#define UTF_PATTERN ((PL_reg_flags & RF_utf8) != 0)
-#define RS_init 1 /* eval environment created */
-#define RS_set 2 /* replsv value is set */
-
#ifndef STATIC
#define STATIC static
#endif
/* Currently these are only used when PL_regkind[OP(rn)] == EXACT so
we don't need this definition. */
#define IS_TEXT(rn) ( OP(rn)==EXACT || OP(rn)==REF || OP(rn)==NREF )
-#define IS_TEXTF(rn) ( (OP(rn)==EXACTFU || OP(rn)==EXACTFA || OP(rn)==EXACTF) || OP(rn)==REFF || OP(rn)==NREFF )
+#define IS_TEXTF(rn) ( OP(rn)==EXACTFU || OP(rn)==EXACTFU_SS || OP(rn)==EXACTFU_TRICKYFOLD || OP(rn)==EXACTFA || OP(rn)==EXACTF || OP(rn)==REFF || OP(rn)==NREFF )
#define IS_TEXTFL(rn) ( OP(rn)==EXACTFL || OP(rn)==REFFL || OP(rn)==NREFFL )
#else
/* ... so we use this as its faster. */
#define IS_TEXT(rn) ( OP(rn)==EXACT )
-#define IS_TEXTFU(rn) ( OP(rn)==EXACTFU || OP(rn) == EXACTFA)
+#define IS_TEXTFU(rn) ( OP(rn)==EXACTFU || OP(rn)==EXACTFU_SS || OP(rn)==EXACTFU_TRICKYFOLD || OP(rn) == EXACTFA)
#define IS_TEXTF(rn) ( OP(rn)==EXACTF )
#define IS_TEXTFL(rn) ( OP(rn)==EXACTFL )
static void restore_pos(pTHX_ void *arg);
-#define REGCP_PAREN_ELEMS 4
-#define REGCP_OTHER_ELEMS 5
+#define REGCP_PAREN_ELEMS 3
+#define REGCP_OTHER_ELEMS 3
#define REGCP_FRAME_ELEMS 1
/* REGCP_FRAME_ELEMS are not part of the REGCP_OTHER_ELEMS and
* are needed for the regexp context stack bookkeeping. */
STATIC CHECKPOINT
-S_regcppush(pTHX_ I32 parenfloor)
+S_regcppush(pTHX_ const regexp *rex, I32 parenfloor)
{
dVAR;
const int retval = PL_savestack_ix;
const int paren_elems_to_push = (PL_regsize - parenfloor) * REGCP_PAREN_ELEMS;
const UV total_elems = paren_elems_to_push + REGCP_OTHER_ELEMS;
const UV elems_shifted = total_elems << SAVE_TIGHT_SHIFT;
- int p;
+ I32 p;
GET_RE_DEBUG_FLAGS_DECL;
+ PERL_ARGS_ASSERT_REGCPPUSH;
+
if (paren_elems_to_push < 0)
- Perl_croak(aTHX_ "panic: paren_elems_to_push < 0");
+ Perl_croak(aTHX_ "panic: paren_elems_to_push, %i < 0",
+ paren_elems_to_push);
if ((elems_shifted >> SAVE_TIGHT_SHIFT) != total_elems)
Perl_croak(aTHX_ "panic: paren_elems_to_push offset %"UVuf
SSGROW(total_elems + REGCP_FRAME_ELEMS);
- for (p = PL_regsize; p > parenfloor; p--) {
+ DEBUG_BUFFERS_r(
+ if ((int)PL_regsize > (int)parenfloor)
+ PerlIO_printf(Perl_debug_log,
+ "rex=0x%"UVxf" offs=0x%"UVxf": saving capture indices:\n",
+ PTR2UV(rex),
+ PTR2UV(rex->offs)
+ );
+ );
+ for (p = parenfloor+1; p <= (I32)PL_regsize; p++) {
/* REGCP_PARENS_ELEMS are pushed per pairs of parentheses. */
- SSPUSHINT(PL_regoffs[p].end);
- SSPUSHINT(PL_regoffs[p].start);
- SSPUSHPTR(PL_reg_start_tmp[p]);
- SSPUSHINT(p);
+ SSPUSHINT(rex->offs[p].end);
+ SSPUSHINT(rex->offs[p].start);
+ SSPUSHINT(rex->offs[p].start_tmp);
DEBUG_BUFFERS_r(PerlIO_printf(Perl_debug_log,
- " saving \\%"UVuf" %"IVdf"(%"IVdf")..%"IVdf"\n",
- (UV)p, (IV)PL_regoffs[p].start,
- (IV)(PL_reg_start_tmp[p] - PL_bostr),
- (IV)PL_regoffs[p].end
+ " \\%"UVuf": %"IVdf"(%"IVdf")..%"IVdf"\n",
+ (UV)p,
+ (IV)rex->offs[p].start,
+ (IV)rex->offs[p].start_tmp,
+ (IV)rex->offs[p].end
));
}
/* REGCP_OTHER_ELEMS are pushed in any case, parentheses or no. */
- SSPUSHPTR(PL_regoffs);
SSPUSHINT(PL_regsize);
- SSPUSHINT(*PL_reglastparen);
- SSPUSHINT(*PL_reglastcloseparen);
- SSPUSHPTR(PL_reginput);
+ SSPUSHINT(rex->lastparen);
+ SSPUSHINT(rex->lastcloseparen);
SSPUSHUV(SAVEt_REGCONTEXT | elems_shifted); /* Magic cookie. */
return retval;
(IV)(cp), (IV)PL_savestack_ix)); \
regcpblow(cp)
-STATIC char *
-S_regcppop(pTHX_ const regexp *rex)
+STATIC void
+S_regcppop(pTHX_ regexp *rex)
{
dVAR;
UV i;
- char *input;
+ U32 paren;
GET_RE_DEBUG_FLAGS_DECL;
PERL_ARGS_ASSERT_REGCPPOP;
i = SSPOPUV;
assert((i & SAVE_MASK) == SAVEt_REGCONTEXT); /* Check that the magic cookie is there. */
i >>= SAVE_TIGHT_SHIFT; /* Parentheses elements to pop. */
- input = (char *) SSPOPPTR;
- *PL_reglastcloseparen = SSPOPINT;
- *PL_reglastparen = SSPOPINT;
+ rex->lastcloseparen = SSPOPINT;
+ rex->lastparen = SSPOPINT;
PL_regsize = SSPOPINT;
- PL_regoffs=(regexp_paren_pair *) SSPOPPTR;
i -= REGCP_OTHER_ELEMS;
/* Now restore the parentheses context. */
+ DEBUG_BUFFERS_r(
+ if (i || rex->lastparen + 1 <= rex->nparens)
+ PerlIO_printf(Perl_debug_log,
+ "rex=0x%"UVxf" offs=0x%"UVxf": restoring capture indices to:\n",
+ PTR2UV(rex),
+ PTR2UV(rex->offs)
+ );
+ );
+ paren = PL_regsize;
for ( ; i > 0; i -= REGCP_PAREN_ELEMS) {
I32 tmps;
- U32 paren = (U32)SSPOPINT;
- PL_reg_start_tmp[paren] = (char *) SSPOPPTR;
- PL_regoffs[paren].start = SSPOPINT;
+ rex->offs[paren].start_tmp = SSPOPINT;
+ rex->offs[paren].start = SSPOPINT;
tmps = SSPOPINT;
- if (paren <= *PL_reglastparen)
- 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_regoffs[paren].start,
- (IV)(PL_reg_start_tmp[paren] - PL_bostr),
- (IV)PL_regoffs[paren].end,
- (paren > *PL_reglastparen ? "(no)" : ""));
+ if (paren <= rex->lastparen)
+ rex->offs[paren].end = tmps;
+ DEBUG_BUFFERS_r( PerlIO_printf(Perl_debug_log,
+ " \\%"UVuf": %"IVdf"(%"IVdf")..%"IVdf"%s\n",
+ (UV)paren,
+ (IV)rex->offs[paren].start,
+ (IV)rex->offs[paren].start_tmp,
+ (IV)rex->offs[paren].end,
+ (paren > rex->lastparen ? "(skipped)" : ""));
);
+ paren--;
}
- DEBUG_BUFFERS_r(
- if (*PL_reglastparen + 1 <= rex->nparens) {
- PerlIO_printf(Perl_debug_log,
- " restoring \\%"IVdf"..\\%"IVdf" to undef\n",
- (IV)(*PL_reglastparen + 1), (IV)rex->nparens);
- }
- );
#if 1
/* It would seem that the similar code in regtry()
* already takes care of this, and in fact it is in
* this code seems to be necessary or otherwise
* this erroneously leaves $1 defined: "1" =~ /^(?:(\d)x)?\d$/
* --jhi updated by dapm */
- for (i = *PL_reglastparen + 1; i <= rex->nparens; i++) {
+ for (i = rex->lastparen + 1; i <= rex->nparens; i++) {
if (i > PL_regsize)
- PL_regoffs[i].start = -1;
- PL_regoffs[i].end = -1;
+ rex->offs[i].start = -1;
+ rex->offs[i].end = -1;
+ DEBUG_BUFFERS_r( PerlIO_printf(Perl_debug_log,
+ " \\%"UVuf": %s ..-1 undeffing\n",
+ (UV)i,
+ (i > PL_regsize) ? "-1" : " "
+ ));
}
#endif
