if (paren_elems_to_push < 0)
Perl_croak(aTHX_ "panic: paren_elems_to_push < 0");
-#define REGCP_OTHER_ELEMS 8
+#define REGCP_OTHER_ELEMS 7
SSGROW(paren_elems_to_push + REGCP_OTHER_ELEMS);
for (p = PL_regsize; p > parenfloor; p--) {
/* REGCP_PARENS_ELEMS are pushed per pairs of parentheses. */
- SSPUSHINT(PL_regendp[p]);
- SSPUSHINT(PL_regstartp[p]);
+ SSPUSHINT(PL_regoffs[p].end);
+ SSPUSHINT(PL_regoffs[p].start);
SSPUSHPTR(PL_reg_start_tmp[p]);
SSPUSHINT(p);
DEBUG_BUFFERS_r(PerlIO_printf(Perl_debug_log,
" saving \\%"UVuf" %"IVdf"(%"IVdf")..%"IVdf"\n",
- (UV)p, (IV)PL_regstartp[p],
+ (UV)p, (IV)PL_regoffs[p].start,
(IV)(PL_reg_start_tmp[p] - PL_bostr),
- (IV)PL_regendp[p]
+ (IV)PL_regoffs[p].end
));
}
/* REGCP_OTHER_ELEMS are pushed in any case, parentheses or no. */
- SSPUSHPTR(PL_regstartp);
- SSPUSHPTR(PL_regendp);
+ SSPUSHPTR(PL_regoffs);
SSPUSHINT(PL_regsize);
SSPUSHINT(*PL_reglastparen);
SSPUSHINT(*PL_reglastcloseparen);
*PL_reglastcloseparen = SSPOPINT;
*PL_reglastparen = SSPOPINT;
PL_regsize = SSPOPINT;
- PL_regendp=(I32 *) SSPOPPTR;
- PL_regstartp=(I32 *) SSPOPPTR;
+ PL_regoffs=(regexp_paren_pair *) SSPOPPTR;
/* Now restore the parentheses context. */
I32 tmps;
U32 paren = (U32)SSPOPINT;
PL_reg_start_tmp[paren] = (char *) SSPOPPTR;
- PL_regstartp[paren] = SSPOPINT;
+ PL_regoffs[paren].start = SSPOPINT;
tmps = SSPOPINT;
if (paren <= *PL_reglastparen)
- PL_regendp[paren] = tmps;
+ PL_regoffs[paren].end = tmps;
DEBUG_BUFFERS_r(
PerlIO_printf(Perl_debug_log,
" restoring \\%"UVuf" to %"IVdf"(%"IVdf")..%"IVdf"%s\n",
- (UV)paren, (IV)PL_regstartp[paren],
+ (UV)paren, (IV)PL_regoffs[paren].start,
(IV)(PL_reg_start_tmp[paren] - PL_bostr),
- (IV)PL_regendp[paren],
+ (IV)PL_regoffs[paren].end,
(paren > *PL_reglastparen ? "(no)" : ""));
);
}
* --jhi */
for (i = *PL_reglastparen + 1; i <= rex->nparens; i++) {
if (i > PL_regsize)
- PL_regstartp[i] = -1;
- PL_regendp[i] = -1;
+ PL_regoffs[i].start = -1;
+ PL_regoffs[i].end = -1;
}
#endif
return input;
if ((!reginfo || regtry(reginfo, &s))) \
goto got_it
+#define REXEC_FBC_CSCAN(CoNdUtF8,CoNd) \
+ if (do_utf8) { \
+ REXEC_FBC_UTF8_CLASS_SCAN(CoNdUtF8); \
+ } \
+ else { \
+ REXEC_FBC_CLASS_SCAN(CoNd); \
+ } \
+ break
+
#define REXEC_FBC_CSCAN_PRELOAD(UtFpReLoAd,CoNdUtF8,CoNd) \
if (do_utf8) { \
UtFpReLoAd; \
!isDIGIT_LC_utf8((U8*)s),
!isDIGIT_LC(*s)
);
+ case LNBREAK:
+ REXEC_FBC_CSCAN(
+ is_LNBREAK_utf8(s),
+ is_LNBREAK_latin1(s)
+ );
+ case VERTWS:
+ REXEC_FBC_CSCAN(
+ is_VERTWS_utf8(s),
+ is_VERTWS_latin1(s)
+ );
+ case NVERTWS:
+ REXEC_FBC_CSCAN(
+ !is_VERTWS_utf8(s),
+ !is_VERTWS_latin1(s)
+ );
+ case HORIZWS:
+ REXEC_FBC_CSCAN(
