*/
/*
- * "One Ring to rule them all, One Ring to find them..."
+ * One Ring to rule them all, One Ring to find them
+ &
+ * [p.v of _The Lord of the Rings_, opening poem]
+ * [p.50 of _The Lord of the Rings_, I/iii: "The Shadow of the Past"]
+ * [p.254 of _The Lord of the Rings_, II/ii: "The Council of Elrond"]
*/
/* This file contains functions for executing a regular expression. See
**** Alterations to Henry's code are...
****
**** Copyright (C) 1991, 1992, 1993, 1994, 1995, 1996, 1997, 1998, 1999,
- **** 2000, 2001, 2002, 2003, 2004, 2005, 2006, 2007 by Larry Wall and others
+ **** 2000, 2001, 2002, 2003, 2004, 2005, 2006, 2007, 2008
+ **** 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 RF_utf8 8 /* Pattern contains multibyte chars? */
-#define UTF ((PL_reg_flags & RF_utf8) != 0)
+#define UTF_PATTERN ((PL_reg_flags & RF_utf8) != 0)
#define RS_init 1 /* eval environment created */
#define RS_set 2 /* replsv value is set */
* Forwards.
*/
-#define CHR_SVLEN(sv) (do_utf8 ? sv_len_utf8(sv) : SvCUR(sv))
+#define CHR_SVLEN(sv) (utf8_target ? sv_len_utf8(sv) : SvCUR(sv))
#define CHR_DIST(a,b) (PL_reg_match_utf8 ? utf8_distance(a,b) : a - b)
#define HOPc(pos,off) \
#define HOP3(pos,off,lim) (PL_reg_match_utf8 ? reghop3((U8*)(pos), off, (U8*)(lim)) : (U8*)(pos + off))
#define HOP3c(pos,off,lim) ((char*)HOP3(pos,off,lim))
+/* these are unrolled below in the CCC_TRY_XXX defined */
#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
+
+/* Doesn't do an assert to verify that is correct */
+#define LOAD_UTF8_CHARCLASS_NO_CHECK(class) STMT_START { \
+ if (!CAT2(PL_utf8_,class)) { bool ok; ENTER; save_re_context(); ok=CAT2(is_utf8_,class)((const U8*)" "); LEAVE; } } STMT_END
+
#define LOAD_UTF8_CHARCLASS_ALNUM() LOAD_UTF8_CHARCLASS(alnum,"a")
#define LOAD_UTF8_CHARCLASS_DIGIT() LOAD_UTF8_CHARCLASS(digit,"0")
#define LOAD_UTF8_CHARCLASS_SPACE() LOAD_UTF8_CHARCLASS(space," ")
-#define LOAD_UTF8_CHARCLASS_MARK() LOAD_UTF8_CHARCLASS(mark, "\xcd\x86")
+
+#define LOAD_UTF8_CHARCLASS_GCB() /* Grapheme cluster boundaries */ \
+ LOAD_UTF8_CHARCLASS(X_begin, " "); \
+ LOAD_UTF8_CHARCLASS(X_non_hangul, "A"); \
+ /* These are utf8 constants, and not utf-ebcdic constants, so the \
+ * assert should likely and hopefully fail on an EBCDIC machine */ \
+ LOAD_UTF8_CHARCLASS(X_extend, "\xcc\x80"); /* U+0300 */ \
+ \
+ /* No asserts are done for these, in case called on an early \
+ * Unicode version in which they map to nothing */ \
+ LOAD_UTF8_CHARCLASS_NO_CHECK(X_prepend);/* U+0E40 "\xe0\xb9\x80" */ \
+ LOAD_UTF8_CHARCLASS_NO_CHECK(X_L); /* U+1100 "\xe1\x84\x80" */ \
+ LOAD_UTF8_CHARCLASS_NO_CHECK(X_LV); /* U+AC00 "\xea\xb0\x80" */ \
+ LOAD_UTF8_CHARCLASS_NO_CHECK(X_LVT); /* U+AC01 "\xea\xb0\x81" */ \
+ LOAD_UTF8_CHARCLASS_NO_CHECK(X_LV_LVT_V);/* U+AC01 "\xea\xb0\x81" */\
+ LOAD_UTF8_CHARCLASS_NO_CHECK(X_T); /* U+11A8 "\xe1\x86\xa8" */ \
+ LOAD_UTF8_CHARCLASS_NO_CHECK(X_V) /* U+1160 "\xe1\x85\xa0" */
+
+/*
+ We dont use PERL_LEGACY_UNICODE_CHARCLASS_MAPPINGS as the direct test
+ so that it is possible to override the option here without having to
+ rebuild the entire core. as we are required to do if we change regcomp.h
+ which is where PERL_LEGACY_UNICODE_CHARCLASS_MAPPINGS is defined.
+*/
+#if PERL_LEGACY_UNICODE_CHARCLASS_MAPPINGS
+#define BROKEN_UNICODE_CHARCLASS_MAPPINGS
+#endif
+
+#ifdef BROKEN_UNICODE_CHARCLASS_MAPPINGS
+#define LOAD_UTF8_CHARCLASS_PERL_WORD() LOAD_UTF8_CHARCLASS_ALNUM()
+#define LOAD_UTF8_CHARCLASS_PERL_SPACE() LOAD_UTF8_CHARCLASS_SPACE()
+#define LOAD_UTF8_CHARCLASS_POSIX_DIGIT() LOAD_UTF8_CHARCLASS_DIGIT()
+#define RE_utf8_perl_word PL_utf8_alnum
+#define RE_utf8_perl_space PL_utf8_space
+#define RE_utf8_posix_digit PL_utf8_digit
+#define perl_word alnum
+#define perl_space space
+#define posix_digit digit
+#else
+#define LOAD_UTF8_CHARCLASS_PERL_WORD() LOAD_UTF8_CHARCLASS(perl_word,"a")
+#define LOAD_UTF8_CHARCLASS_PERL_SPACE() LOAD_UTF8_CHARCLASS(perl_space," ")
+#define LOAD_UTF8_CHARCLASS_POSIX_DIGIT() LOAD_UTF8_CHARCLASS(posix_digit,"0")
+#define RE_utf8_perl_word PL_utf8_perl_word
+#define RE_utf8_perl_space PL_utf8_perl_space
+#define RE_utf8_posix_digit PL_utf8_posix_digit
+#endif
+
+
+#define CCC_TRY_AFF(NAME,NAMEL,CLASS,STR,LCFUNC_utf8,FUNC,LCFUNC) \
+ case NAMEL: \
+ PL_reg_flags |= RF_tainted; \
+ /* FALL THROUGH */ \
+ case NAME: \
+ if (!nextchr) \
+ sayNO; \
+ if (utf8_target && UTF8_IS_CONTINUED(nextchr)) { \
+ if (!CAT2(PL_utf8_,CLASS)) { \
+ bool ok; \
+ ENTER; \
+ save_re_context(); \
+ ok=CAT2(is_utf8_,CLASS)((const U8*)STR); \
+ assert(ok); \
+ LEAVE; \
+ } \
+ if (!(OP(scan) == NAME \
+ ? cBOOL(swash_fetch(CAT2(PL_utf8_,CLASS), (U8*)locinput, utf8_target)) \
+ : LCFUNC_utf8((U8*)locinput))) \
+ { \
+ sayNO; \
+ } \
+ locinput += PL_utf8skip[nextchr]; \
+ nextchr = UCHARAT(locinput); \
+ break; \
+ } \
+ if (!(OP(scan) == NAME ? FUNC(nextchr) : LCFUNC(nextchr))) \
+ sayNO; \
+ nextchr = UCHARAT(++locinput); \
+ break
+
+#define CCC_TRY_NEG(NAME,NAMEL,CLASS,STR,LCFUNC_utf8,FUNC,LCFUNC) \
+ case NAMEL: \
+ PL_reg_flags |= RF_tainted; \
+ /* FALL THROUGH */ \
+ case NAME : \
+ if (!nextchr && locinput >= PL_regeol) \
+ sayNO; \
+ if (utf8_target && UTF8_IS_CONTINUED(nextchr)) { \
+ if (!CAT2(PL_utf8_,CLASS)) { \
+ bool ok; \
+ ENTER; \
+ save_re_context(); \
+ ok=CAT2(is_utf8_,CLASS)((const U8*)STR); \
+ assert(ok); \
+ LEAVE; \
+ } \
+ if ((OP(scan) == NAME \
+ ? cBOOL(swash_fetch(CAT2(PL_utf8_,CLASS), (U8*)locinput, utf8_target)) \
+ : LCFUNC_utf8((U8*)locinput))) \
+ { \
+ sayNO; \
+ } \
+ locinput += PL_utf8skip[nextchr]; \
+ nextchr = UCHARAT(locinput); \
+ break; \
+ } \
+ if ((OP(scan) == NAME ? FUNC(nextchr) : LCFUNC(nextchr))) \
+ sayNO; \
+ nextchr = UCHARAT(++locinput); \
+ break
+
+
+
+
/* TODO: Combine JUMPABLE and HAS_TEXT to cache OP(rn) */
/* for use after a quantifier and before an EXACT-like node -- japhy */
-#define JUMPABLE(rn) ( \
- OP(rn) == OPEN || OP(rn) == CLOSE || OP(rn) == EVAL || \
+/* it would be nice to rework regcomp.sym to generate this stuff. sigh */
+#define JUMPABLE(rn) ( \
+ OP(rn) == OPEN || \
+ (OP(rn) == CLOSE && (!cur_eval || cur_eval->u.eval.close_paren != ARG(rn))) || \
+ OP(rn) == EVAL || \
OP(rn) == SUSPEND || OP(rn) == IFMATCH || \
OP(rn) == PLUS || OP(rn) == MINMOD || \
OP(rn) == KEEPS || (PL_regkind[OP(rn)] == VERB) || \
static void restore_pos(pTHX_ void *arg);
+#define REGCP_PAREN_ELEMS 4
+#define REGCP_OTHER_ELEMS 5
+#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)
{
dVAR;
const int retval = PL_savestack_ix;
-#define REGCP_PAREN_ELEMS 4
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;
GET_RE_DEBUG_FLAGS_DECL;
if (paren_elems_to_push < 0)
Perl_croak(aTHX_ "panic: paren_elems_to_push < 0");
-#define REGCP_OTHER_ELEMS 8
- SSGROW(paren_elems_to_push + REGCP_OTHER_ELEMS);
+ if ((elems_shifted >> SAVE_TIGHT_SHIFT) != total_elems)
+ Perl_croak(aTHX_ "panic: paren_elems_to_push offset %"UVuf
+ " out of range (%lu-%ld)",
+ total_elems, (unsigned long)PL_regsize, (long)parenfloor);
+
+ SSGROW(total_elems + REGCP_FRAME_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_EXECUTE_r(PerlIO_printf(Perl_debug_log,
+ 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);
SSPUSHPTR(PL_reginput);
-#define REGCP_FRAME_ELEMS 2
-/* REGCP_FRAME_ELEMS are part of the REGCP_OTHER_ELEMS and
- * are needed for the regexp context stack bookkeeping. */
- SSPUSHINT(paren_elems_to_push + REGCP_OTHER_ELEMS - REGCP_FRAME_ELEMS);
- SSPUSHINT(SAVEt_REGCONTEXT); /* Magic cookie. */
+ SSPUSHUV(SAVEt_REGCONTEXT | elems_shifted); /* Magic cookie. */
return retval;
}
S_regcppop(pTHX_ const regexp *rex)