- return input;
}
#define regcpblow(cp) LEAVE_SCOPE(cp) /* Ignores regcppush()ed data. */
I32 ml_anch;
register char *other_last = NULL; /* other substr checked before this */
char *check_at = NULL; /* check substr found at this pos */
+ char *checked_upto = NULL; /* how far into the string we have already checked using find_byclass*/
const I32 multiline = prog->extflags & RXf_PMf_MULTILINE;
RXi_GET_DECL(prog,progi);
#ifdef DEBUGGING
GET_RE_DEBUG_FLAGS_DECL;
PERL_ARGS_ASSERT_RE_INTUIT_START;
+ PERL_UNUSED_ARG(flags);
+ PERL_UNUSED_ARG(data);
RX_MATCH_UTF8_set(rx,utf8_target);
{
I32 srch_start_shift = start_shift;
I32 srch_end_shift = end_shift;
+ U8* start_point;
+ U8* end_point;
if (srch_start_shift < 0 && strbeg - s > srch_start_shift) {
srch_end_shift -= ((strbeg - s) - srch_start_shift);
srch_start_shift = strbeg - s;
(IV)prog->check_end_shift);
});
- if ((flags & REXEC_SCREAM) && SvSCREAM(sv)) {
- I32 p = -1; /* Internal iterator of scream. */
- I32 * const pp = data ? data->scream_pos : &p;
- const MAGIC *mg;
- bool found = FALSE;
-
- assert(SvMAGICAL(sv));
- mg = mg_find(sv, PERL_MAGIC_study);
- assert(mg);
-
- if (mg->mg_private == 1) {
- found = ((U8 *)mg->mg_ptr)[BmRARE(check)] != (U8)~0;
- } else if (mg->mg_private == 2) {
- found = ((U16 *)mg->mg_ptr)[BmRARE(check)] != (U16)~0;
- } else {
- assert (mg->mg_private == 4);
- found = ((U32 *)mg->mg_ptr)[BmRARE(check)] != (U32)~0;
- }
-
- if (found
- || ( BmRARE(check) == '\n'
- && (BmPREVIOUS(check) == SvCUR(check) - 1)
- && SvTAIL(check) ))
- s = screaminstr(sv, check,
- srch_start_shift + (s - strbeg), srch_end_shift, pp, 0);
- else
- goto fail_finish;
- /* we may be pointing at the wrong string */
- if (s && RXp_MATCH_COPIED(prog))
- s = strbeg + (s - SvPVX_const(sv));
- if (data)
- *data->scream_olds = s;
- }
- else {
- U8* start_point;
- U8* end_point;
if (prog->extflags & RXf_CANY_SEEN) {
start_point= (U8*)(s + srch_start_shift);
end_point= (U8*)(strend - srch_end_shift);
s = fbm_instr( start_point, end_point,
check, multiline ? FBMrf_MULTILINE : 0);
}
- }
/* Update the count-of-usability, remove useless subpatterns,
unshift s. */
else
endpos= strend;
- DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "start_shift: %"IVdf" check_at: %"IVdf" s: %"IVdf" endpos: %"IVdf"\n",
- (IV)start_shift, (IV)(check_at - strbeg), (IV)(s - strbeg), (IV)(endpos - strbeg)));
-
+ if (checked_upto < s)
+ checked_upto = s;
+ DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "start_shift: %"IVdf" check_at: %"IVdf" s: %"IVdf" endpos: %"IVdf" checked_upto: %"IVdf"\n",
+ (IV)start_shift, (IV)(check_at - strbeg), (IV)(s - strbeg), (IV)(endpos - strbeg), (IV)(checked_upto- strbeg)));
+
t = s;
- s = find_byclass(prog, progi->regstclass, s, endpos, NULL);
- if (!s) {
+ s = find_byclass(prog, progi->regstclass, checked_upto, endpos, NULL);
+ if (s) {
+ checked_upto = s;
+ } else {
#ifdef DEBUGGING
const char *what = NULL;
#endif
"This position contradicts STCLASS...\n") );
if ((prog->extflags & RXf_ANCH) && !ml_anch)
goto fail;
+ checked_upto = HOPBACKc(endpos, start_shift);
+ DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "start_shift: %"IVdf" check_at: %"IVdf" endpos: %"IVdf" checked_upto: %"IVdf"\n",
+ (IV)start_shift, (IV)(check_at - strbeg), (IV)(endpos - strbeg), (IV)(checked_upto- strbeg)));
/* Contradict one of substrings */
if (prog->anchored_substr || prog->anchored_utf8) {
if ((utf8_target ? prog->anchored_utf8 : prog->anchored_substr) == check) {
#define DECL_TRIE_TYPE(scan) \
const enum { trie_plain, trie_utf8, trie_utf8_fold, trie_latin_utf8_fold } \
- trie_type = (scan->flags != EXACT) \
- ? (utf8_target ? trie_utf8_fold : (UTF_PATTERN ? trie_latin_utf8_fold : trie_plain)) \
- : (utf8_target ? trie_utf8 : trie_plain)
-
-#define REXEC_TRIE_READ_CHAR(trie_type, trie, widecharmap, uc, uscan, len, \
-uvc, charid, foldlen, foldbuf, uniflags) STMT_START { \
- switch (trie_type) { \
- case trie_utf8_fold: \
- if ( foldlen>0 ) { \
- uvc = utf8n_to_uvuni( uscan, UTF8_MAXLEN, &len, uniflags ); \
- foldlen -= len; \
- uscan += len; \
- len=0; \
- } else { \
- uvc = to_utf8_fold( (U8 *) uc, foldbuf, &foldlen ); \
- len = UTF8SKIP(uc); \
- foldlen -= UNISKIP( uvc ); \
- uscan = foldbuf + UNISKIP( uvc ); \
- } \
- break; \
- case trie_latin_utf8_fold: \
- if ( foldlen>0 ) { \
- uvc = utf8n_to_uvuni( uscan, UTF8_MAXLEN, &len, uniflags ); \
- foldlen -= len; \
- uscan += len; \
- len=0; \
- } else { \
- len = 1; \
- uvc = to_uni_fold( *(U8*)uc, 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 (widecharmap) { \
- SV** const svpp = hv_fetch(widecharmap, \
- (char*)&uvc, sizeof(UV), 0); \
- if (svpp) \
- charid = (U16)SvIV(*svpp); \
- } \
- } \
+ trie_type = ((scan->flags == EXACT) \
+ ? (utf8_target ? trie_utf8 : trie_plain) \
+ : (utf8_target ? trie_utf8_fold : trie_latin_utf8_fold))
+
+#define REXEC_TRIE_READ_CHAR(trie_type, trie, widecharmap, uc, uscan, len, \
+uvc, charid, foldlen, foldbuf, uniflags) STMT_START { \
+ STRLEN skiplen; \
+ switch (trie_type) { \
+ case trie_utf8_fold: \
+ if ( foldlen>0 ) { \
+ uvc = utf8n_to_uvuni( (const U8*) uscan, UTF8_MAXLEN, &len, uniflags ); \
+ foldlen -= len; \
+ uscan += len; \
+ len=0; \
+ } else { \
+ uvc = to_utf8_fold( (const U8*) uc, foldbuf, &foldlen ); \
+ len = UTF8SKIP(uc); \
+ skiplen = UNISKIP( uvc ); \
+ foldlen -= skiplen; \
+ uscan = foldbuf + skiplen; \
+ } \
+ break; \
+ case trie_latin_utf8_fold: \
+ if ( foldlen>0 ) { \
+ uvc = utf8n_to_uvuni( (const U8*) uscan, UTF8_MAXLEN, &len, uniflags ); \
+ foldlen -= len; \
+ uscan += len; \
+ len=0; \
+ } else { \
+ len = 1; \
+ uvc = _to_fold_latin1( (U8) *uc, foldbuf, &foldlen, 1); \
+ skiplen = UNISKIP( uvc ); \
+ foldlen -= skiplen; \
+ uscan = foldbuf + skiplen; \
+ } \
+ break; \
+ case trie_utf8: \
+ uvc = utf8n_to_uvuni( (const 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 (widecharmap) { \
+ SV** const svpp = hv_fetch(widecharmap, \
+ (char*)&uvc, sizeof(UV), 0); \