+ is_HORIZWS_utf8(s),
+ is_HORIZWS_latin1(s)
+ );
+ case NHORIZWS:
+ REXEC_FBC_CSCAN(
+ !is_HORIZWS_utf8(s),
+ !is_HORIZWS_latin1(s)
+ );
case AHOCORASICKC:
case AHOCORASICK:
{
static void
S_swap_match_buff (pTHX_ regexp *prog) {
- I32 *t;
+ regexp_paren_pair *t;
if (!prog->swap) {
/* We have to be careful. If the previous successful match
to the re, and switch the buffer each match. If we fail
we switch it back, otherwise we leave it swapped.
*/
- Newxz(prog->swap, 1, regexp_paren_ofs);
- /* no need to copy these */
- Newxz(prog->swap->startp, 2 * (prog->nparens + 1), I32);
- prog->swap->endp = prog->swap->startp + prog->nparens + 1;
+ Newxz(prog->swap, (prog->nparens + 1), regexp_paren_pair);
}
- t = prog->swap->startp;
- prog->swap->startp = prog->startp;
- prog->startp = t;
- t = prog->swap->endp;
- prog->swap->endp = prog->endp;
- prog->endp = t;
+ t = prog->swap;
+ prog->swap = prog->offs;
+ prog->offs = t;
}
I32 end_shift = 0; /* Same for the end. */ /* CC */
I32 scream_pos = -1; /* Internal iterator of scream. */
char *scream_olds = NULL;
- SV* const oreplsv = GvSV(PL_replgv);
const bool do_utf8 = (bool)DO_UTF8(sv);
I32 multiline;
RXi_GET_DECL(prog,progi);
got_it:
RX_MATCH_TAINTED_set(prog, PL_reg_flags & RF_tainted);
- if (PL_reg_eval_set) {
- /* Preserve the current value of $^R */
- if (oreplsv != GvSV(PL_replgv))
- sv_setsv(oreplsv, GvSV(PL_replgv));/* So that when GvSV(replgv) is
- restored, the value remains
- the same. */
+ if (PL_reg_eval_set)
restore_pos(aTHX_ prog);
- }
if (prog->paren_names)
(void)hv_iterinit(prog->paren_names);
prog->sublen = PL_regeol - PL_bostr; /* strend may have been modified */
}
DEBUG_EXECUTE_r(PL_reg_starttry = *startpos);
- prog->startp[0] = *startpos - PL_bostr;
+ prog->offs[0].start = *startpos - PL_bostr;
PL_reginput = *startpos;
PL_reglastparen = &prog->lastparen;
PL_reglastcloseparen = &prog->lastcloseparen;
prog->lastparen = 0;
prog->lastcloseparen = 0;
PL_regsize = 0;
- PL_regstartp = prog->startp;
- PL_regendp = prog->endp;
+ PL_regoffs = prog->offs;
if (PL_reg_start_tmpl <= prog->nparens) {
PL_reg_start_tmpl = prog->nparens*3/2 + 3;
if(PL_reg_start_tmp)
* --jhi */
#if 1
if (prog->nparens) {
- I32 *sp = PL_regstartp;
- I32 *ep = PL_regendp;
+ regexp_paren_pair *pp = PL_regoffs;
register I32 i;
for (i = prog->nparens; i > (I32)*PL_reglastparen; i--) {
- *++sp = -1;
- *++ep = -1;
+ ++pp;
+ pp->start = -1;
+ pp->end = -1;
}
}
#endif
REGCP_SET(lastcp);
if (regmatch(reginfo, progi->program + 1)) {
- PL_regendp[0] = PL_reginput - PL_bostr;
+ PL_regoffs[0].end = PL_reginput - PL_bostr;
return 1;
}
if (reginfo->cutpoint)