{
dVAR;
- U32 i;
+ UV i;
char *input;
-
GET_RE_DEBUG_FLAGS_DECL;
+ PERL_ARGS_ASSERT_REGCPPOP;
+
/* Pop REGCP_OTHER_ELEMS before the parentheses loop starts. */
- i = SSPOPINT;
- assert(i == SAVEt_REGCONTEXT); /* Check that the magic cookie is there. */
- i = SSPOPINT; /* Parentheses elements to pop. */
+ 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;
PL_regsize = SSPOPINT;
- PL_regendp=(I32 *) SSPOPPTR;
- PL_regstartp=(I32 *) SSPOPPTR;
+ PL_regoffs=(regexp_paren_pair *) SSPOPPTR;
-
+ i -= REGCP_OTHER_ELEMS;
/* Now restore the parentheses context. */
- for (i -= (REGCP_OTHER_ELEMS - REGCP_FRAME_ELEMS);
- i > 0; i -= REGCP_PAREN_ELEMS) {
+ for ( ; i > 0; i -= REGCP_PAREN_ELEMS) {
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;
- DEBUG_EXECUTE_r(
+ 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)" : ""));
);
}
- DEBUG_EXECUTE_r(
+ DEBUG_BUFFERS_r(
if (*PL_reglastparen + 1 <= rex->nparens) {
PerlIO_printf(Perl_debug_log,
" restoring \\%"IVdf"..\\%"IVdf" to undef\n",
* requiring null fields (pat.t#187 and split.t#{13,14}
* (as of patchlevel 7877) will fail. Then again,
* this code seems to be necessary or otherwise
- * building DynaLoader will fail:
- * "Error: '*' not in typemap in DynaLoader.xs, line 164"
- * --jhi */
+ * this erroneously leaves $1 defined: "1" =~ /^(?:(\d)x)?\d$/
+ * --jhi updated by dapm */
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;
- pregexec - match a regexp against a string
*/
I32
-Perl_pregexec(pTHX_ register regexp *prog, char *stringarg, register char *strend,
+Perl_pregexec(pTHX_ REGEXP * const prog, char* stringarg, register char *strend,
char *strbeg, I32 minend, SV *screamer, U32 nosave)
/* strend: pointer to null at end of string */
/* strbeg: real beginning of string */
/* minend: end of match must be >=minend after stringarg. */
/* nosave: For optimizations. */
{
+ PERL_ARGS_ASSERT_PREGEXEC;
+
return
regexec_flags(prog, stringarg, strend, strbeg, minend, screamer, NULL,
nosave ? 0 : REXEC_COPY_STR);
deleted from the finite automaton. */
char *
-Perl_re_intuit_start(pTHX_ regexp *prog, SV *sv, char *strpos,
- char *strend, U32 flags, re_scream_pos_data *data)
+Perl_re_intuit_start(pTHX_ REGEXP * const rx, SV *sv, char *strpos,
+ char *strend, const U32 flags, re_scream_pos_data *data)
{
dVAR;
+ struct regexp *const prog = (struct regexp *)SvANY(rx);
register I32 start_shift = 0;
/* Should be nonnegative! */
register I32 end_shift = 0;
register SV *check;
char *strbeg;
char *t;
- const bool do_utf8 = (sv && SvUTF8(sv)) ? 1 : 0; /* if no sv we have to assume bytes */
+ const bool utf8_target = (sv && SvUTF8(sv)) ? 1 : 0; /* if no sv we have to assume bytes */
I32 ml_anch;
register char *other_last = NULL; /* other substr checked before this */
char *check_at = NULL; /* check substr found at this pos */
#ifdef DEBUGGING
const char * const i_strpos = strpos;
#endif
-
GET_RE_DEBUG_FLAGS_DECL;
- RX_MATCH_UTF8_set(prog,do_utf8);
+ PERL_ARGS_ASSERT_RE_INTUIT_START;
+
+ RX_MATCH_UTF8_set(rx,utf8_target);
- if (prog->extflags & RXf_UTF8) {
+ if (RX_UTF8(rx)) {
PL_reg_flags |= RF_utf8;
}
DEBUG_EXECUTE_r(
- debug_start_match(prog, do_utf8, strpos, strend,
+ debug_start_match(rx, utf8_target, strpos, strend,
sv ? "Guessing start of match in sv for"
: "Guessing start of match in string for");
);
strbeg = (sv && SvPOK(sv)) ? strend - SvCUR(sv) : strpos;
PL_regeol = strend;
- if (do_utf8) {
+ if (utf8_target) {
if (!prog->check_utf8 && prog->check_substr)
to_utf8_substr(prog);
check = prog->check_utf8;
#ifdef QDEBUGGING /* 7/99: reports of failure (with the older version) */
if (end_shift < 0)
Perl_croak(aTHX_ "panic: end_shift: %"IVdf" pattern:\n%s\n ",
- (IV)end_shift, prog->precomp);
+ (IV)end_shift, RX_PRECOMP(prog));
#endif
restart:
else
goto fail_finish;
/* we may be pointing at the wrong string */
- if (s && RX_MATCH_COPIED(prog))
+ if (s && RXp_MATCH_COPIED(prog))
s = strbeg + (s - SvPVX_const(sv));
if (data)
*data->scream_olds = s;
unshift s. */
DEBUG_EXECUTE_r({
- RE_PV_QUOTED_DECL(quoted, do_utf8, PERL_DEBUG_PAD_ZERO(0),
+ RE_PV_QUOTED_DECL(quoted, utf8_target, PERL_DEBUG_PAD_ZERO(0),
SvPVX_const(check), RE_SV_DUMPLEN(check), 30);
PerlIO_printf(Perl_debug_log, "%s %s substr %s%s%s",
(s ? "Found" : "Did not find"),
- (check == (do_utf8 ? prog->anchored_utf8 : prog->anchored_substr)
+ (check == (utf8_target ? prog->anchored_utf8 : prog->anchored_substr)
? "anchored" : "floating"),
quoted,
RE_SV_TAIL(check),
Probably it is right to do no SCREAM here...
*/
- if (do_utf8 ? (prog->float_utf8 && prog->anchored_utf8)
+ if (utf8_target ? (prog->float_utf8 && prog->anchored_utf8)
: (prog->float_substr && prog->anchored_substr))
{
/* Take into account the "other" substring. */
/* XXXX May be hopelessly wrong for UTF... */
if (!other_last)
other_last = strpos;
- if (check == (do_utf8 ? prog->float_utf8 : prog->float_substr)) {
+ if (check == (utf8_target ? prog->float_utf8 : prog->float_substr)) {
do_other_anchored:
{
char * const last = HOP3c(s, -start_shift, strbeg);
t = s - prog->check_offset_max;
if (s - strpos > prog->check_offset_max /* signed-corrected t > strpos */
- && (!do_utf8
+ && (!utf8_target
|| ((t = (char*)reghopmaybe3((U8*)s, -(prog->check_offset_max), (U8*)strpos))
&& t > strpos)))
NOOP;
*/
/* On end-of-str: see comment below. */
- must = do_utf8 ? prog->anchored_utf8 : prog->anchored_substr;
+ must = utf8_target ? prog->anchored_utf8 : prog->anchored_substr;
if (must == &PL_sv_undef) {
s = (char*)NULL;
DEBUG_r(must = prog->anchored_utf8); /* for debug */
multiline ? FBMrf_MULTILINE : 0
);
DEBUG_EXECUTE_r({
- RE_PV_QUOTED_DECL(quoted, do_utf8, PERL_DEBUG_PAD_ZERO(0),
+ RE_PV_QUOTED_DECL(quoted, utf8_target, PERL_DEBUG_PAD_ZERO(0),
SvPVX_const(must), RE_SV_DUMPLEN(must), 30);
PerlIO_printf(Perl_debug_log, "%s anchored substr %s%s",
(s ? "Found" : "Contradicts"),
if (s < other_last)
s = other_last;
/* XXXX It is not documented what units *_offsets are in. Assume bytes. */
- must = do_utf8 ? prog->float_utf8 : prog->float_substr;
+ must = utf8_target ? prog->float_utf8 : prog->float_substr;
/* fbm_instr() takes into account exact value of end-of-str
if the check is SvTAIL(ed). Since false positives are OK,
and end-of-str is not later than strend we are OK. */
- (SvTAIL(must)!=0),
must, multiline ? FBMrf_MULTILINE : 0);
DEBUG_EXECUTE_r({
- RE_PV_QUOTED_DECL(quoted, do_utf8, PERL_DEBUG_PAD_ZERO(0),
+ RE_PV_QUOTED_DECL(quoted, utf8_target, PERL_DEBUG_PAD_ZERO(0),
SvPVX_const(must), RE_SV_DUMPLEN(must), 30);
PerlIO_printf(Perl_debug_log, "%s floating substr %s%s",
(s ? "Found" : "Contradicts"),
);
if (s - strpos > prog->check_offset_max /* signed-corrected t > strpos */
- && (!do_utf8
+ && (!utf8_target
|| ((t = (char*)reghopmaybe3((U8*)s, -prog->check_offset_max, (U8*) ((prog->check_offset_max<0) ? strend : strpos)))
&& t > strpos)))
{
while (t < strend - prog->minlen) {
if (*t == '\n') {
if (t < check_at - prog->check_offset_min) {
- if (do_utf8 ? prog->anchored_utf8 : prog->anchored_substr) {
+ if (utf8_target ? prog->anchored_utf8 : prog->anchored_substr) {
/* Since we moved from the found position,
we definitely contradict the found anchored
substr. Due to the above check we do not
}
s = t;
set_useful:
- ++BmUSEFUL(do_utf8 ? prog->check_utf8 : prog->check_substr); /* hooray/5 */
+ ++BmUSEFUL(utf8_target ? prog->check_utf8 : prog->check_substr); /* hooray/5 */
}
else {
/* The found string does not prohibit matching at strpos,
);
success_at_start:
if (!(prog->intflags & PREGf_NAUGHTY) /* XXXX If strpos moved? */
- && (do_utf8 ? (
+ && (utf8_target ? (