+ if (svpp) \
+ charid = (U16)SvIV(*svpp); \
+ } \
+ } \
} STMT_END
#define REXEC_FBC_EXACTISH_SCAN(CoNd) \
goto do_exactf_non_utf8; /* isn't dealt with by these */
case EXACTF:
- if (UTF_PATTERN || utf8_target) {
+ if (utf8_target) {
/* regcomp.c already folded this if pattern is in UTF-8 */
- utf8_fold_flags = (UTF_PATTERN) ? FOLDEQ_S2_ALREADY_FOLDED : 0;
+ utf8_fold_flags = 0;
goto do_exactf_utf8;
}
fold_array = PL_fold;
folder = foldEQ_locale;
goto do_exactf_non_utf8;
+ case EXACTFU_SS:
+ if (UTF_PATTERN) {
+ utf8_fold_flags = FOLDEQ_S2_ALREADY_FOLDED;
+ }
+ goto do_exactf_utf8;
+
+ case EXACTFU_TRICKYFOLD:
case EXACTFU:
if (UTF_PATTERN || utf8_target) {
utf8_fold_flags = (UTF_PATTERN) ? FOLDEQ_S2_ALREADY_FOLDED : 0;
/* FALL THROUGH */
- do_exactf_non_utf8: /* Neither pattern nor string are UTF8 */
+ do_exactf_non_utf8: /* Neither pattern nor string are UTF8, and there
+ are no glitches with fold-length differences
+ between the target string and pattern */
/* The idea in the non-utf8 EXACTF* cases is to first find the
* first character of the EXACTF* node and then, if necessary,
}
PL_reg_flags = 0;
- PL_reg_eval_set = 0;
+ PL_reg_state.re_state_eval_setup_done = FALSE;
PL_reg_maxiter = 0;
if (RX_UTF8(rx))
swap = prog->offs;
/* do we need a save destructor here for eval dies? */
Newxz(prog->offs, (prog->nparens + 1), regexp_paren_pair);
+ DEBUG_BUFFERS_r(PerlIO_printf(Perl_debug_log,
+ "rex=0x%"UVxf" saving offs: orig=0x%"UVxf" new=0x%"UVxf"\n",
+ PTR2UV(prog),
+ PTR2UV(swap),
+ PTR2UV(prog->offs)
+ ));
}
if (!(flags & REXEC_CHECKED) && (prog->check_substr != NULL || prog->check_utf8 != NULL)) {
re_scream_pos_data d;
/*XXX: The s-- is almost definitely wrong here under unicode - demeprhq*/
s--;
}
- /* We can use a more efficient search as newlines are the same in unicode as they are in latin */
- while (s < end) {
+ /* We can use a more efficient search as newlines are the same in unicode as they are in latin */
+ while (s <= end) { /* note it could be possible to match at the end of the string */
if (*s++ == '\n') { /* don't need PL_utf8skip here */
if (regtry(®info, &s))
goto got_it;
dontbother = end_shift;
strend = HOPc(strend, -dontbother);
while ( (s <= last) &&
- ((flags & REXEC_SCREAM) && SvSCREAM(sv)
- ? (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, (back_min<0 ? strbeg : strend)),
+ (s = fbm_instr((unsigned char*)HOP3(s, back_min, (back_min<0 ? strbeg : strend)),
(unsigned char*)strend, must,
- multiline ? FBMrf_MULTILINE : 0))) ) {
- /* we may be pointing at the wrong string */
- if ((flags & REXEC_SCREAM) && RXp_MATCH_COPIED(prog))
- s = strbeg + (s - SvPVX_const(sv));
+ multiline ? FBMrf_MULTILINE : 0)) ) {
DEBUG_EXECUTE_r( did_match = 1 );
if (HOPc(s, -back_max) > last1) {
last1 = HOPc(s, -back_min);
dontbother = 0;
if (prog->float_substr != NULL || prog->float_utf8 != NULL) {
/* Trim the end. */
- char *last;
+ char *last= NULL;
SV* float_real;
+ STRLEN len;
+ const char *little;
if (!(utf8_target ? prog->float_utf8 : prog->float_substr))
utf8_target ? to_utf8_substr(prog) : to_byte_substr(prog);
float_real = utf8_target ? prog->float_utf8 : prog->float_substr;
- if ((flags & REXEC_SCREAM) && SvSCREAM(sv)) {
- last = screaminstr(sv, float_real, s - strbeg,
- end_shift, &scream_pos, 1); /* last one */
- if (!last)
- last = scream_olds; /* Only one occurrence. */
- /* we may be pointing at the wrong string */
- else if (RXp_MATCH_COPIED(prog))
- s = strbeg + (s - SvPVX_const(sv));
- }
- else {
- STRLEN len;
- const char * const little = SvPV_const(float_real, len);
-
- if (SvTAIL(float_real)) {
- if (memEQ(strend - len + 1, little, len - 1))
- last = strend - len + 1;
- else if (!multiline)
- last = memEQ(strend - len, little, len)
- ? strend - len : NULL;
- else
+ little = SvPV_const(float_real, len);
+ if (SvTAIL(float_real)) {
+ /* This means that float_real contains an artificial \n on the end
+ * due to the presence of something like this: /foo$/
+ * where we can match both "foo" and "foo\n" at the end of the string.
+ * So we have to compare the end of the string first against the float_real
+ * without the \n and then against the full float_real with the string.
+ * We have to watch out for cases where the string might be smaller
+ * than the float_real or the float_real without the \n.
+ */
+ char *checkpos= strend - len;
+ DEBUG_OPTIMISE_r(
+ PerlIO_printf(Perl_debug_log,
+ "%sChecking for float_real.%s\n",
+ PL_colors[4], PL_colors[5]));
+ if (checkpos + 1 < strbeg) {
+ /* can't match, even if we remove the trailing \n string is too short to match */
+ DEBUG_EXECUTE_r(
+ PerlIO_printf(Perl_debug_log,
+ "%sString shorter than required trailing substring, cannot match.%s\n",
+ PL_colors[4], PL_colors[5]));
+ goto phooey;
+ } else if (memEQ(checkpos + 1, little, len - 1)) {
+ /* can match, the end of the string matches without the "\n" */
+ last = checkpos + 1;
+ } else if (checkpos < strbeg) {
+ /* cant match, string is too short when the "\n" is included */
+ DEBUG_EXECUTE_r(
+ PerlIO_printf(Perl_debug_log,
+ "%sString does not contain required trailing substring, cannot match.%s\n",
+ PL_colors[4], PL_colors[5]));
+ goto phooey;
+ } else if (!multiline) {
+ /* non multiline match, so compare with the "\n" at the end of the string */
+ if (memEQ(checkpos, little, len)) {
+ last= checkpos;
+ } else {
+ DEBUG_EXECUTE_r(
+ PerlIO_printf(Perl_debug_log,
+ "%sString does not contain required trailing substring, cannot match.%s\n",
+ PL_colors[4], PL_colors[5]));
+ goto phooey;
+ }
+ } else {
+ /* multiline match, so we have to search for a place where the full string is located */
goto find_last;
- } else {
+ }
+ } else {
find_last:
if (len)
last = rninstr(s, strend, little, little + len);
else
last = strend; /* matching "$" */
- }
}
- if (last == NULL) {
+ if (!last) {
+ /* at one point this block contained a comment which was probably
+ * incorrect, which said that this was a "should not happen" case.