I32 *nums=(I32*)SvPVX(sv_dat);
for ( n=0; n<SvIVX(sv_dat); n++ ) {
if ((I32)*PL_reglastparen >= nums[n] &&
- PL_regendp[nums[n]] != -1)
+ PL_regoffs[nums[n]].end != -1)
{
return nums[n];
}
return 0;
}
+
+/* free all slabs above current one - called during LEAVE_SCOPE */
+
+STATIC void
+S_clear_backtrack_stack(pTHX_ void *p)
+{
+ regmatch_slab *s = PL_regmatch_slab->next;
+ PERL_UNUSED_ARG(p);
+
+ if (!s)
+ return;
+ PL_regmatch_slab->next = NULL;
+ while (s) {
+ regmatch_slab * const osl = s;
+ s = s->next;
+ Safefree(osl);
+ }
+}
+
+
#define SETREX(Re1,Re2) \
if (PL_reg_eval_set) PM_SETRE((PL_reg_curpm), (Re2)); \
Re1 = (Re2)
regexp *rex = reginfo->prog;
RXi_GET_DECL(rex,rexi);
- regmatch_slab *orig_slab;
- regmatch_state *orig_state;
+ I32 oldsave;
/* the current state. This is a cached copy of PL_regmatch_state */
register regmatch_state *st;
during a successfull match */
U32 lastopen = 0; /* last open we saw */
bool has_cutgroup = RX_HAS_CUTGROUP(rex) ? 1 : 0;
+
+ SV* const oreplsv = GvSV(PL_replgv);
/* these three flags are set by various ops to signal information to
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;
+ oldsave = PL_savestack_ix;
+ SAVEDESTRUCTOR_X(S_clear_backtrack_stack, NULL);
+ SAVEVPTR(PL_regmatch_slab);
+ SAVEVPTR(PL_regmatch_state);
/* grab next free state slot */
st = ++PL_regmatch_state;
case KEEPS:
/* update the startpoint */
- st->u.keeper.val = PL_regstartp[0];
+ st->u.keeper.val = PL_regoffs[0].start;
PL_reginput = locinput;
- PL_regstartp[0] = locinput - PL_bostr;
+ PL_regoffs[0].start = locinput - PL_bostr;
PUSH_STATE_GOTO(KEEPS_next, next);
/*NOT-REACHED*/
case KEEPS_next_fail:
/* rollback the start point change */
- PL_regstartp[0] = st->u.keeper.val;
+ PL_regoffs[0].start = st->u.keeper.val;
sayNO_SILENT;
/*NOT-REACHED*/
case EOL:
if ( ST.jump) {
REGCP_UNWIND(ST.cp);
for (n = *PL_reglastparen; n > ST.lastparen; n--)
- PL_regendp[n] = -1;
+ PL_regoffs[n].end = -1;
*PL_reglastparen = n;
}
trie_first_try:
* pack("U0U*", 0xDF) =~ /ss/i,
* the 0xC3 0x9F are the UTF-8
* byte sequence for the U+00DF. */
+
if (!(do_utf8 &&
- toLOWER(s[0]) == 's' &&
+ toLOWER(s[0]) == 's' &&
ln >= 2 &&
toLOWER(s[1]) == 's' &&
(U8)l[0] == 0xC3 &&
n = ARG(scan); /* which paren pair */
type = OP(scan);
do_ref:
- ln = PL_regstartp[n];
+ ln = PL_regoffs[n].start;
PL_reg_leftiter = PL_reg_maxiter; /* Void cache */
if (*PL_reglastparen < n || ln == -1)
sayNO; /* Do not match unless seen CLOSEn. */
- if (ln == PL_regendp[n])
+ if (ln == PL_regoffs[n].end)
break;
s = PL_bostr + ln;
if (do_utf8 && type != REF) { /* REF can do byte comparison */
char *l = locinput;
