && s < lim) {
/* XXX could check well-formedness here */
static void
restore_pos(pTHX_ void *arg)
{
+ dVAR;
PERL_UNUSED_ARG(arg);
if (PL_reg_eval_set) {
if (PL_reg_oldsaved) {
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);
projects / p5sagit/p5-mst-13.2.git / blobdiff