prog->check_utf8 /* Could be deleted already */
&& --BmUSEFUL(prog->check_utf8) < 0
&& (prog->check_utf8 == prog->float_utf8)
{
/* If flags & SOMETHING - do not do it many times on the same match */
DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "... Disabling check substring...\n"));
- 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);
+ /* XXX Does the destruction order has to change with utf8_target? */
+ SvREFCNT_dec(utf8_target ? prog->check_utf8 : prog->check_substr);
+ SvREFCNT_dec(utf8_target ? prog->check_substr : prog->check_utf8);
prog->check_substr = prog->check_utf8 = NULL; /* disable */
prog->float_substr = prog->float_utf8 = NULL; /* clear */
check = NULL; /* abort */
goto fail;
/* Contradict one of substrings */
if (prog->anchored_substr || prog->anchored_utf8) {
- if ((do_utf8 ? prog->anchored_utf8 : prog->anchored_substr) == check) {
+ if ((utf8_target ? prog->anchored_utf8 : prog->anchored_substr) == check) {
DEBUG_EXECUTE_r( what = "anchored" );
hop_and_restart:
s = HOP3c(t, 1, strend);
PL_colors[0], PL_colors[1], (long)(t - i_strpos)) );
goto try_at_offset;
}
- if (!(do_utf8 ? prog->float_utf8 : prog->float_substr)) /* Could have been deleted */
+ if (!(utf8_target ? prog->float_utf8 : prog->float_substr)) /* Could have been deleted */
goto fail;
/* Check is floating subtring. */
retry_floating_check:
fail_finish: /* Substring not found */
if (prog->check_substr || prog->check_utf8) /* could be removed already */
- BmUSEFUL(do_utf8 ? prog->check_utf8 : prog->check_substr) += 5; /* hooray */
+ BmUSEFUL(utf8_target ? prog->check_utf8 : prog->check_substr) += 5; /* hooray */
fail:
DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "%sMatch rejected by optimizer%s\n",
PL_colors[4], PL_colors[5]));
return NULL;
}
-
+#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 ); \
+ uvc = utf8n_to_uvuni( uscan, UTF8_MAXLEN, &len, uniflags ); \
foldlen -= len; \
uscan += len; \
len=0; \
} else { \
- uvc = utf8n_to_uvuni( (U8*)uc, UTF8_MAXLEN, &len, uniflags ); \
+ uvc = utf8n_to_uvuni( (U8*)uc, UTF8_MAXLEN, &len, uniflags ); \
uvc = to_uni_fold( uvc, foldbuf, &foldlen ); \
foldlen -= UNISKIP( uvc ); \
uscan = foldbuf + UNISKIP( uvc ); \
} \
break; \
+ case trie_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; \
uvc = (UV)*uc; \
len = 1; \
} \
- \
if (uvc < 256) { \
charid = trie->charmap[ uvc ]; \
} \
} \
} STMT_END
-#define REXEC_FBC_EXACTISH_CHECK(CoNd) \
+#define REXEC_FBC_EXACTISH_CHECK(CoNd) \
+{ \
+ char *my_strend= (char *)strend; \
if ( (CoNd) \
&& (ln == len || \
- ibcmp_utf8(s, NULL, 0, do_utf8, \
- m, NULL, ln, (bool)UTF)) \
- && (!reginfo || regtry(reginfo, &s)) ) \
+ foldEQ_utf8(s, &my_strend, 0, utf8_target, \
+ m, NULL, ln, cBOOL(UTF_PATTERN))) \
+ && (!reginfo || regtry(reginfo, &s)) ) \
goto got_it; \
else { \
U8 foldbuf[UTF8_MAXBYTES_CASE+1]; \
f = to_utf8_fold(tmpbuf, foldbuf, &foldlen); \
if ( f != c \
&& (f == c1 || f == c2) \
- && (ln == foldlen || \
- !ibcmp_utf8((char *) foldbuf, \
- NULL, foldlen, do_utf8, \
- m, \
- NULL, ln, (bool)UTF)) \
- && (!reginfo || regtry(reginfo, &s)) ) \
+ && (ln == len || \
+ foldEQ_utf8(s, &my_strend, 0, utf8_target,\
+ m, NULL, ln, cBOOL(UTF_PATTERN)))\
+ && (!reginfo || regtry(reginfo, &s)) ) \
goto got_it; \
} \
- s += len
+} \
+s += len
#define REXEC_FBC_EXACTISH_SCAN(CoNd) \
STMT_START { \
while (s <= e) { \
if ( (CoNd) \
- && (ln == 1 || !(OP(c) == EXACTF \
- ? ibcmp(s, m, ln) \
- : ibcmp_locale(s, m, ln))) \
+ && (ln == 1 || (OP(c) == EXACTF \
+ ? foldEQ(s, m, ln) \
+ : foldEQ_locale(s, m, ln))) \
&& (!reginfo || regtry(reginfo, &s)) ) \
goto got_it; \
s++; \
if ((!reginfo || regtry(reginfo, &s))) \
goto got_it
+#define REXEC_FBC_CSCAN(CoNdUtF8,CoNd) \
+ if (utf8_target) { \
+ REXEC_FBC_UTF8_CLASS_SCAN(CoNdUtF8); \
+ } \
+ else { \
+ REXEC_FBC_CLASS_SCAN(CoNd); \
+ } \
+ break
+
#define REXEC_FBC_CSCAN_PRELOAD(UtFpReLoAd,CoNdUtF8,CoNd) \
- if (do_utf8) { \
+ if (utf8_target) { \
UtFpReLoAd; \
REXEC_FBC_UTF8_CLASS_SCAN(CoNdUtF8); \
} \
#define REXEC_FBC_CSCAN_TAINT(CoNdUtF8,CoNd) \
PL_reg_flags |= RF_tainted; \
- if (do_utf8) { \
+ if (utf8_target) { \
REXEC_FBC_UTF8_CLASS_SCAN(CoNdUtF8); \
} \
else { \
unsigned int c2;
char *e;
register I32 tmp = 1; /* Scratch variable? */
- register const bool do_utf8 = PL_reg_match_utf8;
+ register const bool utf8_target = PL_reg_match_utf8;
RXi_GET_DECL(prog,progi);
+
+ PERL_ARGS_ASSERT_FIND_BYCLASS;
/* We know what class it must start with. */
switch (OP(c)) {
case ANYOF:
- if (do_utf8) {
+ if (utf8_target) {
REXEC_FBC_UTF8_CLASS_SCAN((ANYOF_FLAGS(c) & ANYOF_UNICODE) ||
!UTF8_IS_INVARIANT((U8)s[0]) ?
- reginclass(prog, c, (U8*)s, 0, do_utf8) :
+ reginclass(prog, c, (U8*)s, 0, utf8_target) :
REGINCLASS(prog, c, (U8*)s));
}
else {
m = STRING(c);
ln = STR_LEN(c); /* length to match in octets/bytes */
lnc = (I32) ln; /* length to match in characters */
- if (UTF) {
+ if (UTF_PATTERN) {
STRLEN ulen1, ulen2;
U8 *sm = (U8 *) m;
U8 tmpbuf1[UTF8_MAXBYTES_CASE+1];
U8 tmpbuf2[UTF8_MAXBYTES_CASE+1];
- const U32 uniflags = UTF8_ALLOW_DEFAULT;
-
- to_utf8_lower((U8*)m, tmpbuf1, &ulen1);
- to_utf8_upper((U8*)m, tmpbuf2, &ulen2);
-
+ /* used by commented-out code below */
+ /*const U32 uniflags = UTF8_ALLOW_DEFAULT;*/
+
+ /* XXX: Since the node will be case folded at compile
+ time this logic is a little odd, although im not
+ sure that its actually wrong. --dmq */
+
+ c1 = to_utf8_lower((U8*)m, tmpbuf1, &ulen1);
+ c2 = to_utf8_upper((U8*)m, tmpbuf2, &ulen2);
+
+ /* XXX: This is kinda strange. to_utf8_XYZ returns the
+ codepoint of the first character in the converted
+ form, yet originally we did the extra step.
+ No tests fail by commenting this code out however
+ so Ive left it out. -- dmq.
+
c1 = utf8n_to_uvchr(tmpbuf1, UTF8_MAXBYTES_CASE,
0, uniflags);
c2 = utf8n_to_uvchr(tmpbuf2, UTF8_MAXBYTES_CASE,
0, uniflags);
+ */
+
lnc = 0;
while (sm < ((U8 *) m + ln)) {
lnc++;
* than just upper and lower: one needs to use
* the so-called folding case for case-insensitive
* matching (called "loose matching" in Unicode).
- * ibcmp_utf8() will do just that. */
+ * foldEQ_utf8() will do just that. */
- if (do_utf8) {
+ if (utf8_target || UTF_PATTERN) {
UV c, f;
U8 tmpbuf [UTF8_MAXBYTES+1];
- STRLEN len, foldlen;
+ STRLEN len = 1;
+ STRLEN foldlen;
const U32 uniflags = UTF8_ALLOW_DEFAULT;
if (c1 == c2) {
/* Upper and lower of 1st char are equal -
* probably not a "letter". */
while (s <= e) {
- c = utf8n_to_uvchr((U8*)s, UTF8_MAXBYTES, &len,
+ if (utf8_target) {
+ c = utf8n_to_uvchr((U8*)s, UTF8_MAXBYTES, &len,
uniflags);
+ } else {
+ c = *((U8*)s);
+ }
REXEC_FBC_EXACTISH_CHECK(c == c1);
}
}
else {
while (s <= e) {
- c = utf8n_to_uvchr((U8*)s, UTF8_MAXBYTES, &len,
+ if (utf8_target) {
+ c = utf8n_to_uvchr((U8*)s, UTF8_MAXBYTES, &len,
uniflags);
+ } else {
+ c = *((U8*)s);
+ }
/* Handle some of the three Greek sigmas cases.
* Note that not all the possible combinations
}
}
else {
+ /* Neither pattern nor string are UTF8 */
if (c1 == c2)
REXEC_FBC_EXACTISH_SCAN(*(U8*)s == c1);
else
PL_reg_flags |= RF_tainted;
/* FALL THROUGH */
case BOUND:
- if (do_utf8) {
+ if (utf8_target) {
if (s == PL_bostr)
tmp = '\n';
else {
LOAD_UTF8_CHARCLASS_ALNUM();
REXEC_FBC_UTF8_SCAN(
if (tmp == !(OP(c) == BOUND ?
- (bool)swash_fetch(PL_utf8_alnum, (U8*)s, do_utf8) :
+ cBOOL(swash_fetch(PL_utf8_alnum, (U8*)s, utf8_target)) :
isALNUM_LC_utf8((U8*)s)))
{
tmp = !tmp;
PL_reg_flags |= RF_tainted;
/* FALL THROUGH */
case NBOUND:
- if (do_utf8) {
+ if (utf8_target) {
if (s == PL_bostr)
tmp = '\n';
else {
LOAD_UTF8_CHARCLASS_ALNUM();
REXEC_FBC_UTF8_SCAN(
if (tmp == !(OP(c) == NBOUND ?