+ * Even if it was true when it was written I am pretty sure it is
+ * not anymore, so I have removed the comment and replaced it with
+ * this one. Yves */
DEBUG_EXECUTE_r(
PerlIO_printf(Perl_debug_log,
- "%sCan't trim the tail, match fails (should not happen)%s\n",
- PL_colors[4], PL_colors[5]));
- goto phooey; /* Should not happen! */
+ "String does not contain required substring, cannot match.\n"
+ ));
+ goto phooey;
}
dontbother = strend - last + prog->float_min_offset;
}
goto phooey;
got_it:
+ DEBUG_BUFFERS_r(
+ if (swap)
+ PerlIO_printf(Perl_debug_log,
+ "rex=0x%"UVxf" freeing offs: 0x%"UVxf"\n",
+ PTR2UV(prog),
+ PTR2UV(swap)
+ );
+ );
Safefree(swap);
RX_MATCH_TAINTED_set(rx, PL_reg_flags & RF_tainted);
- if (PL_reg_eval_set)
+ if (PL_reg_state.re_state_eval_setup_done)
restore_pos(aTHX_ prog);
if (RXp_PAREN_NAMES(prog))
(void)hv_iterinit(RXp_PAREN_NAMES(prog));
phooey:
DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "%sMatch failed%s\n",
PL_colors[4], PL_colors[5]));
- if (PL_reg_eval_set)
+ if (PL_reg_state.re_state_eval_setup_done)
restore_pos(aTHX_ prog);
if (swap) {
/* we failed :-( roll it back */
+ DEBUG_BUFFERS_r(PerlIO_printf(Perl_debug_log,
+ "rex=0x%"UVxf" rolling back offs: freeing=0x%"UVxf" restoring=0x%"UVxf"\n",
+ PTR2UV(prog),
+ PTR2UV(prog->offs),
+ PTR2UV(swap)
+ ));
Safefree(prog->offs);
prog->offs = swap;
}
reginfo->cutpoint=NULL;
- if ((prog->extflags & RXf_EVAL_SEEN) && !PL_reg_eval_set) {
+ if ((prog->extflags & RXf_EVAL_SEEN)
+ && !PL_reg_state.re_state_eval_setup_done)
+ {
MAGIC *mg;
- PL_reg_eval_set = RS_init;
- DEBUG_EXECUTE_r(DEBUG_s(
- PerlIO_printf(Perl_debug_log, " setting stack tmpbase at %"IVdf"\n",
- (IV)(PL_stack_sp - PL_stack_base));
- ));
- SAVESTACK_CXPOS();
- cxstack[cxstack_ix].blk_oldsp = PL_stack_sp - PL_stack_base;
- /* Otherwise OP_NEXTSTATE will free whatever on stack now. */
- SAVETMPS;
- /* Apparently this is not needed, judging by wantarray. */
- /* SAVEI8(cxstack[cxstack_ix].blk_gimme);
- cxstack[cxstack_ix].blk_gimme = G_SCALAR; */
-
+ PL_reg_state.re_state_eval_setup_done = TRUE;
if (reginfo->sv) {
/* Make $_ available to executed code. */
if (reginfo->sv != DEFSV) {
prog->subbeg = PL_bostr;
prog->sublen = PL_regeol - PL_bostr; /* strend may have been modified */
}
- DEBUG_EXECUTE_r(PL_reg_starttry = *startpos);
+#ifdef DEBUGGING
+ PL_reg_starttry = *startpos;
+#endif
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_regoffs = prog->offs;
- if (PL_reg_start_tmpl <= prog->nparens) {
- PL_reg_start_tmpl = prog->nparens*3/2 + 3;
- if(PL_reg_start_tmp)
- Renew(PL_reg_start_tmp, PL_reg_start_tmpl, char*);
- else
- Newx(PL_reg_start_tmp, PL_reg_start_tmpl, char*);
- }
/* XXXX What this code is doing here?!!! There should be no need
- to do this again and again, PL_reglastparen should take care of
+ to do this again and again, prog->lastparen should take care of
this! --ilya*/
/* Tests pat.t#187 and split.t#{13,14} seem to depend on this code.
* Actually, the code in regcppop() (which Ilya may be meaning by
- * PL_reglastparen), is not needed at all by the test suite
+ * prog->lastparen), is not needed at all by the test suite
* (op/regexp, op/pat, op/split), but that code is needed otherwise
* this erroneously leaves $1 defined: "1" =~ /^(?:(\d)x)?\d$/
* Meanwhile, this code *is* needed for the
* --jhi updated by dapm */
#if 1
if (prog->nparens) {
- regexp_paren_pair *pp = PL_regoffs;
+ regexp_paren_pair *pp = prog->offs;
register I32 i;
- for (i = prog->nparens; i > (I32)*PL_reglastparen; i--) {
+ for (i = prog->nparens; i > (I32)prog->lastparen; i--) {
++pp;
pp->start = -1;
pp->end = -1;
#endif
REGCP_SET(lastcp);
if (regmatch(reginfo, progi->program + 1)) {
- PL_regoffs[0].end = PL_reginput - PL_bostr;
+ prog->offs[0].end = PL_reginput - PL_bostr;
return 1;
}
if (reginfo->cutpoint)
PERL_ARGS_ASSERT_REG_CHECK_NAMED_BUFF_MATCHED;
for ( n=0; n<SvIVX(sv_dat); n++ ) {
- if ((I32)*PL_reglastparen >= nums[n] &&
- PL_regoffs[nums[n]].end != -1)
+ if ((I32)rex->lastparen >= nums[n] &&
+ rex->offs[nums[n]].end != -1)
{
return nums[n];
}
#define SETREX(Re1,Re2) \
- if (PL_reg_eval_set) PM_SETRE((PL_reg_curpm), (Re2)); \
+ if (PL_reg_state.re_state_eval_setup_done) \
+ PM_SETRE((PL_reg_curpm), (Re2)); \
Re1 = (Re2)
STATIC I32 /* 0 failure, 1 success */
false: plain (?=foo)
true: used as a condition: (?(?=foo))
*/
+ PAD* last_pad = NULL;
+ dMULTICALL;
+ I32 gimme = G_SCALAR;
+ CV *caller_cv = NULL; /* who called us */
+ CV *last_pushed_cv = NULL; /* most recently called (?{}) CV */
+
#ifdef DEBUGGING
GET_RE_DEBUG_FLAGS_DECL;
#endif
+ /* shut up 'may be used uninitialized' compiler warnings for dMULTICALL */
+ multicall_oldcatch = 0;
+ multicall_cv = NULL;
+ cx = NULL;
+
+
PERL_ARGS_ASSERT_REGMATCH;
DEBUG_OPTIMISE_r( DEBUG_EXECUTE_r({
reenter_switch:
- assert(PL_reglastparen == &rex->lastparen);
- assert(PL_reglastcloseparen == &rex->lastcloseparen);
- assert(PL_regoffs == rex->offs);
-
switch (state_num) {
case BOL:
if (locinput == PL_bostr)
case KEEPS:
/* update the startpoint */
- st->u.keeper.val = PL_regoffs[0].start;
+ st->u.keeper.val = rex->offs[0].start;
PL_reginput = locinput;
- PL_regoffs[0].start = locinput - PL_bostr;
+ rex->offs[0].start = locinput - PL_bostr;
PUSH_STATE_GOTO(KEEPS_next, next);
/*NOT-REACHED*/
case KEEPS_next_fail:
/* rollback the start point change */
- PL_regoffs[0].start = st->u.keeper.val;
+ rex->offs[0].start = st->u.keeper.val;
sayNO_SILENT;
/*NOT-REACHED*/
case EOL:
/* In this case the charclass data is available inline so
we can fail fast without a lot of extra overhead.