- const char *e = PL_bostr + PL_regendp[n];
+ const char *e = PL_bostr + PL_regoffs[n].end;
/*
* Note that we can't do the "other character" lookup trick as
* in the 8-bit case (no pun intended) because in Unicode we
(UCHARAT(s) != (type == REFF
? PL_fold : PL_fold_locale)[nextchr])))
sayNO;
- ln = PL_regendp[n] - ln;
+ ln = PL_regoffs[n].end - ln;
if (locinput + ln > PL_regeol)
sayNO;
if (ln > 1 && (type == REF
DEBUG_STATE_r( PerlIO_printf(Perl_debug_log,
" re_eval 0x%"UVxf"\n", PTR2UV(PL_op)) );
PAD_SAVE_LOCAL(old_comppad, (PAD*)rexi->data->data[n + 2]);
- PL_regendp[0] = PL_reg_magic->mg_len = locinput - PL_bostr;
+ PL_regoffs[0].end = PL_reg_magic->mg_len = locinput - PL_bostr;
if (sv_yes_mark) {
SV *sv_mrk = get_sv("REGMARK", 1);
re = reg_temp_copy((regexp *)mg->mg_obj); /*XXX:dmq*/
}
else {
- STRLEN len;
- const char * const t = SvPV_const(ret, len);
- PMOP pm;
+ U32 pm_flags = 0;
const I32 osize = PL_regsize;
- Zero(&pm, 1, PMOP);
- if (DO_UTF8(ret)) pm.op_pmdynflags |= PMdf_DYN_UTF8;
- re = CALLREGCOMP((char*)t, (char*)t + len, &pm);
+ if (DO_UTF8(ret)) pm_flags |= RXf_UTF8;
+ re = CALLREGCOMP(ret, pm_flags);
if (!(SvFLAGS(ret)
& (SVs_TEMP | SVs_PADTMP | SVf_READONLY
| SVs_GMG)))
ST.cp = regcppush(0); /* Save *all* the positions. */
REGCP_SET(ST.lastcp);
- PL_regstartp = re->startp; /* essentially NOOP on GOSUB */
- PL_regendp = re->endp; /* essentially NOOP on GOSUB */
+ PL_regoffs = re->offs; /* essentially NOOP on GOSUB */
*PL_reglastparen = 0;
*PL_reglastcloseparen = 0;
break;
case CLOSE:
n = ARG(scan); /* which paren pair */
- PL_regstartp[n] = PL_reg_start_tmp[n] - PL_bostr;
- PL_regendp[n] = locinput - PL_bostr;
+ PL_regoffs[n].start = PL_reg_start_tmp[n] - PL_bostr;
+ PL_regoffs[n].end = locinput - PL_bostr;
/*if (n > PL_regsize)
PL_regsize = n;*/
if (n > *PL_reglastparen)
if ( OP(cursor)==CLOSE ){
n = ARG(cursor);
if ( n <= lastopen ) {
- PL_regstartp[n] = PL_reg_start_tmp[n] - PL_bostr;
- PL_regendp[n] = locinput - PL_bostr;
+ PL_regoffs[n].start
+ = PL_reg_start_tmp[n] - PL_bostr;
+ PL_regoffs[n].end = locinput - PL_bostr;
/*if (n > PL_regsize)
PL_regsize = n;*/
if (n > *PL_reglastparen)
/*NOTREACHED*/
case GROUPP:
n = ARG(scan); /* which paren pair */
- sw = (bool)(*PL_reglastparen >= n && PL_regendp[n] != -1);
+ sw = (bool)(*PL_reglastparen >= n && PL_regoffs[n].end != -1);
break;
case NGROUPP:
/* reg_check_named_buff_matched returns 0 for no match */
}
case CURLYX_end: /* just finished matching all of A*B */
- if (PL_reg_eval_set){
- SV *pres= GvSV(PL_replgv);
- SvREFCNT_inc(pres);
- regcpblow(ST.cp);
- sv_setsv(GvSV(PL_replgv), pres);