- (bool)swash_fetch(PL_utf8_alnum, (U8*)s, do_utf8) :
+ cBOOL(swash_fetch(PL_utf8_alnum, (U8*)s, utf8_target)) :
isALNUM_LC_utf8((U8*)s)))
tmp = !tmp;
else REXEC_FBC_TRYIT;
break;
case ALNUM:
REXEC_FBC_CSCAN_PRELOAD(
- LOAD_UTF8_CHARCLASS_ALNUM(),
- swash_fetch(PL_utf8_alnum, (U8*)s, do_utf8),
+ LOAD_UTF8_CHARCLASS_PERL_WORD(),
+ swash_fetch(RE_utf8_perl_word, (U8*)s, utf8_target),
isALNUM(*s)
);
case ALNUML:
);
case NALNUM:
REXEC_FBC_CSCAN_PRELOAD(
- LOAD_UTF8_CHARCLASS_ALNUM(),
- !swash_fetch(PL_utf8_alnum, (U8*)s, do_utf8),
+ LOAD_UTF8_CHARCLASS_PERL_WORD(),
+ !swash_fetch(RE_utf8_perl_word, (U8*)s, utf8_target),
!isALNUM(*s)
);
case NALNUML:
);
case SPACE:
REXEC_FBC_CSCAN_PRELOAD(
- LOAD_UTF8_CHARCLASS_SPACE(),
- *s == ' ' || swash_fetch(PL_utf8_space,(U8*)s, do_utf8),
+ LOAD_UTF8_CHARCLASS_PERL_SPACE(),
+ *s == ' ' || swash_fetch(RE_utf8_perl_space,(U8*)s, utf8_target),
isSPACE(*s)
);
case SPACEL:
);
case NSPACE:
REXEC_FBC_CSCAN_PRELOAD(
- LOAD_UTF8_CHARCLASS_SPACE(),
- !(*s == ' ' || swash_fetch(PL_utf8_space,(U8*)s, do_utf8)),
+ LOAD_UTF8_CHARCLASS_PERL_SPACE(),
+ !(*s == ' ' || swash_fetch(RE_utf8_perl_space,(U8*)s, utf8_target)),
!isSPACE(*s)
);
case NSPACEL:
);
case DIGIT:
REXEC_FBC_CSCAN_PRELOAD(
- LOAD_UTF8_CHARCLASS_DIGIT(),
- swash_fetch(PL_utf8_digit,(U8*)s, do_utf8),
+ LOAD_UTF8_CHARCLASS_POSIX_DIGIT(),
+ swash_fetch(RE_utf8_posix_digit,(U8*)s, utf8_target),
isDIGIT(*s)
);
case DIGITL:
);
case NDIGIT:
REXEC_FBC_CSCAN_PRELOAD(
- LOAD_UTF8_CHARCLASS_DIGIT(),
- !swash_fetch(PL_utf8_digit,(U8*)s, do_utf8),
+ LOAD_UTF8_CHARCLASS_POSIX_DIGIT(),
+ !swash_fetch(RE_utf8_posix_digit,(U8*)s, utf8_target),
!isDIGIT(*s)
);
case NDIGITL:
!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:
{
- const enum { trie_plain, trie_utf8, trie_utf8_fold }
- trie_type = do_utf8 ?
- (c->flags == EXACT ? trie_utf8 : trie_utf8_fold)
- : trie_plain;
+ DECL_TRIE_TYPE(c);
/* what trie are we using right now */
reg_ac_data *aho
= (reg_ac_data*)progi->data->data[ ARG( c ) ];
reg_trie_data *trie
= (reg_trie_data*)progi->data->data[ aho->trie ];
- HV *widecharmap = (HV*) progi->data->data[ aho->trie + 1 ];
+ HV *widecharmap = MUTABLE_HV(progi->data->data[ aho->trie + 1 ]);
const char *last_start = strend - trie->minlen;
#ifdef DEBUGGING
U8 **points; /* map of where we were in the input string
when reading a given char. For ASCII this
is unnecessary overhead as the relationship
- is always 1:1, but for unicode, especially
- case folded unicode this is not true. */
+ is always 1:1, but for Unicode, especially
+ case folded Unicode this is not true. */
U8 foldbuf[ UTF8_MAXBYTES_CASE + 1 ];
U8 *bitmap=NULL;
DEBUG_TRIE_EXECUTE_r(
if ( uc <= (U8*)last_start && !BITMAP_TEST(bitmap,*uc) ) {
dump_exec_pos( (char *)uc, c, strend, real_start,
- (char *)uc, do_utf8 );
+ (char *)uc, utf8_target );
PerlIO_printf( Perl_debug_log,
" Scanning for legal start char...\n");
}
}
if ( word ) {
- U8 *lpos= points[ (pointpos - trie->wordlen[word-1] ) % maxlen ];
+ U8 *lpos= points[ (pointpos - trie->wordinfo[word].len) % maxlen ];
if (!leftmost || lpos < leftmost) {
DEBUG_r(accepted_word=word);
leftmost= lpos;
foldbuf, uniflags);
DEBUG_TRIE_EXECUTE_r({
dump_exec_pos( (char *)uc, c, strend, real_start,
- s, do_utf8 );
+ s, utf8_target );
PerlIO_printf(Perl_debug_log,
" Charid:%3u CP:%4"UVxf" ",
charid, uvc);
DEBUG_TRIE_EXECUTE_r({
if (failed)
dump_exec_pos( (char *)uc, c, strend, real_start,
- s, do_utf8 );
+ s, utf8_target );
PerlIO_printf( Perl_debug_log,
"%sState: %4"UVxf", word=%"UVxf,
failed ? " Fail transition to " : "",
});
if ( base ) {
U32 tmp;
+ I32 offset;
if (charid &&
- (base + charid > trie->uniquecharcount )
- && (base + charid - 1 - trie->uniquecharcount
- < trie->lasttrans)
- && trie->trans[base + charid - 1 -
- trie->uniquecharcount].check == state
- && (tmp=trie->trans[base + charid - 1 -
- trie->uniquecharcount ].next))
+ ( ((offset = base + charid
+ - 1 - trie->uniquecharcount)) >= 0)
+ && ((U32)offset < trie->lasttrans)
+ && trie->trans[offset].check == state
+ && (tmp=trie->trans[offset].next))
{
DEBUG_TRIE_EXECUTE_r(
PerlIO_printf( Perl_debug_log," - legal\n"));
}
}
if ( aho->states[ state ].wordnum ) {
- U8 *lpos = points[ (pointpos - trie->wordlen[aho->states[ state ].wordnum-1]) % maxlen ];
+ U8 *lpos = points[ (pointpos - trie->wordinfo[aho->states[ state ].wordnum].len) % maxlen ];
if (!leftmost || lpos < leftmost) {
DEBUG_r(accepted_word=aho->states[ state ].wordnum);
leftmost = lpos;
return s;
}
+
/*
- regexec_flags - match a regexp against a string
*/
I32
-Perl_regexec_flags(pTHX_ register regexp *prog, char *stringarg, register char *strend,
+Perl_regexec_flags(pTHX_ REGEXP * const rx, char *stringarg, register char *strend,
char *strbeg, I32 minend, SV *sv, void *data, U32 flags)
/* strend: pointer to null at end of string */
/* strbeg: real beginning of string */
/* nosave: For optimizations. */
{
dVAR;
+ struct regexp *const prog = (struct regexp *)SvANY(rx);
/*register*/ char *s;
register regnode *c;
/*register*/ char *startpos = stringarg;
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);
+ const bool utf8_target = cBOOL(DO_UTF8(sv));
I32 multiline;
RXi_GET_DECL(prog,progi);
regmatch_info reginfo; /* create some info to pass to regtry etc */
-
+ regexp_paren_pair *swap = NULL;
GET_RE_DEBUG_FLAGS_DECL;
+ PERL_ARGS_ASSERT_REGEXEC_FLAGS;
PERL_UNUSED_ARG(data);
/* Be paranoid... */
}
multiline = prog->extflags & RXf_PMf_MULTILINE;
- reginfo.prog = prog;
+ reginfo.prog = rx; /* Yes, sorry that this is confusing. */
- RX_MATCH_UTF8_set(prog, do_utf8);
+ RX_MATCH_UTF8_set(rx, utf8_target);
DEBUG_EXECUTE_r(
- debug_start_match(prog, do_utf8, startpos, strend,
+ debug_start_match(rx, utf8_target, startpos, strend,
"Matching");
);
PL_reg_eval_set = 0;
PL_reg_maxiter = 0;
- if (prog->extflags & RXf_UTF8)
+ if (RX_UTF8(rx))
PL_reg_flags |= RF_utf8;
/* Mark beginning of line for ^ and lookbehind. */
if (prog->extflags & RXf_GPOS_SEEN) { /* Need to set reginfo->ganch */
MAGIC *mg;
-
- if (flags & REXEC_IGNOREPOS) /* Means: check only at start */
+ if (flags & REXEC_IGNOREPOS){ /* Means: check only at start */
reginfo.ganch = startpos + prog->gofs;
- else if (sv && SvTYPE(sv) >= SVt_PVMG
+ DEBUG_GPOS_r(PerlIO_printf(Perl_debug_log,
+ "GPOS IGNOREPOS: reginfo.ganch = startpos + %"UVxf"\n",(UV)prog->gofs));
+ } else if (sv && SvTYPE(sv) >= SVt_PVMG
&& SvMAGIC(sv)
&& (mg = mg_find(sv, PERL_MAGIC_regex_global))
&& mg->mg_len >= 0) {
reginfo.ganch = strbeg + mg->mg_len; /* Defined pos() */
+ DEBUG_GPOS_r(PerlIO_printf(Perl_debug_log,
+ "GPOS MAGIC: reginfo.ganch = strbeg + %"IVdf"\n",(IV)mg->mg_len));
+
if (prog->extflags & RXf_ANCH_GPOS) {
if (s > reginfo.ganch)
goto phooey;
s = reginfo.ganch - prog->gofs;
+ DEBUG_GPOS_r(PerlIO_printf(Perl_debug_log,
+ "GPOS ANCH_GPOS: s = ganch - %"UVxf"\n",(UV)prog->gofs));
+ if (s < strbeg)
+ goto phooey;
}
}
else if (data) {
reginfo.ganch = strbeg + PTR2UV(data);
- } else /* pos() not defined */
+ DEBUG_GPOS_r(PerlIO_printf(Perl_debug_log,
+ "GPOS DATA: reginfo.ganch= strbeg + %"UVxf"\n",PTR2UV(data)));
+
+ } else { /* pos() not defined */
reginfo.ganch = strbeg;
+ DEBUG_GPOS_r(PerlIO_printf(Perl_debug_log,
+ "GPOS: reginfo.ganch = strbeg\n"));
+ }
}
- if (PL_curpm && (PM_GETRE(PL_curpm) == prog)) {
- I32 *t;
- if (!progi->swap) {
+ if (PL_curpm && (PM_GETRE(PL_curpm) == rx)) {
/* We have to be careful. If the previous successful match
- was from this regex we don't want a subsequent paritally
- successful match to clobber the old results.
+ was from this regex we don't want a subsequent partially
+ successful match to clobber the old results.
So when we detect this possibility we add a swap buffer
to the re, and switch the buffer each match. If we fail
we switch it back, otherwise we leave it swapped.