*/
- if (scan->flags == EXACT || !utf8_target) {
- 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 */
- }
+ 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:
HV * widecharmap = MUTABLE_HV(rexi->data->data[ ARG( scan ) + 1 ]);
U32 state = trie->startstate;
- if (trie->bitmap && trie_type != trie_utf8_fold &&
- !TRIE_BITMAP_TEST(trie,*locinput)
- ) {
+ if (trie->bitmap && !TRIE_BITMAP_TEST(trie,*locinput) ) {
if (trie->states[ state ].wordnum) {
DEBUG_EXECUTE_r(
PerlIO_printf(Perl_debug_log,
U32 charcount = 0; /* how many input chars we have matched */
U32 accepted = 0; /* have we seen any accepting states? */
- ST.B = next;
ST.jump = trie->jump;
ST.me = scan;
ST.firstpos = NULL;
case TRIE_next_fail: /* we failed - try next alternative */
if ( ST.jump) {
REGCP_UNWIND(ST.cp);
- for (n = *PL_reglastparen; n > ST.lastparen; n--)
- PL_regoffs[n].end = -1;
- *PL_reglastparen = n;
+ for (n = rex->lastparen; n > ST.lastparen; n--)
+ rex->offs[n].end = -1;
+ rex->lastparen = n;
+ rex->lastcloseparen = ST.lastcloseparen;
}
if (!--ST.accepted) {
DEBUG_EXECUTE_r({
}
if ( ST.jump) {
- ST.lastparen = *PL_reglastparen;
+ ST.lastparen = rex->lastparen;
+ ST.lastcloseparen = rex->lastcloseparen;
REGCP_SET(ST.cp);
}
PL_reginput = (char *)uc;
}
- scan = (ST.jump && ST.jump[ST.nextword])
- ? ST.me + ST.jump[ST.nextword]
- : ST.B;
+ scan = ST.me + ((ST.jump && ST.jump[ST.nextword])
+ ? ST.jump[ST.nextword]
+ : NEXT_OFF(ST.me));
DEBUG_EXECUTE_r({
PerlIO_printf( Perl_debug_log,
fold_utf8_flags = FOLDEQ_UTF8_LOCALE;
goto do_exactf;
+ case EXACTFU_SS:
+ case EXACTFU_TRICKYFOLD:
case EXACTFU:
folder = foldEQ_latin1;
fold_array = PL_fold_latin1;
case EXACTF:
folder = foldEQ;
fold_array = PL_fold;
- fold_utf8_flags = (UTF_PATTERN) ? FOLDEQ_S1_ALREADY_FOLDED : 0;
+ fold_utf8_flags = 0;
do_exactf:
s = STRING(scan);
ln = STR_LEN(scan);
- if (utf8_target || UTF_PATTERN) {
- /* Either target or the pattern are utf8. */
+ if (utf8_target || UTF_PATTERN || state_num == EXACTFU_SS) {
+ /* Either target or the pattern are utf8, or has the issue where
+ * the fold lengths may differ. */
const char * const l = locinput;
char *e = PL_regeol;
if (! foldEQ_utf8_flags(s, 0, ln, cBOOL(UTF_PATTERN),
- l, &e, 0, utf8_target, fold_utf8_flags))
+ l, &e, 0, utf8_target, fold_utf8_flags))
{
sayNO;
}
n = ARG(scan); /* which paren pair */
do_nref_ref_common:
- ln = PL_regoffs[n].start;
+ ln = rex->offs[n].start;
PL_reg_leftiter = PL_reg_maxiter; /* Void cache */
- if (*PL_reglastparen < n || ln == -1)
+ if (rex->lastparen < n || ln == -1)
sayNO; /* Do not match unless seen CLOSEn. */
- if (ln == PL_regoffs[n].end)
+ if (ln == rex->offs[n].end)
break;
s = PL_bostr + ln;
* not going off the end given by PL_regeol, and returns in
* limit upon success, how much of the current input was
* matched */
- if (! foldEQ_utf8_flags(s, NULL, PL_regoffs[n].end - ln, utf8_target,
+ if (! foldEQ_utf8_flags(s, NULL, rex->offs[n].end - ln, utf8_target,
locinput, &limit, 0, utf8_target, utf8_fold_flags))
{
sayNO;
(type == REF ||
UCHARAT(s) != fold_array[nextchr]))
sayNO;
- ln = PL_regoffs[n].end - ln;
+ ln = rex->offs[n].end - ln;
if (locinput + ln > PL_regeol)
sayNO;
if (ln > 1 && (type == REF
}
{
/* execute the code in the {...} */
+
dSP;
- SV ** const before = SP;
- OP_4tree * const oop = PL_op;
+ SV ** before;
+ OP * const oop = PL_op;
COP * const ocurcop = PL_curcop;
- PAD *old_comppad;
+ OP *nop;
char *saved_regeol = PL_regeol;
struct re_save_state saved_state;
+ CV *newcv;
/* To not corrupt the existing regex state while executing the
* eval we would normally put it on the save stack, like with
*/
Copy(&PL_reg_state, &saved_state, 1, struct re_save_state);
+ PL_reg_state.re_reparsing = FALSE;
+
+ if (!caller_cv)
+ caller_cv = find_runcv(NULL);
+
n = ARG(scan);
- PL_op = (OP_4tree*)rexi->data->data[n];
+
+ if (rexi->data->what[n] == 'r') { /* code from an external qr */
+ newcv = ((struct regexp *)SvANY(
+ (REGEXP*)(rexi->data->data[n])
+ ))->qr_anoncv
+ ;
+ nop = (OP*)rexi->data->data[n+1];
+ }
+ else if (rexi->data->what[n] == 'l') { /* literal code */
+ newcv = caller_cv;
+ nop = (OP*)rexi->data->data[n];
+ assert(CvDEPTH(newcv));
+ }
+ else {
+ /* literal with own CV */
+ assert(rexi->data->what[n] == 'L');
+ newcv = rex->qr_anoncv;
+ nop = (OP*)rexi->data->data[n];
+ }
+
+ /* the initial nextstate you would normally execute
+ * at the start of an eval (which would cause error
+ * messages to come from the eval), may be optimised
+ * away from the execution path in the regex code blocks;
+ * so manually set PL_curcop to it initially */
+ {
+ OP *o = cUNOPx(nop)->op_first;
+ assert(o->op_type == OP_NULL);
+ if (o->op_targ == OP_SCOPE) {
+ o = cUNOPo->op_first;
+ }
+ else {
+ assert(o->op_targ == OP_LEAVE);
+ o = cUNOPo->op_first;
+ assert(o->op_type == OP_ENTER);
+ o = o->op_sibling;
+ }
+
+ if (o->op_type != OP_STUB) {
+ assert( o->op_type == OP_NEXTSTATE
+ || o->op_type == OP_DBSTATE
+ || (o->op_type == OP_NULL
+ && ( o->op_targ == OP_NEXTSTATE
+ || o->op_targ == OP_DBSTATE
+ )
+ )
+ );
+ PL_curcop = (COP*)o;
+ }
+ }
+ nop = nop->op_next;
+
DEBUG_STATE_r( PerlIO_printf(Perl_debug_log,
- " re_eval 0x%"UVxf"\n", PTR2UV(PL_op)) );
- /* wrap the call in two SAVECOMPPADs. This ensures that
- * when the save stack is eventually unwound, all the
- * accumulated SAVEt_CLEARSV's will be processed with
- * interspersed SAVEt_COMPPAD's to ensure that lexicals
- * are cleared in the right pad */
- SAVECOMPPAD();
- PAD_SAVE_LOCAL(old_comppad, (PAD*)rexi->data->data[n + 2]);
- PL_regoffs[0].end = PL_reg_magic->mg_len = locinput - PL_bostr;
+ " re EVAL PL_op=0x%"UVxf"\n", PTR2UV(nop)) );
+
+ /* normally if we're about to execute code from the same
+ * CV that we used previously, we just use the existing
+ * CX stack entry. However, its possible that in the
+ * meantime we may have backtracked, popped from the save
+ * stack, and undone the SAVECOMPPAD(s) associated with
+ * PUSH_MULTICALL; in which case PL_comppad no longer
+ * points to newcv's pad. */
+ if (newcv != last_pushed_cv || PL_comppad != last_pad)
+ {
+ I32 depth = (newcv == caller_cv) ? 0 : 1;
+ if (last_pushed_cv) {
+ CHANGE_MULTICALL_WITHDEPTH(newcv, depth);
+ }
+ else {
+ PUSH_MULTICALL_WITHDEPTH(newcv, depth);
+ }
+ last_pushed_cv = newcv;
+ }
+ last_pad = PL_comppad;
+
+ rex->offs[0].end = PL_reg_magic->mg_len = locinput - PL_bostr;
if (sv_yes_mark) {
SV *sv_mrk = get_sv("REGMARK", 1);
sv_setsv(sv_mrk, sv_yes_mark);
}
+ /* we don't use MULTICALL here as we want to call the
+ * first op of the block of interest, rather than the
+ * first op of the sub */
+ before = SP;