- SvREFCNT_dec(pres);
- } else {
- regcpblow(ST.cp);
- }
cur_curlyx = ST.prev_curlyx;
sayYES;
/* NOTREACHED */
}
REGCP_UNWIND(ST.cp);
for (n = *PL_reglastparen; n > ST.lastparen; n--)
- PL_regendp[n] = -1;
+ PL_regoffs[n].end = -1;
*PL_reglastparen = n;
/*dmq: *PL_reglastcloseparen = n; */
scan = ST.next_branch;
/* mark current A as captured */
I32 paren = ST.me->flags;
if (ST.count) {
- PL_regstartp[paren]
+ PL_regoffs[paren].start
= HOPc(PL_reginput, -ST.alen) - PL_bostr;
- PL_regendp[paren] = PL_reginput - PL_bostr;
+ PL_regoffs[paren].end = PL_reginput - PL_bostr;
/*dmq: *PL_reglastcloseparen = paren; */
}
else
- PL_regendp[paren] = -1;
+ PL_regoffs[paren].end = -1;
if (cur_eval && cur_eval->u.eval.close_paren &&
cur_eval->u.eval.close_paren == (U32)ST.me->flags)
{
#define CURLY_SETPAREN(paren, success) \
if (paren) { \
if (success) { \
- PL_regstartp[paren] = HOPc(locinput, -1) - PL_bostr; \
- PL_regendp[paren] = locinput - PL_bostr; \
+ PL_regoffs[paren].start = HOPc(locinput, -1) - PL_bostr; \
+ PL_regoffs[paren].end = locinput - PL_bostr; \
*PL_reglastcloseparen = paren; \
} \
else \
- PL_regendp[paren] = -1; \
+ PL_regoffs[paren].end = -1; \
}
case STAR: /* /A*B/ where A is width 1 */
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_regoffs[ST.paren].end = -1;
PL_reginput = locinput; /* Could be reset... */
REGCP_UNWIND(ST.cp);
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;
+ PL_regoffs[ST.paren].end = -1;
REGCP_UNWIND(ST.cp);
/* failed -- move forward one */
case CURLY_B_max_fail:
/* failed to find B in a greedy match */
if (ST.paren && ST.count)
- PL_regendp[ST.paren] = -1;
+ PL_regoffs[ST.paren].end = -1;
REGCP_UNWIND(ST.cp);
/* back up. */
sayNO;
/* NOTREACHED */
#undef ST
+ case FOLDCHAR:
+ n = ARG(scan);
+ if ( n == what_len_TRICKYFOLD(locinput,do_utf8,ln) ) {
+ locinput += ln;
+ } else if ( 0xDF == n && !do_utf8 && !UTF ) {
+ sayNO;
+ } else {
+ U8 folded[UTF8_MAXBYTES_CASE+1];
+ STRLEN foldlen;
+ const char * const l = locinput;
+ char *e = PL_regeol;
+ to_uni_fold(n, folded, &foldlen);
+
+ if (ibcmp_utf8(folded, 0, foldlen, 1,
+ l, &e, 0, do_utf8)) {
+ sayNO;
+ }
+ locinput = e;
+ }
+ nextchr = UCHARAT(locinput);
+ break;
+ case LNBREAK:
+ if ((n=is_LNBREAK(locinput,do_utf8))) {
+ locinput += n;
+ nextchr = UCHARAT(locinput);
+ } else
+ sayNO;
+ break;
+
+#define CASE_CLASS(nAmE) \
+ case nAmE: \
+ if ((n=is_##nAmE(locinput,do_utf8))) { \
+ locinput += n; \
+ nextchr = UCHARAT(locinput); \
+ } else \
+ sayNO; \
+ break; \
+ case N##nAmE: \
+ if ((n=is_##nAmE(locinput,do_utf8))) { \
+ sayNO; \
+ } else { \
+ locinput += UTF8SKIP(locinput); \