*/
- Newxz(progi->swap, 1, regexp_paren_ofs);
- /* no need to copy these */
- Newxz(progi->swap->startp, prog->nparens + 1, I32);
- Newxz(progi->swap->endp, prog->nparens + 1, I32);
- }
- t = progi->swap->startp;
- progi->swap->startp = prog->startp;
- prog->startp = t;
- t = progi->swap->endp;
- progi->swap->endp = prog->endp;
- prog->endp = t;
+ swap = prog->offs;
+ /* do we need a save destructor here for eval dies? */
+ Newxz(prog->offs, (prog->nparens + 1), regexp_paren_pair);
}
if (!(flags & REXEC_CHECKED) && (prog->check_substr != NULL || prog->check_utf8 != NULL)) {
re_scream_pos_data d;
d.scream_olds = &scream_olds;
d.scream_pos = &scream_pos;
- s = re_intuit_start(prog, sv, s, strend, flags, &d);
+ s = re_intuit_start(rx, sv, s, strend, flags, &d);
if (!s) {
DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "Not present...\n"));
goto phooey; /* not present */
if (regtry(®info, &s))
goto got_it;
after_try:
- if (s >= end)
+ if (s > end)
goto phooey;
if (prog->extflags & RXf_USE_INTUIT) {
- s = re_intuit_start(prog, sv, s + 1, strend, flags, NULL);
+ s = re_intuit_start(rx, sv, s + 1, strend, flags, NULL);
if (!s)
goto phooey;
}
is bogus -- we set it above, when prog->extflags & RXf_GPOS_SEEN
and we only enter this block when the same bit is set. */
char *tmp_s = reginfo.ganch - prog->gofs;
- if (regtry(®info, &tmp_s))
+
+ if (tmp_s >= strbeg && regtry(®info, &tmp_s))
goto got_it;
goto phooey;
}
/* Messy cases: unanchored match. */
if ((prog->anchored_substr || prog->anchored_utf8) && prog->intflags & PREGf_SKIP) {
/* we have /x+whatever/ */
- /* it must be a one character string (XXXX Except UTF?) */
+ /* it must be a one character string (XXXX Except UTF_PATTERN?) */
char ch;
#ifdef DEBUGGING
int did_match = 0;
#endif
- if (!(do_utf8 ? prog->anchored_utf8 : prog->anchored_substr))
- do_utf8 ? to_utf8_substr(prog) : to_byte_substr(prog);
- ch = SvPVX_const(do_utf8 ? prog->anchored_utf8 : prog->anchored_substr)[0];
+ if (!(utf8_target ? prog->anchored_utf8 : prog->anchored_substr))
+ utf8_target ? to_utf8_substr(prog) : to_byte_substr(prog);
+ ch = SvPVX_const(utf8_target ? prog->anchored_utf8 : prog->anchored_substr)[0];
- if (do_utf8) {
+ if (utf8_target) {
REXEC_FBC_SCAN(
if (*s == ch) {
DEBUG_EXECUTE_r( did_match = 1 );
int did_match = 0;
#endif
if (prog->anchored_substr || prog->anchored_utf8) {
- if (!(do_utf8 ? prog->anchored_utf8 : prog->anchored_substr))
- do_utf8 ? to_utf8_substr(prog) : to_byte_substr(prog);
- must = do_utf8 ? prog->anchored_utf8 : prog->anchored_substr;
+ if (!(utf8_target ? prog->anchored_utf8 : prog->anchored_substr))
+ utf8_target ? to_utf8_substr(prog) : to_byte_substr(prog);
+ must = utf8_target ? prog->anchored_utf8 : prog->anchored_substr;
back_max = back_min = prog->anchored_offset;
} else {
- if (!(do_utf8 ? prog->float_utf8 : prog->float_substr))
- do_utf8 ? to_utf8_substr(prog) : to_byte_substr(prog);
- must = do_utf8 ? prog->float_utf8 : prog->float_substr;
+ if (!(utf8_target ? prog->float_utf8 : prog->float_substr))
+ utf8_target ? to_utf8_substr(prog) : to_byte_substr(prog);
+ must = utf8_target ? prog->float_utf8 : prog->float_substr;
back_max = prog->float_max_offset;
back_min = prog->float_min_offset;
}
(unsigned char*)strend, must,
multiline ? FBMrf_MULTILINE : 0))) ) {
/* we may be pointing at the wrong string */
- if ((flags & REXEC_SCREAM) && RX_MATCH_COPIED(prog))
+ if ((flags & REXEC_SCREAM) && RXp_MATCH_COPIED(prog))
s = strbeg + (s - SvPVX_const(sv));
DEBUG_EXECUTE_r( did_match = 1 );
if (HOPc(s, -back_max) > last1) {
last1 = HOPc(s, -back_min);
s = t;
}
- if (do_utf8) {
+ if (utf8_target) {
while (s <= last1) {
if (regtry(®info, &s))
goto got_it;
}
}
DEBUG_EXECUTE_r(if (!did_match) {
- RE_PV_QUOTED_DECL(quoted, do_utf8, PERL_DEBUG_PAD_ZERO(0),
+ RE_PV_QUOTED_DECL(quoted, utf8_target, PERL_DEBUG_PAD_ZERO(0),
SvPVX_const(must), RE_SV_DUMPLEN(must), 30);
PerlIO_printf(Perl_debug_log, "Did not find %s substr %s%s...\n",
((must == prog->anchored_substr || must == prog->anchored_utf8)
SV * const prop = sv_newmortal();
regprop(prog, prop, c);
{
- RE_PV_QUOTED_DECL(quoted,UTF,PERL_DEBUG_PAD_ZERO(1),
+ RE_PV_QUOTED_DECL(quoted,utf8_target,PERL_DEBUG_PAD_ZERO(1),
s,strend-s,60);
PerlIO_printf(Perl_debug_log,
- "Matching stclass %.*s against %s (%d chars)\n",
+ "Matching stclass %.*s against %s (%d bytes)\n",
(int)SvCUR(prop), SvPVX_const(prop),
quoted, (int)(strend - s));
}
char *last;
SV* float_real;
- if (!(do_utf8 ? prog->float_utf8 : prog->float_substr))
- do_utf8 ? to_utf8_substr(prog) : to_byte_substr(prog);
- float_real = do_utf8 ? prog->float_utf8 : prog->float_substr;
+ 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) {
last = screaminstr(sv, float_real, s - strbeg,
if (!last)
last = scream_olds; /* Only one occurrence. */
/* we may be pointing at the wrong string */
- else if (RX_MATCH_COPIED(prog))
+ else if (RXp_MATCH_COPIED(prog))
s = strbeg + (s - SvPVX_const(sv));
}
else {
dontbother = minlen - 1;
strend -= dontbother; /* this one's always in bytes! */
/* We don't know much -- general case. */
- if (do_utf8) {
+ if (utf8_target) {
for (;;) {
if (regtry(®info, &s))
goto got_it;
goto phooey;
got_it:
- RX_MATCH_TAINTED_set(prog, PL_reg_flags & RF_tainted);
+ Safefree(swap);
+ RX_MATCH_TAINTED_set(rx, 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);
+ if (RXp_PAREN_NAMES(prog))
+ (void)hv_iterinit(RXp_PAREN_NAMES(prog));
/* make sure $`, $&, $', and $digit will work later */
if ( !(flags & REXEC_NOT_FIRST) ) {
- RX_MATCH_COPY_FREE(prog);
+ RX_MATCH_COPY_FREE(rx);
if (flags & REXEC_COPY_STR) {
const I32 i = PL_regeol - startpos + (stringarg - strbeg);
#ifdef PERL_OLD_COPY_ON_WRITE
} else
#endif
{
- RX_MATCH_COPIED_on(prog);
+ RX_MATCH_COPIED_on(rx);
s = savepvn(strbeg, i);
prog->subbeg = s;
}
PL_colors[4], PL_colors[5]));
if (PL_reg_eval_set)
restore_pos(aTHX_ prog);
- if (progi->swap) {
+ if (swap) {
/* we failed :-( roll it back */
- I32 *t;
- t = progi->swap->startp;
- progi->swap->startp = prog->startp;
- prog->startp = t;
- t = progi->swap->endp;
- progi->swap->endp = prog->endp;
- prog->endp = t;
+ Safefree(prog->offs);
+ prog->offs = swap;
}
+
return 0;
}
-
-
/*
- regtry - try match at specific point
*/
S_regtry(pTHX_ regmatch_info *reginfo, char **startpos)
{
dVAR;
- register I32 *sp;
- register I32 *ep;
CHECKPOINT lastcp;
- regexp *prog = reginfo->prog;
+ REGEXP *const rx = reginfo->prog;
+ regexp *const prog = (struct regexp *)SvANY(rx);
RXi_GET_DECL(prog,progi);
GET_RE_DEBUG_FLAGS_DECL;
+
+ PERL_ARGS_ASSERT_REGTRY;
+
reginfo->cutpoint=NULL;
if ((prog->extflags & RXf_EVAL_SEEN) && !PL_reg_eval_set) {
/* Make $_ available to executed code. */
if (reginfo->sv != DEFSV) {
SAVE_DEFSV;
- DEFSV = reginfo->sv;
+ DEFSV_set(reginfo->sv);
}
if (!(SvTYPE(reginfo->sv) >= SVt_PVMG && SvMAGIC(reginfo->sv)
Newxz(PL_reg_curpm, 1, PMOP);
#ifdef USE_ITHREADS
{
- SV* const repointer = newSViv(0);
- /* so we know which PL_regex_padav element is PL_reg_curpm */
- SvFLAGS(repointer) |= SVf_BREAK;
- av_push(PL_regex_padav,repointer);
+ SV* const repointer = &PL_sv_undef;
+ /* this regexp is also owned by the new PL_reg_curpm, which
+ will try to free it. */
+ av_push(PL_regex_padav, repointer);
PL_reg_curpm->op_pmoffset = av_len(PL_regex_padav);
PL_regex_pad = AvARRAY(PL_regex_padav);
}
#endif
}
- PM_SETRE(PL_reg_curpm, prog);
+#ifdef USE_ITHREADS
+ /* It seems that non-ithreads works both with and without this code.
+ So for efficiency reasons it seems best not to have the code
+ compiled when it is not needed. */
+ /* This is safe against NULLs: */
+ ReREFCNT_dec(PM_GETRE(PL_reg_curpm));
+ /* PM_reg_curpm owns a reference to this regexp. */
+ ReREFCNT_inc(rx);
+#endif
+ PM_SETRE(PL_reg_curpm, rx);
PL_reg_oldcurpm = PL_curpm;
PL_curpm = PL_reg_curpm;
- if (RX_MATCH_COPIED(prog)) {
+ if (RXp_MATCH_COPIED(prog)) {
/* Here is a serious problem: we cannot rewrite subbeg,
since it may be needed if this match fails. Thus
$` inside (?{}) could fail... */
#ifdef PERL_OLD_COPY_ON_WRITE
PL_nrs = prog->saved_copy;
#endif
- RX_MATCH_COPIED_off(prog);
+ RXp_MATCH_COPIED_off(prog);
}
else
PL_reg_oldsaved = NULL;
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)
/* 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
- * (op/regexp, op/pat, op/split), but that code is needed, oddly
- * enough, for building DynaLoader, or otherwise this
- * "Error: '*' not in typemap in DynaLoader.xs, line 164"
- * will happen. Meanwhile, this code *is* needed for the
+ * (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
* above-mentioned test suite tests to succeed. The common theme
* on those tests seems to be returning null fields from matches.