+ PL_op = nop;
CALLRUNOPS(aTHX); /* Scalar context. */
SPAGAIN;
if (SP == before)
Copy(&saved_state, &PL_reg_state, 1, struct re_save_state);
+ /* *** Note that at this point we don't restore
+ * PL_comppad, (or pop the CxSUB) on the assumption it may
+ * be used again soon. This is safe as long as nothing
+ * in the regexp code uses the pad ! */
PL_op = oop;
- SAVECOMPPAD();
- PAD_RESTORE_LOCAL(old_comppad);
PL_curcop = ocurcop;
PL_regeol = saved_regeol;
if (!logical) {
);
startpoint = rei->program + 1;
ST.close_paren = 0; /* only used for GOSUB */
- /* borrowed from regtry */
- if (PL_reg_start_tmpl <= re->nparens) {
- PL_reg_start_tmpl = re->nparens*3/2 + 3;
- if(PL_reg_start_tmp)
- Renew(PL_reg_start_tmp, PL_reg_start_tmpl, char*);
- else
- Newx(PL_reg_start_tmp, PL_reg_start_tmpl, char*);
- }
eval_recurse_doit: /* Share code with GOSUB below this line */
/* run the pattern returned from (??{...}) */
- ST.cp = regcppush(0); /* Save *all* the positions. */
+ ST.cp = regcppush(rex, 0); /* Save *all* the positions. */
REGCP_SET(ST.lastcp);
- PL_regoffs = re->offs; /* essentially NOOP on GOSUB */
-
- /* see regtry, specifically PL_reglast(?:close)?paren is a pointer! (i dont know why) :dmq */
- PL_reglastparen = &re->lastparen;
- PL_reglastcloseparen = &re->lastcloseparen;
re->lastparen = 0;
re->lastcloseparen = 0;
cur_eval = ST.prev_eval;
cur_curlyx = ST.prev_curlyx;
- /* rex was changed so update the pointer in PL_reglastparen and PL_reglastcloseparen */
- PL_reglastparen = &rex->lastparen;
- PL_reglastcloseparen = &rex->lastcloseparen;
- /* also update PL_regoffs */
- PL_regoffs = rex->offs;
-
/* XXXX This is too dramatic a measure... */
PL_reg_maxiter = 0;
if ( nochange_depth )
SETREX(rex_sv,ST.prev_rex);
rex = (struct regexp *)SvANY(rex_sv);
rexi = RXi_GET(rex);
- /* rex was changed so update the pointer in PL_reglastparen and PL_reglastcloseparen */
- PL_reglastparen = &rex->lastparen;
- PL_reglastcloseparen = &rex->lastcloseparen;
PL_reginput = locinput;
REGCP_UNWIND(ST.lastcp);
case OPEN:
n = ARG(scan); /* which paren pair */
- PL_reg_start_tmp[n] = locinput;
+ rex->offs[n].start_tmp = locinput - PL_bostr;
if (n > PL_regsize)
PL_regsize = n;
+ DEBUG_BUFFERS_r(PerlIO_printf(Perl_debug_log,
+ "rex=0x%"UVxf" offs=0x%"UVxf": \\%"UVuf": set %"IVdf" tmp; regsize=%"UVuf"\n",
+ PTR2UV(rex),
+ PTR2UV(rex->offs),
+ (UV)n,
+ (IV)rex->offs[n].start_tmp,
+ (UV)PL_regsize
+ ));
lastopen = n;
break;
+
+/* XXX really need to log other places start/end are set too */
+#define CLOSE_CAPTURE \
+ rex->offs[n].start = rex->offs[n].start_tmp; \
+ rex->offs[n].end = locinput - PL_bostr; \
+ DEBUG_BUFFERS_r(PerlIO_printf(Perl_debug_log, \
+ "rex=0x%"UVxf" offs=0x%"UVxf": \\%"UVuf": set %"IVdf"..%"IVdf"\n", \
+ PTR2UV(rex), \
+ PTR2UV(rex->offs), \
+ (UV)n, \
+ (IV)rex->offs[n].start, \
+ (IV)rex->offs[n].end \
+ ))
+
case CLOSE:
n = ARG(scan); /* which paren pair */
- PL_regoffs[n].start = PL_reg_start_tmp[n] - PL_bostr;
- PL_regoffs[n].end = locinput - PL_bostr;
+ CLOSE_CAPTURE;
/*if (n > PL_regsize)
PL_regsize = n;*/
- if (n > *PL_reglastparen)
- *PL_reglastparen = n;
- *PL_reglastcloseparen = n;
+ if (n > rex->lastparen)
+ rex->lastparen = n;
+ rex->lastcloseparen = n;
if (cur_eval && cur_eval->u.eval.close_paren == n) {
goto fake_end;
}
if ( OP(cursor)==CLOSE ){
n = ARG(cursor);
if ( n <= lastopen ) {
- PL_regoffs[n].start
- = PL_reg_start_tmp[n] - PL_bostr;
- PL_regoffs[n].end = locinput - PL_bostr;
+ CLOSE_CAPTURE;
/*if (n > PL_regsize)
PL_regsize = n;*/
- if (n > *PL_reglastparen)
- *PL_reglastparen = n;
- *PL_reglastcloseparen = n;
+ if (n > rex->lastparen)
+ rex->lastparen = n;
+ rex->lastcloseparen = n;
if ( n == ARG(scan) || (cur_eval &&
cur_eval->u.eval.close_paren == n))
break;
/*NOTREACHED*/
case GROUPP:
n = ARG(scan); /* which paren pair */
- sw = cBOOL(*PL_reglastparen >= n && PL_regoffs[n].end != -1);
+ sw = cBOOL(rex->lastparen >= n && rex->offs[n].end != -1);
break;
case NGROUPP:
/* reg_check_named_buff_matched returns 0 for no match */
/* XXXX Probably it is better to teach regpush to support
parenfloor > PL_regsize... */
- if (parenfloor > (I32)*PL_reglastparen)
- parenfloor = *PL_reglastparen; /* Pessimization... */
+ if (parenfloor > (I32)rex->lastparen)
+ parenfloor = rex->lastparen; /* Pessimization... */
ST.prev_curlyx= cur_curlyx;
cur_curlyx = st;
/* First just match a string of min A's. */
if (n < min) {
- ST.cp = regcppush(cur_curlyx->u.curlyx.parenfloor);
+ ST.cp = regcppush(rex, cur_curlyx->u.curlyx.parenfloor);
cur_curlyx->u.curlyx.lastloc = locinput;
REGCP_SET(ST.lastcp);
if (cur_curlyx->u.curlyx.minmod) {
ST.save_curlyx = cur_curlyx;
cur_curlyx = cur_curlyx->u.curlyx.prev_curlyx;
- ST.cp = regcppush(ST.save_curlyx->u.curlyx.parenfloor);
+ ST.cp = regcppush(rex, ST.save_curlyx->u.curlyx.parenfloor);
REGCP_SET(ST.lastcp);
PUSH_YES_STATE_GOTO(WHILEM_B_min, ST.save_curlyx->u.curlyx.B);
/* NOTREACHED */
/* Prefer A over B for maximal matching. */
if (n < max) { /* More greed allowed? */
- ST.cp = regcppush(cur_curlyx->u.curlyx.parenfloor);
+ ST.cp = regcppush(rex, cur_curlyx->u.curlyx.parenfloor);
cur_curlyx->u.curlyx.lastloc = locinput;
REGCP_SET(ST.lastcp);
PUSH_STATE_GOTO(WHILEM_A_max, A);
/* Try grabbing another A and see if it helps. */
PL_reginput = locinput;
cur_curlyx->u.curlyx.lastloc = locinput;
- ST.cp = regcppush(cur_curlyx->u.curlyx.parenfloor);
+ ST.cp = regcppush(rex, cur_curlyx->u.curlyx.parenfloor);
REGCP_SET(ST.lastcp);
PUSH_STATE_GOTO(WHILEM_A_min,
/*A*/ NEXTOPER(ST.save_curlyx->u.curlyx.me) + EXTRA_STEP_2ARGS);
case BRANCH: /* /(...|A|...)/ */
scan = NEXTOPER(scan); /* scan now points to inner node */
- ST.lastparen = *PL_reglastparen;
+ ST.lastparen = rex->lastparen;
+ ST.lastcloseparen = rex->lastcloseparen;
ST.next_branch = next;
REGCP_SET(ST.cp);
PL_reginput = locinput;
no_final = 0;
}
REGCP_UNWIND(ST.cp);
- for (n = *PL_reglastparen; n > ST.lastparen; n--)
- PL_regoffs[n].end = -1;
- *PL_reglastparen = n;
- /*dmq: *PL_reglastcloseparen = n; */
+ for (n = rex->lastparen; n > ST.lastparen; n--)
+ rex->offs[n].end = -1;
+ rex->lastparen = n;
+ rex->lastcloseparen = ST.lastcloseparen;
scan = ST.next_branch;
/* no more branches? */
if (!scan || (OP(scan) != BRANCH && OP(scan) != BRANCHJ)) {
ST.me = scan;
scan = NEXTOPER(scan) + NODE_STEP_REGNODE;
+ ST.lastparen = rex->lastparen;
+ ST.lastcloseparen = rex->lastcloseparen;
+
/* if paren positive, emulate an OPEN/CLOSE around A */
if (ST.me->flags) {
U32 paren = ST.me->flags;
if (paren > PL_regsize)
PL_regsize = paren;
- if (paren > *PL_reglastparen)
- *PL_reglastparen = paren;