+ nextchr = UCHARAT(locinput); \
+ } \
+ break
+
+ CASE_CLASS(VERTWS);
+ CASE_CLASS(HORIZWS);
+#undef CASE_CLASS
default:
PerlIO_printf(Perl_error_log, "%"UVxf" %d\n",
DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "%sMatch successful!%s\n",
PL_colors[4], PL_colors[5]));
+ if (PL_reg_eval_set) {
+ /* each successfully executed (?{...}) block does the equivalent of
+ * local $^R = do {...}
+ * When popping the save stack, all these locals would be undone;
+ * bypass this by setting the outermost saved $^R to the latest
+ * value */
+ if (oreplsv != GvSV(PL_replgv))
+ sv_setsv(oreplsv, GvSV(PL_replgv));
+ }
result = 1;
goto final_exit;
sv_setsv(sv_err, sv_commit);
sv_setsv(sv_mrk, sv_yes_mark);
}
- /* restore original high-water mark */
- PL_regmatch_slab = orig_slab;
- PL_regmatch_state = orig_state;
-
- /* 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);
- }
- }
+
+ /* clean up; in particular, free all slabs above current one */
+ LEAVE_SCOPE(oldsave);
return result;
}
while (scan < loceol && !isDIGIT(*scan))
scan++;
}
+ case LNBREAK:
+ if (do_utf8) {
+ loceol = PL_regeol;
+ while (hardcount < max && scan < loceol && (c=is_LNBREAK_utf8(scan))) {
+ scan += c;
+ hardcount++;
+ }
+ } else {
+ /*
+ LNBREAK can match two latin chars, which is ok,
+ because we have a null terminated string, but we
+ have to use hardcount in this situation
+ */
+ while (scan < loceol && (c=is_LNBREAK_latin1(scan))) {
+ scan+=c;
+ hardcount++;
+ }
+ }
+ break;
+ case HORIZWS:
+ if (do_utf8) {
+ loceol = PL_regeol;
+ while (hardcount < max && scan < loceol && (c=is_HORIZWS_utf8(scan))) {
+ scan += c;
+ hardcount++;
+ }
+ } else {
+ while (scan < loceol && is_HORIZWS_latin1(scan))
+ scan++;
+ }
+ break;
+ case NHORIZWS:
+ if (do_utf8) {
+ loceol = PL_regeol;
+ while (hardcount < max && scan < loceol && !is_HORIZWS_utf8(scan)) {
+ scan += UTF8SKIP(scan);
+ hardcount++;
+ }
+ } else {
+ while (scan < loceol && !is_HORIZWS_latin1(scan))
+ scan++;
+
+ }
+ break;
+ case VERTWS:
+ if (do_utf8) {
+ loceol = PL_regeol;
+ while (hardcount < max && scan < loceol && (c=is_VERTWS_utf8(scan))) {
+ scan += c;
+ hardcount++;
+ }
+ } else {
+ while (scan < loceol && is_VERTWS_latin1(scan))
+ scan++;
+
+ }
break;
+ case NVERTWS:
+ if (do_utf8) {
+ loceol = PL_regeol;
+ while (hardcount < max && scan < loceol && !is_VERTWS_utf8(scan)) {
+ scan += UTF8SKIP(scan);
+ hardcount++;
+ }
+ } else {
+ while (scan < loceol && !is_VERTWS_latin1(scan))
+ scan++;
+
+ }
+ break;
+
default: /* Called on something of 0 width. */
break; /* So match right here or not at all. */
}
projects / p5sagit/p5-mst-13.2.git / blobdiff