- * --jhi */
+ * --jhi updated by dapm */
#if 1
- sp = PL_regstartp;
- ep = PL_regendp;
if (prog->nparens) {
+ 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)
#define REPORT_CODE_OFF 32
-/* Make sure there is a test for this +1 options in re_tests */
-#define TRIE_INITAL_ACCEPT_BUFFLEN 4;
-
#define CHRTEST_UNINIT -1001 /* c1/c2 haven't been calculated yet */
#define CHRTEST_VOID -1000 /* the c1/c2 "next char" test should be skipped */
STATIC regmatch_state *
S_push_slab(pTHX)
{
-#if PERL_VERSION < 9
+#if PERL_VERSION < 9 && !defined(PERL_CORE)
dMY_CXT;
#endif
regmatch_slab *s = PL_regmatch_slab->next;
#define DEBUG_STATE_pp(pp) \
DEBUG_STATE_r({ \
- DUMP_EXEC_POS(locinput, scan, do_utf8); \
+ DUMP_EXEC_POS(locinput, scan, utf8_target); \
PerlIO_printf(Perl_debug_log, \
" %*s"pp" %s%s%s%s%s\n", \
depth*2, "", \
- reg_name[st->resume_state], \
+ PL_reg_name[st->resume_state], \
((st==yes_state||st==mark_state) ? "[" : ""), \
((st==yes_state) ? "Y" : ""), \
((st==mark_state) ? "M" : ""), \
#ifdef DEBUGGING
STATIC void
-S_debug_start_match(pTHX_ const regexp *prog, const bool do_utf8,
+S_debug_start_match(pTHX_ const REGEXP *prog, const bool utf8_target,
const char *start, const char *end, const char *blurb)
{
- const bool utf8_pat= prog->extflags & RXf_UTF8 ? 1 : 0;
+ const bool utf8_pat = RX_UTF8(prog) ? 1 : 0;
+
+ PERL_ARGS_ASSERT_DEBUG_START_MATCH;
+
if (!PL_colorset)
reginitcolors();
{
RE_PV_QUOTED_DECL(s0, utf8_pat, PERL_DEBUG_PAD_ZERO(0),
- prog->precomp, prog->prelen, 60);
+ RX_PRECOMP_const(prog), RX_PRELEN(prog), 60);
- RE_PV_QUOTED_DECL(s1, do_utf8, PERL_DEBUG_PAD_ZERO(1),
+ RE_PV_QUOTED_DECL(s1, utf8_target, PERL_DEBUG_PAD_ZERO(1),
start, end - start, 60);
PerlIO_printf(Perl_debug_log,
"%s%s REx%s %s against %s\n",
PL_colors[4], blurb, PL_colors[5], s0, s1);
- if (do_utf8||utf8_pat)
+ if (utf8_target||utf8_pat)
PerlIO_printf(Perl_debug_log, "UTF-8 %s%s%s...\n",
utf8_pat ? "pattern" : "",
- utf8_pat && do_utf8 ? " and " : "",
- do_utf8 ? "string" : ""
+ utf8_pat && utf8_target ? " and " : "",
+ utf8_target ? "string" : ""
);
}
}
const char *loc_regeol,
const char *loc_bostr,
const char *loc_reg_starttry,
- const bool do_utf8)
+ const bool utf8_target)
{
const int docolor = *PL_colors[0] || *PL_colors[2] || *PL_colors[4];
const int taill = (docolor ? 10 : 7); /* 3 chars for "> <" */
? (5 + taill) - l : locinput - loc_bostr;
int pref0_len;
- while (do_utf8 && UTF8_IS_CONTINUATION(*(U8*)(locinput - pref_len)))
+ PERL_ARGS_ASSERT_DUMP_EXEC_POS;
+
+ while (utf8_target && UTF8_IS_CONTINUATION(*(U8*)(locinput - pref_len)))
pref_len++;
pref0_len = pref_len - (locinput - loc_reg_starttry);
if (l + pref_len < (5 + taill) && l < loc_regeol - locinput)
l = ( loc_regeol - locinput > (5 + taill) - pref_len
? (5 + taill) - pref_len : loc_regeol - locinput);
- while (do_utf8 && UTF8_IS_CONTINUATION(*(U8*)(locinput + l)))
+ while (utf8_target && UTF8_IS_CONTINUATION(*(U8*)(locinput + l)))
l--;
if (pref0_len < 0)
pref0_len = 0;
if (pref0_len > pref_len)
pref0_len = pref_len;
{
- const int is_uni = (do_utf8 && OP(scan) != CANY) ? 1 : 0;
+ const int is_uni = (utf8_target && OP(scan) != CANY) ? 1 : 0;
RE_PV_COLOR_DECL(s0,len0,is_uni,PERL_DEBUG_PAD(0),
(locinput - pref_len),pref0_len, 60, 4, 5);
* or 0 if non of the buffers matched.
*/
STATIC I32
-S_reg_check_named_buff_matched(pTHX_ const regexp *rex, const regnode *scan) {
+S_reg_check_named_buff_matched(pTHX_ const regexp *rex, const regnode *scan)
+{
I32 n;
RXi_GET_DECL(rex,rexi);
- SV *sv_dat=(SV*)rexi->data->data[ ARG( scan ) ];
+ SV *sv_dat= MUTABLE_SV(rexi->data->data[ ARG( scan ) ]);
I32 *nums=(I32*)SvPVX(sv_dat);
+
+ PERL_ARGS_ASSERT_REG_CHECK_NAMED_BUFF_MATCHED;
+
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)
+
STATIC I32 /* 0 failure, 1 success */
S_regmatch(pTHX_ regmatch_info *reginfo, regnode *prog)
{
-#if PERL_VERSION < 9
+#if PERL_VERSION < 9 && !defined(PERL_CORE)
dMY_CXT;
#endif
dVAR;
- register const bool do_utf8 = PL_reg_match_utf8;
+ register const bool utf8_target = PL_reg_match_utf8;
const U32 uniflags = UTF8_ALLOW_DEFAULT;
-
- regexp *rex = reginfo->prog;
+ REGEXP *rex_sv = reginfo->prog;
+ regexp *rex = (struct regexp *)SvANY(rex_sv);
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;
-
/* cache heavy used fields of st in registers */
register regnode *scan;
register regnode *next;
bool result = 0; /* return value of S_regmatch */
int depth = 0; /* depth of backtrack stack */
- int nochange_depth = 0; /* depth of GOSUB recursion with nochange*/
+ U32 nochange_depth = 0; /* depth of GOSUB recursion with nochange */
+ const U32 max_nochange_depth =
+ (3 * rex->nparens > MAX_RECURSE_EVAL_NOCHANGE_DEPTH) ?
+ 3 * rex->nparens : MAX_RECURSE_EVAL_NOCHANGE_DEPTH;
regmatch_state *yes_state = NULL; /* state to pop to on success of
subpattern */
/* mark_state piggy backs on the yes_state logic so that when we unwind
the stack on success we can update the mark_state as we go */
regmatch_state *mark_state = NULL; /* last mark state we have seen */
-
regmatch_state *cur_eval = NULL; /* most recent EVAL_AB state */
struct regmatch_state *cur_curlyx = NULL; /* most recent curlyx */
U32 state_num;
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
* the very next op. They have a useful lifetime of exactly one loop
* iteration, and are not preserved or restored by state pushes/pops
false: plain (?=foo)
true: used as a condition: (?(?=foo))
*/
-
#ifdef DEBUGGING
GET_RE_DEBUG_FLAGS_DECL;
#endif
- DEBUG_OPTIMISE_r( {
+ PERL_ARGS_ASSERT_REGMATCH;
+
+ DEBUG_OPTIMISE_r( DEBUG_EXECUTE_r({
PerlIO_printf(Perl_debug_log,"regmatch start\n");
- });
+ }));
/* on first ever call to regmatch, allocate first slab */
if (!PL_regmatch_slab) {
Newx(PL_regmatch_slab, 1, regmatch_slab);
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;
DEBUG_EXECUTE_r( {
SV * const prop = sv_newmortal();
regnode *rnext=regnext(scan);
- DUMP_EXEC_POS( locinput, scan, do_utf8 );
+ DUMP_EXEC_POS( locinput, scan, utf8_target );
regprop(rex, prop, scan);
PerlIO_printf(Perl_debug_log,
state_num = OP(scan);
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_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:
case SANY:
if (!nextchr && locinput >= PL_regeol)
sayNO;
- if (do_utf8) {
+ if (utf8_target) {
locinput += PL_utf8skip[nextchr];
if (locinput > PL_regeol)
sayNO;
case REG_ANY:
if ((!nextchr && locinput >= PL_regeol) || nextchr == '\n')
sayNO;
- if (do_utf8) {
+ if (utf8_target) {
locinput += PL_utf8skip[nextchr];
if (locinput > PL_regeol)
sayNO;
/* In this case the charclass data is available inline so
we can fail fast without a lot of extra overhead.
*/
- if (scan->flags == EXACT || !do_utf8) {
+ if (scan->flags == EXACT || !utf8_target) {
if(!ANYOF_BITMAP_TEST(scan, *locinput)) {
DEBUG_EXECUTE_r(
PerlIO_printf(Perl_debug_log,
}
/* FALL THROUGH */
case TRIE:
+ /* the basic plan of execution of the trie is:
+ * At the beginning, run though all the states, and
+ * find the longest-matching word. Also remember the position
+ * of the shortest matching word. For example, this pattern:
+ * 1 2 3 4 5
+ * ab|a|x|abcd|abc
+ * when matched against the string "abcde", will generate
+ * accept states for all words except 3, with the longest
+ * matching word being 4, and the shortest being 1 (with
+ * the position being after char 1 of the string).
+ *
+ * Then for each matching word, in word order (i.e. 1,2,4,5),
+ * we run the remainder of the pattern; on each try setting
+ * the current position to the character following the word,
+ * returning to try the next word on failure.
+ *
+ * We avoid having to build a list of words at runtime by
+ * using a compile-time structure, wordinfo[].prev, which
+ * gives, for each word, the previous accepting word (if any).
+ * In the case above it would contain the mappings 1->2, 2->0,
+ * 3->0, 4->5, 5->1. We can use this table to generate, from
+ * the longest word (4 above), a list of all words, by
+ * following the list of prev pointers; this gives us the
+ * unordered list 4,5,1,2. Then given the current word we have
+ * just tried, we can go through the list and find the
+ * next-biggest word to try (so if we just failed on word 2,
+ * the next in the list is 4).
+ *
+ * Since at runtime we don't record the matching position in
+ * the string for each word, we have to work that out for
+ * each word we're about to process. The wordinfo table holds
+ * the character length of each word; given that we recorded
+ * at the start: the position of the shortest word and its
+ * length in chars, we just need to move the pointer the
+ * difference between the two char lengths. Depending on
+ * Unicode status and folding, that's cheap or expensive.