scan += NEXT_OFF(scan); /* Skip former OPEN. */
}
ST.A = scan;
switch (OP(text_node)) {
case EXACTF: ST.c2 = PL_fold[ST.c1]; break;
case EXACTFA:
+ case EXACTFU_SS:
+ case EXACTFU_TRICKYFOLD:
case EXACTFU: ST.c2 = PL_fold_latin1[ST.c1]; break;
case EXACTFL: ST.c2 = PL_fold_locale[ST.c1]; break;
default: ST.c2 = ST.c1;
}
if (ST.me->flags) {
- /* mark current A as captured */
+ /* emulate CLOSE: mark current A as captured */
I32 paren = ST.me->flags;
if (ST.count) {
- PL_regoffs[paren].start
+ rex->offs[paren].start
= HOPc(PL_reginput, -ST.alen) - PL_bostr;
- PL_regoffs[paren].end = PL_reginput - PL_bostr;
- /*dmq: *PL_reglastcloseparen = paren; */
+ rex->offs[paren].end = PL_reginput - PL_bostr;
+ if ((U32)paren > rex->lastparen)
+ rex->lastparen = paren;
+ rex->lastcloseparen = paren;
}
else
- PL_regoffs[paren].end = -1;
+ rex->offs[paren].end = -1;
if (cur_eval && cur_eval->u.eval.close_paren &&
cur_eval->u.eval.close_paren == (U32)ST.me->flags)
{
case CURLYM_B_fail: /* just failed to match a B */
REGCP_UNWIND(ST.cp);
+ rex->lastparen = ST.lastparen;
+ rex->lastcloseparen = ST.lastcloseparen;
if (ST.minmod) {
I32 max = ARG2(ST.me);
if (max != REG_INFTY && ST.count == max)
#define CURLY_SETPAREN(paren, success) \
if (paren) { \
if (success) { \
- PL_regoffs[paren].start = HOPc(locinput, -1) - PL_bostr; \
- PL_regoffs[paren].end = locinput - PL_bostr; \
- *PL_reglastcloseparen = paren; \
+ rex->offs[paren].start = HOPc(locinput, -1) - PL_bostr; \
+ rex->offs[paren].end = locinput - PL_bostr; \
+ if (paren > rex->lastparen) \
+ rex->lastparen = paren; \
+ rex->lastcloseparen = paren; \
+ } \
+ else { \
+ rex->offs[paren].end = -1; \
+ rex->lastparen = ST.lastparen; \
+ rex->lastcloseparen = ST.lastcloseparen; \
} \
- else \
- PL_regoffs[paren].end = -1; \
}
case STAR: /* /A*B/ where A is width 1 */
goto repeat;
case CURLYN: /* /(A){m,n}B/ where A is width 1 */
ST.paren = scan->flags; /* Which paren to set */
+ ST.lastparen = rex->lastparen;
+ ST.lastcloseparen = rex->lastcloseparen;
if (ST.paren > PL_regsize)
PL_regsize = ST.paren;
- if (ST.paren > *PL_reglastparen)
- *PL_reglastparen = ST.paren;
ST.min = ARG1(scan); /* min to match */
ST.max = ARG2(scan); /* max to match */
if (cur_eval && cur_eval->u.eval.close_paren &&
switch (OP(text_node)) {
case EXACTF: ST.c2 = PL_fold[ST.c1]; break;
case EXACTFA:
+ case EXACTFU_SS:
+ case EXACTFU_TRICKYFOLD:
case EXACTFU: ST.c2 = PL_fold_latin1[ST.c1]; break;
case EXACTFL: ST.c2 = PL_fold_locale[ST.c1]; break;
default: ST.c2 = ST.c1; break;
}
else { /* UTF_PATTERN */
if (IS_TEXTFU(text_node) || IS_TEXTF(text_node)) {
- STRLEN ulen1, ulen2;
- U8 tmpbuf1[UTF8_MAXBYTES_CASE+1];
- U8 tmpbuf2[UTF8_MAXBYTES_CASE+1];
+ STRLEN ulen;
+ U8 tmpbuf[UTF8_MAXBYTES_CASE+1];
- to_utf8_lower((U8*)s, tmpbuf1, &ulen1);
- to_utf8_upper((U8*)s, tmpbuf2, &ulen2);
-#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,
+ to_utf8_fold((U8*)s, tmpbuf, &ulen);
+ ST.c1 = ST.c2 = utf8n_to_uvchr(tmpbuf, UTF8_MAXLEN, 0,
uniflags);
- ST.c2 = utf8n_to_uvuni(tmpbuf2, UTF8_MAXBYTES, 0,
- uniflags);
-#endif
}
else {
ST.c2 = ST.c1 = utf8n_to_uvchr(s, UTF8_MAXBYTES, 0,
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_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_regoffs[ST.paren].end = -1;
REGCP_UNWIND(ST.cp);
/* failed -- move forward one */
/* FALL THROUGH */
case CURLY_B_max_fail:
/* failed to find B in a greedy match */
- if (ST.paren && ST.count)
- PL_regoffs[ST.paren].end = -1;
REGCP_UNWIND(ST.cp);
/* back up. */
PL_reg_flags ^= st->u.eval.toggle_reg_flags;
st->u.eval.prev_rex = rex_sv; /* inner */
+ st->u.eval.cp = regcppush(rex, 0); /* Save *all* the positions. */
SETREX(rex_sv,cur_eval->u.eval.prev_rex);
rex = (struct regexp *)SvANY(rex_sv);
rexi = RXi_GET(rex);
cur_curlyx = cur_eval->u.eval.prev_curlyx;
(void)ReREFCNT_inc(rex_sv);
- st->u.eval.cp = regcppush(0); /* Save *all* the positions. */
-
- /* rex was changed so update the pointer in PL_reglastparen and PL_reglastcloseparen */
- PL_reglastparen = &rex->lastparen;
- PL_reglastcloseparen = &rex->lastcloseparen;
REGCP_SET(st->u.eval.lastcp);
PL_reginput = locinput;
sayNO;
/* NOTREACHED */
#undef ST
- case FOLDCHAR:
- n = ARG(scan);
- if ( n == (U32)what_len_TRICKYFOLD(locinput,utf8_target,ln) ) {
- locinput += ln;
- } else if ( LATIN_SMALL_LETTER_SHARP_S == n && !utf8_target && !UTF_PATTERN ) {
- 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 (! foldEQ_utf8((const char*) folded, 0, foldlen, 1,
- l, &e, 0, utf8_target)) {
- sayNO;
- }
- locinput = e;
- }
- nextchr = UCHARAT(locinput);
- break;
case LNBREAK:
if ((n=is_LNBREAK(locinput,utf8_target))) {
locinput += n;
DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "%sMatch successful!%s\n",
PL_colors[4], PL_colors[5]));
- if (PL_reg_eval_set) {
+ if (PL_reg_state.re_state_eval_setup_done) {
/* each successfully executed (?{...}) block does the equivalent of
* local $^R = do {...}
* When popping the save stack, all these locals would be undone;
sv_setsv(sv_mrk, sv_yes_mark);
}
+
+ if (last_pushed_cv) {
+ dSP;
+ POP_MULTICALL;
+ }
+
/* clean up; in particular, free all slabs above current one */
LEAVE_SCOPE(oldsave);
goto do_exactf;
case EXACTF:
+ utf8_flags = 0;
+ goto do_exactf;
+
+ case EXACTFU_SS:
+ case EXACTFU_TRICKYFOLD:
case EXACTFU:
utf8_flags = (UTF_PATTERN) ? FOLDEQ_S2_ALREADY_FOLDED : 0;
c = (U8)*STRING(p);
assert(! UTF_PATTERN || UNI_IS_INVARIANT(c));
- if (utf8_target) { /* Use full Unicode fold matching */
+ if (utf8_target || OP(p) == EXACTFU_SS) { /* Use full Unicode fold matching */
char *tmpeol = loceol;
while (hardcount < max
&& foldEQ_utf8_flags(scan, &tmpeol, 0, utf8_target,
switch (OP(p)) {
case EXACTF: folded = PL_fold[c]; break;
case EXACTFA:
+ case EXACTFU_TRICKYFOLD:
case EXACTFU: folded = PL_fold_latin1[c]; break;
case EXACTFL: folded = PL_fold_locale[c]; break;
default: Perl_croak(aTHX_ "panic: Unexpected op %u", OP(p));
#if !defined(PERL_IN_XSUB_RE) || defined(PLUGGABLE_RE_EXTENSION)
/*
-- regclass_swash - prepare the utf8 swash
-*/
-
+- regclass_swash - prepare the utf8 swash. Wraps the shared core version to
+create a copy so that changes the caller makes won't change the shared one
+ */
SV *
Perl_regclass_swash(pTHX_ const regexp *prog, register const regnode* node, bool doinit, SV** listsvp, SV **altsvp)
{
+ PERL_ARGS_ASSERT_REGCLASS_SWASH;
+ return newSVsv(core_regclass_swash(prog, node, doinit, listsvp, altsvp));
+}
+#endif
+
+STATIC SV *
+S_core_regclass_swash(pTHX_ const regexp *prog, register const regnode* node, bool doinit, SV** listsvp, SV **altsvp)
+{
/* Returns the swash for the input 'node' in the regex 'prog'.