+ *
+ * This algorithm is optimised for the case where are only a
+ * small number of accept states, i.e. 0,1, or maybe 2.
+ * With lots of accepts states, and having to try all of them,
+ * it becomes quadratic on number of accept states to find all
+ * the next words.
+ */
+
{
/* what type of TRIE am I? (utf8 makes this contextual) */
- const enum { trie_plain, trie_utf8, trie_utf8_fold }
- trie_type = do_utf8 ?
- (scan->flags == EXACT ? trie_utf8 : trie_utf8_fold)
- : trie_plain;
+ DECL_TRIE_TYPE(scan);
/* what trie are we using right now */
reg_trie_data * const trie
= (reg_trie_data*)rexi->data->data[ ARG( scan ) ];
- HV * widecharmap = (HV *)rexi->data->data[ ARG( scan ) + 1 ];
+ HV * widecharmap = MUTABLE_HV(rexi->data->data[ ARG( scan ) + 1 ]);
U32 state = trie->startstate;
if (trie->bitmap && trie_type != trie_utf8_fold &&
STRLEN len = 0;
STRLEN foldlen = 0;
U8 *uscan = (U8*)NULL;
- STRLEN bufflen=0;
- SV *sv_accept_buff = NULL;
U8 foldbuf[ UTF8_MAXBYTES_CASE + 1 ];
+ U32 charcount = 0; /* how many input chars we have matched */
+ U32 accepted = 0; /* have we seen any accepting states? */
- ST.accepted = 0; /* how many accepting states we have seen */
ST.B = next;
ST.jump = trie->jump;
ST.me = scan;
-
- /*
- traverse the TRIE keeping track of all accepting states
- we transition through until we get to a failing node.
- */
+ ST.firstpos = NULL;
+ ST.longfold = FALSE; /* char longer if folded => it's harder */
+ ST.nextword = 0;
+
+ /* fully traverse the TRIE; note the position of the
+ shortest accept state and the wordnum of the longest
+ accept state */
while ( state && uc <= (U8*)PL_regeol ) {
U32 base = trie->states[ state ].trans.base;
UV uvc = 0;
U16 charid;
- /* We use charid to hold the wordnum as we don't use it
- for charid until after we have done the wordnum logic.
- We define an alias just so that the wordnum logic reads
- more naturally. */
-
-#define got_wordnum charid
- got_wordnum = trie->states[ state ].wordnum;
-
- if ( got_wordnum ) {
- if ( ! ST.accepted ) {
- ENTER;
- SAVETMPS;
- bufflen = TRIE_INITAL_ACCEPT_BUFFLEN;
- sv_accept_buff=newSV(bufflen *
- sizeof(reg_trie_accepted) - 1);
- SvCUR_set(sv_accept_buff, 0);
- SvPOK_on(sv_accept_buff);
- sv_2mortal(sv_accept_buff);
- SAVETMPS;
- ST.accept_buff =
- (reg_trie_accepted*)SvPV_nolen(sv_accept_buff );
- }
- do {
- if (ST.accepted >= bufflen) {
- bufflen *= 2;
- ST.accept_buff =(reg_trie_accepted*)
- SvGROW(sv_accept_buff,
- bufflen * sizeof(reg_trie_accepted));
+ U16 wordnum;
+ wordnum = trie->states[ state ].wordnum;
+
+ if (wordnum) { /* it's an accept state */
+ if (!accepted) {
+ accepted = 1;
+ /* record first match position */
+ if (ST.longfold) {
+ ST.firstpos = (U8*)locinput;
+ ST.firstchars = 0;
}
- SvCUR_set(sv_accept_buff,SvCUR(sv_accept_buff)
- + sizeof(reg_trie_accepted));
-
-
- ST.accept_buff[ST.accepted].wordnum = got_wordnum;
- ST.accept_buff[ST.accepted].endpos = uc;
- ++ST.accepted;
- } while (trie->nextword && (got_wordnum= trie->nextword[got_wordnum]));
+ else {
+ ST.firstpos = uc;
+ ST.firstchars = charcount;
+ }
+ }
+ if (!ST.nextword || wordnum < ST.nextword)
+ ST.nextword = wordnum;
+ ST.topword = wordnum;
}
-#undef got_wordnum
DEBUG_TRIE_EXECUTE_r({
- DUMP_EXEC_POS( (char *)uc, scan, do_utf8 );
+ DUMP_EXEC_POS( (char *)uc, scan, utf8_target );
PerlIO_printf( Perl_debug_log,
- "%*s %sState: %4"UVxf" Accepted: %4"UVxf" ",
+ "%*s %sState: %4"UVxf" Accepted: %c ",
2+depth * 2, "", PL_colors[4],
- (UV)state, (UV)ST.accepted );
+ (UV)state, (accepted ? 'Y' : 'N'));
});
+ /* read a char and goto next state */
if ( base ) {
+ I32 offset;
REXEC_TRIE_READ_CHAR(trie_type, trie, widecharmap, uc,
uscan, len, uvc, charid, foldlen,
foldbuf, uniflags);
-
+ charcount++;
+ if (foldlen>0)
+ ST.longfold = TRUE;
if (charid &&
- (base + charid > trie->uniquecharcount )
- && (base + charid - 1 - trie->uniquecharcount
- < trie->lasttrans)
- && trie->trans[base + charid - 1 -
- trie->uniquecharcount].check == state)
+ ( ((offset =
+ base + charid - 1 - trie->uniquecharcount)) >= 0)
+
+ && ((U32)offset < trie->lasttrans)
+ && trie->trans[offset].check == state)
{
- state = trie->trans[base + charid - 1 -
- trie->uniquecharcount ].next;
+ state = trie->trans[offset].next;
}
else {
state = 0;
charid, uvc, (UV)state, PL_colors[5] );
);
}
- if (!ST.accepted )
+ if (!accepted)
sayNO;
+ /* calculate total number of accept states */
+ {
+ U16 w = ST.topword;
+ accepted = 0;
+ while (w) {
+ w = trie->wordinfo[w].prev;
+ accepted++;
+ }
+ ST.accepted = accepted;
+ }
+
DEBUG_EXECUTE_r(
PerlIO_printf( Perl_debug_log,
"%*s %sgot %"IVdf" possible matches%s\n",
REPORT_CODE_OFF + depth * 2, "",
PL_colors[4], (IV)ST.accepted, PL_colors[5] );
);
+ goto trie_first_try; /* jump into the fail handler */
}}
+ /* NOTREACHED */
- /* FALL THROUGH */
- case TRIE_next_fail: /* we failed - try next alterative */
- if (do_cutgroup) {
- do_cutgroup = 0;
- no_final = 0;
- }
- if ( ST.accepted == 1 ) {
- /* only one choice left - just continue */
- DEBUG_EXECUTE_r({
- AV *const trie_words
- = (AV *) rexi->data->data[ARG(ST.me)+TRIE_WORDS_OFFSET];
- SV ** const tmp = av_fetch( trie_words,
- ST.accept_buff[ 0 ].wordnum-1, 0 );
- SV *sv= tmp ? sv_newmortal() : NULL;
-
- PerlIO_printf( Perl_debug_log,
- "%*s %sonly one match left: #%d <%s>%s\n",
- REPORT_CODE_OFF+depth*2, "", PL_colors[4],
- ST.accept_buff[ 0 ].wordnum,
- tmp ? pv_pretty(sv, SvPV_nolen_const(*tmp), SvCUR(*tmp), 0,
- PL_colors[0], PL_colors[1],
- (SvUTF8(*tmp) ? PERL_PV_ESCAPE_UNI : 0)
- )
- : "not compiled under -Dr",
- PL_colors[5] );
- });
- PL_reginput = (char *)ST.accept_buff[ 0 ].endpos;
- /* in this case we free tmps/leave before we call regmatch
- as we wont be using accept_buff again. */
-
- locinput = PL_reginput;
- nextchr = UCHARAT(locinput);
- if ( !ST.jump || !ST.jump[ST.accept_buff[0].wordnum])
- scan = ST.B;
- else
- scan = ST.me + ST.jump[ST.accept_buff[0].wordnum];
- if (!has_cutgroup) {
- FREETMPS;
- LEAVE;
- } else {
- ST.accepted--;
- PUSH_YES_STATE_GOTO(TRIE_next, scan);
- }
-
- continue; /* execute rest of RE */
+ 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;
}
-
- if (!ST.accepted-- ) {
+ if (!--ST.accepted) {
DEBUG_EXECUTE_r({
PerlIO_printf( Perl_debug_log,
"%*s %sTRIE failed...%s\n",
PL_colors[4],
PL_colors[5] );
});
- FREETMPS;
- LEAVE;
sayNO_SILENT;
}
+ {
+ /* Find next-highest word to process. Note that this code
+ * is O(N^2) per trie run (O(N) per branch), so keep tight */
+ register U16 min = 0;
+ register U16 word;
+ register U16 const nextword = ST.nextword;
+ register reg_trie_wordinfo * const wordinfo
+ = ((reg_trie_data*)rexi->data->data[ARG(ST.me)])->wordinfo;
+ for (word=ST.topword; word; word=wordinfo[word].prev) {
+ if (word > nextword && (!min || word < min))
+ min = word;
+ }
+ ST.nextword = min;
+ }
- /*
- There are at least two accepting states left. Presumably
- the number of accepting states is going to be low,
- typically two. So we simply scan through to find the one
- with lowest wordnum. Once we find it, we swap the last
- state into its place and decrement the size. We then try to
- match the rest of the pattern at the point where the word
- ends. If we succeed, control just continues along the
- regex; if we fail we return here to try the next accepting
- state
- */
+ trie_first_try:
+ if (do_cutgroup) {
+ do_cutgroup = 0;
+ no_final = 0;
+ }
- {
- U32 best = 0;
- U32 cur;
- for( cur = 1 ; cur <= ST.accepted ; cur++ ) {
- DEBUG_TRIE_EXECUTE_r(
- PerlIO_printf( Perl_debug_log,
- "%*s %sgot %"IVdf" (%d) as best, looking at %"IVdf" (%d)%s\n",
- REPORT_CODE_OFF + depth * 2, "", PL_colors[4],
- (IV)best, ST.accept_buff[ best ].wordnum, (IV)cur,
- ST.accept_buff[ cur ].wordnum, PL_colors[5] );
- );
+ if ( ST.jump) {
+ ST.lastparen = *PL_reglastparen;
+ REGCP_SET(ST.cp);
+ }
- if (ST.accept_buff[cur].wordnum <
- ST.accept_buff[best].wordnum)
- best = cur;
+ /* find start char of end of current word */
+ {
+ U32 chars; /* how many chars to skip */
+ U8 *uc = ST.firstpos;
+ reg_trie_data * const trie
+ = (reg_trie_data*)rexi->data->data[ARG(ST.me)];
+
+ assert((trie->wordinfo[ST.nextword].len - trie->prefixlen)
+ >= ST.firstchars);
+ chars = (trie->wordinfo[ST.nextword].len - trie->prefixlen)
+ - ST.firstchars;
+
+ if (ST.longfold) {
+ /* the hard option - fold each char in turn and find
+ * its folded length (which may be different */
+ U8 foldbuf[UTF8_MAXBYTES_CASE + 1];
+ STRLEN foldlen;
+ STRLEN len;
+ UV uvc;
+ U8 *uscan;
+
+ while (chars) {
+ if (utf8_target) {
+ uvc = utf8n_to_uvuni((U8*)uc, UTF8_MAXLEN, &len,
+ uniflags);
+ uc += len;
+ }
+ else {
+ uvc = *uc;
+ uc++;
+ }
+ uvc = to_uni_fold(uvc, foldbuf, &foldlen);
+ uscan = foldbuf;
+ while (foldlen) {
+ if (!--chars)
+ break;
+ uvc = utf8n_to_uvuni(uscan, UTF8_MAXLEN, &len,
+ uniflags);
+ uscan += len;
+ foldlen -= len;
+ }
+ }
+ }
+ else {
+ if (utf8_target)
+ while (chars--)
+ uc += UTF8SKIP(uc);
+ else
+ uc += chars;
}
+ PL_reginput = (char *)uc;
+ }
- DEBUG_EXECUTE_r({
- AV *const trie_words
- = (AV *) rexi->data->data[ARG(ST.me)+TRIE_WORDS_OFFSET];
- SV ** const tmp = av_fetch( trie_words,
- ST.accept_buff[ best ].wordnum - 1, 0 );
- regnode *nextop=(!ST.jump || !ST.jump[ST.accept_buff[best].wordnum]) ?