* If <doinit> is true, will attempt to create the swash if not already
* done.
SV *sw = NULL;
SV *si = NULL;
SV *alt = NULL;
+ SV* invlist = NULL;
+
RXi_GET_DECL(prog,progi);
const struct reg_data * const data = prog ? progi->data : NULL;
- PERL_ARGS_ASSERT_REGCLASS_SWASH;
+ PERL_ARGS_ASSERT_CORE_REGCLASS_SWASH;
assert(ANYOF_NONBITMAP(node));
SV * const rv = MUTABLE_SV(data->data[n]);
AV * const av = MUTABLE_AV(SvRV(rv));
SV **const ary = AvARRAY(av);
+ bool invlist_has_user_defined_property;
- /* See the end of regcomp.c:S_regclass() for
- * documentation of these array elements. */
-
si = *ary; /* ary[0] = the string to initialize the swash with */
+ /* Elements 3 and 4 are either both present or both absent. [3] is
+ * any inversion list generated at compile time; [4] indicates if
+ * that inversion list has any user-defined properties in it. */
+ if (av_len(av) >= 3) {
+ invlist = ary[3];
+ invlist_has_user_defined_property = cBOOL(SvUV(ary[4]));
+ }
+ else {
+ invlist = NULL;
+ invlist_has_user_defined_property = FALSE;
+ }
+
/* Element [1] is reserved for the set-up swash. If already there,
* return it; if not, create it and store it there */
if (SvROK(ary[1])) {
sw = ary[1];
}
else if (si && doinit) {
- sw = swash_init("utf8", "", si, 1, 0);
+
+ sw = _core_swash_init("utf8", /* the utf8 package */
+ "", /* nameless */
+ si,
+ 1, /* binary */
+ 0, /* not from tr/// */
+ FALSE, /* is error if can't find
+ property */
+ invlist,
+ invlist_has_user_defined_property);
(void)av_store(av, 1, sw);
}
}
}
- if (listsvp)
- *listsvp = si;
+ if (listsvp) {
+ SV* matches_string = newSVpvn("", 0);
+ SV** invlistsvp;
+
+ /* Use the swash, if any, which has to have incorporated into it all
+ * possibilities */
+ if ( sw
+ && SvROK(sw)
+ && SvTYPE(SvRV(sw)) == SVt_PVHV
+ && (invlistsvp = hv_fetchs(MUTABLE_HV(SvRV(sw)), "INVLIST", FALSE)))
+ {
+ invlist = *invlistsvp;
+ }
+ else if (si && si != &PL_sv_undef) {
+
+ /* If no swash, use the input nitialization string, if available */
+ sv_catsv(matches_string, si);
+ }
+
+ /* Add the inversion list to whatever we have. This may have come from
+ * the swash, or from an input parameter */
+ if (invlist) {
+ sv_catsv(matches_string, _invlist_contents(invlist));
+ }
+ *listsvp = matches_string;
+ }
+
if (altsvp)
*altsvp = alt;
return sw;
}
-#endif
/*
- reginclass - determine if a character falls into a character class
(ANYOF_CLASS_TEST(n, ANYOF_NALNUMC) && !isALNUMC_LC(c)) ||
(ANYOF_CLASS_TEST(n, ANYOF_ALPHA) && isALPHA_LC(c)) ||
(ANYOF_CLASS_TEST(n, ANYOF_NALPHA) && !isALPHA_LC(c)) ||
- (ANYOF_CLASS_TEST(n, ANYOF_ASCII) && isASCII(c)) ||
- (ANYOF_CLASS_TEST(n, ANYOF_NASCII) && !isASCII(c)) ||
+ (ANYOF_CLASS_TEST(n, ANYOF_ASCII) && isASCII_LC(c)) ||
+ (ANYOF_CLASS_TEST(n, ANYOF_NASCII) && !isASCII_LC(c)) ||
(ANYOF_CLASS_TEST(n, ANYOF_CNTRL) && isCNTRL_LC(c)) ||
(ANYOF_CLASS_TEST(n, ANYOF_NCNTRL) && !isCNTRL_LC(c)) ||
(ANYOF_CLASS_TEST(n, ANYOF_GRAPH) && isGRAPH_LC(c)) ||
(ANYOF_CLASS_TEST(n, ANYOF_NXDIGIT) && !isXDIGIT(c)) ||
(ANYOF_CLASS_TEST(n, ANYOF_PSXSPC) && isPSXSPC(c)) ||
(ANYOF_CLASS_TEST(n, ANYOF_NPSXSPC) && !isPSXSPC(c)) ||
- (ANYOF_CLASS_TEST(n, ANYOF_BLANK) && isBLANK(c)) ||
- (ANYOF_CLASS_TEST(n, ANYOF_NBLANK) && !isBLANK(c))
+ (ANYOF_CLASS_TEST(n, ANYOF_BLANK) && isBLANK_LC(c)) ||
+ (ANYOF_CLASS_TEST(n, ANYOF_NBLANK) && !isBLANK_LC(c))
) /* How's that for a conditional? */
) {
match = TRUE;
|| (flags & ANYOF_IS_SYNTHETIC)))))
{
AV *av;
- SV * const sw = regclass_swash(prog, n, TRUE, 0, (SV**)&av);
+ SV * const sw = core_regclass_swash(prog, n, TRUE, 0, (SV**)&av);
if (sw) {
U8 * utf8_p;
{
dVAR;
regexp * const rex = (regexp *)arg;
- if (PL_reg_eval_set) {
+ if (PL_reg_state.re_state_eval_setup_done) {
if (PL_reg_oldsaved) {
rex->subbeg = PL_reg_oldsaved;
rex->sublen = PL_reg_oldsavedlen;
RXp_MATCH_COPIED_on(rex);
}
PL_reg_magic->mg_len = PL_reg_oldpos;
- PL_reg_eval_set = 0;
+ PL_reg_state.re_state_eval_setup_done = FALSE;
PL_curpm = PL_reg_oldcurpm;
}
}
* Local variables:
* c-indentation-style: bsd
* c-basic-offset: 4
- * indent-tabs-mode: t
+ * indent-tabs-mode: nil
* End:
*
- * ex: set ts=8 sts=4 sw=4 noet:
+ * ex: set ts=8 sts=4 sw=4 et:
*/
https://perl5.git.perl.org / perl5.git / blobdiff