- ST.B :
- ST.me + ST.jump[ST.accept_buff[best].wordnum];
- SV *sv= tmp ? sv_newmortal() : NULL;
-
- PerlIO_printf( Perl_debug_log,
- "%*s %strying alternation #%d <%s> at node #%d %s\n",
- REPORT_CODE_OFF+depth*2, "", PL_colors[4],
- ST.accept_buff[best].wordnum,
- tmp ? pv_pretty(sv, SvPV_nolen_const(*tmp), SvCUR(*tmp), 0,
- PL_colors[0], PL_colors[1],
- (SvUTF8(*tmp) ? PERL_PV_ESCAPE_UNI : 0)
- ) : "not compiled under -Dr",
- REG_NODE_NUM(nextop),
- PL_colors[5] );
- });
+ scan = (ST.jump && ST.jump[ST.nextword])
+ ? ST.me + ST.jump[ST.nextword]
+ : ST.B;
- if ( best<ST.accepted ) {
- reg_trie_accepted tmp = ST.accept_buff[ best ];
- ST.accept_buff[ best ] = ST.accept_buff[ ST.accepted ];
- ST.accept_buff[ ST.accepted ] = tmp;
- best = ST.accepted;
- }
- PL_reginput = (char *)ST.accept_buff[ best ].endpos;
- if ( !ST.jump || !ST.jump[ST.accept_buff[best].wordnum]) {
- scan = ST.B;
- /* NOTREACHED */
- } else {
- scan = ST.me + ST.jump[ST.accept_buff[best].wordnum];
- /* NOTREACHED */
- }
- if (has_cutgroup) {
- PUSH_YES_STATE_GOTO(TRIE_next, scan);
- /* NOTREACHED */
- } else {
- PUSH_STATE_GOTO(TRIE_next, scan);
- /* NOTREACHED */
- }
- /* NOTREACHED */
+ DEBUG_EXECUTE_r({
+ PerlIO_printf( Perl_debug_log,
+ "%*s %sTRIE matched word #%d, continuing%s\n",
+ REPORT_CODE_OFF+depth*2, "",
+ PL_colors[4],
+ ST.nextword,
+ PL_colors[5]
+ );
+ });
+
+ if (ST.accepted > 1 || has_cutgroup) {
+ PUSH_STATE_GOTO(TRIE_next, scan);
+ /* NOTREACHED */
}
+ /* only one choice left - just continue */
+ DEBUG_EXECUTE_r({
+ AV *const trie_words
+ = MUTABLE_AV(rexi->data->data[ARG(ST.me)+TRIE_WORDS_OFFSET]);
+ SV ** const tmp = av_fetch( trie_words,
+ ST.nextword-1, 0 );
+ SV *sv= tmp ? sv_newmortal() : NULL;
+
+ PerlIO_printf( Perl_debug_log,
+ "%*s %sonly one match left, short-circuiting: #%d <%s>%s\n",
+ REPORT_CODE_OFF+depth*2, "", PL_colors[4],
+ ST.nextword,
+ tmp ? pv_pretty(sv, SvPV_nolen_const(*tmp), SvCUR(*tmp), 0,
+ PL_colors[0], PL_colors[1],
+ (SvUTF8(*tmp) ? PERL_PV_ESCAPE_UNI : 0)
+ )
+ : "not compiled under -Dr",
+ PL_colors[5] );
+ });
+
+ locinput = PL_reginput;
+ nextchr = UCHARAT(locinput);
+ continue; /* execute rest of RE */
/* NOTREACHED */
- case TRIE_next:
- FREETMPS;
- LEAVE;
- sayYES;
#undef ST
case EXACT: {
char *s = STRING(scan);
ln = STR_LEN(scan);
- if (do_utf8 != UTF) {
+ if (utf8_target != UTF_PATTERN) {
/* The target and the pattern have differing utf8ness. */
char *l = locinput;
const char * const e = s + ln;
- if (do_utf8) {
+ if (utf8_target) {
/* The target is utf8, the pattern is not utf8. */
while (s < e) {
STRLEN ulen;
char * const s = STRING(scan);
ln = STR_LEN(scan);
- if (do_utf8 || UTF) {
+ if (utf8_target || UTF_PATTERN) {
/* Either target or the pattern are utf8. */
const char * const l = locinput;
char *e = PL_regeol;
- if (ibcmp_utf8(s, 0, ln, (bool)UTF,
- l, &e, 0, do_utf8)) {
+ if (! foldEQ_utf8(s, 0, ln, cBOOL(UTF_PATTERN),
+ l, &e, 0, utf8_target)) {
/* One more case for the sharp s:
* 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' &&
+
+ if (!(utf8_target &&
+ toLOWER(s[0]) == 's' &&
ln >= 2 &&
toLOWER(s[1]) == 's' &&
(U8)l[0] == 0xC3 &&
if (PL_regeol - locinput < ln)
sayNO;
if (ln > 1 && (OP(scan) == EXACTF
- ? ibcmp(s, locinput, ln)
- : ibcmp_locale(s, locinput, ln)))
+ ? ! foldEQ(s, locinput, ln)
+ : ! foldEQ_locale(s, locinput, ln)))
sayNO;
locinput += ln;
nextchr = UCHARAT(locinput);
break;
}
- case ANYOF:
- if (do_utf8) {
- STRLEN inclasslen = PL_regeol - locinput;
-
- if (!reginclass(rex, scan, (U8*)locinput, &inclasslen, do_utf8))
- goto anyof_fail;
- if (locinput >= PL_regeol)
- sayNO;
- locinput += inclasslen ? inclasslen : UTF8SKIP(locinput);
- nextchr = UCHARAT(locinput);
- break;
- }
- else {
- if (nextchr < 0)
- nextchr = UCHARAT(locinput);
- if (!REGINCLASS(rex, scan, (U8*)locinput))
- goto anyof_fail;
- if (!nextchr && locinput >= PL_regeol)
- sayNO;
- nextchr = UCHARAT(++locinput);
- break;
- }
- anyof_fail:
- /* If we might have the case of the German sharp s
- * in a casefolding Unicode character class. */
-
- if (ANYOF_FOLD_SHARP_S(scan, locinput, PL_regeol)) {
- locinput += SHARP_S_SKIP;
- nextchr = UCHARAT(locinput);
- }
- else
- sayNO;
- break;
- case ALNUML:
- PL_reg_flags |= RF_tainted;
- /* FALL THROUGH */
- case ALNUM:
- if (!nextchr)
- sayNO;
- if (do_utf8) {
- LOAD_UTF8_CHARCLASS_ALNUM();
- if (!(OP(scan) == ALNUM
- ? (bool)swash_fetch(PL_utf8_alnum, (U8*)locinput, do_utf8)
- : isALNUM_LC_utf8((U8*)locinput)))
- {
- sayNO;
- }
- locinput += PL_utf8skip[nextchr];
- nextchr = UCHARAT(locinput);
- break;
- }
- if (!(OP(scan) == ALNUM
- ? isALNUM(nextchr) : isALNUM_LC(nextchr)))
- sayNO;
- nextchr = UCHARAT(++locinput);
- break;
- case NALNUML:
- PL_reg_flags |= RF_tainted;
- /* FALL THROUGH */
- case NALNUM:
- if (!nextchr && locinput >= PL_regeol)
- sayNO;
- if (do_utf8) {
- LOAD_UTF8_CHARCLASS_ALNUM();
- if (OP(scan) == NALNUM
- ? (bool)swash_fetch(PL_utf8_alnum, (U8*)locinput, do_utf8)
- : isALNUM_LC_utf8((U8*)locinput))
- {
- sayNO;
- }
- locinput += PL_utf8skip[nextchr];
- nextchr = UCHARAT(locinput);
- break;
- }
- if (OP(scan) == NALNUM
- ? isALNUM(nextchr) : isALNUM_LC(nextchr))
- sayNO;
- nextchr = UCHARAT(++locinput);
- break;
case BOUNDL:
case NBOUNDL:
PL_reg_flags |= RF_tainted;
case BOUND:
case NBOUND:
/* was last char in word? */
- if (do_utf8) {
+ if (utf8_target) {
if (locinput == PL_bostr)
ln = '\n';
else {
const U8 * const r = reghop3((U8*)locinput, -1, (U8*)PL_bostr);
-
+
ln = utf8n_to_uvchr(r, UTF8SKIP(r), 0, uniflags);
}
if (OP(scan) == BOUND || OP(scan) == NBOUND) {
ln = isALNUM_uni(ln);
LOAD_UTF8_CHARCLASS_ALNUM();
- n = swash_fetch(PL_utf8_alnum, (U8*)locinput, do_utf8);
+ n = swash_fetch(PL_utf8_alnum, (U8*)locinput, utf8_target);
}
else {
ln = isALNUM_LC_uvchr(UNI_TO_NATIVE(ln));
OP(scan) == BOUNDL))
sayNO;
break;
- case SPACEL:
- PL_reg_flags |= RF_tainted;
- /* FALL THROUGH */
- case SPACE:
- if (!nextchr)
- sayNO;
- if (do_utf8) {
- if (UTF8_IS_CONTINUED(nextchr)) {
- LOAD_UTF8_CHARCLASS_SPACE();
- if (!(OP(scan) == SPACE
- ? (bool)swash_fetch(PL_utf8_space, (U8*)locinput, do_utf8)
- : isSPACE_LC_utf8((U8*)locinput)))
- {
- sayNO;
- }
- locinput += PL_utf8skip[nextchr];
- nextchr = UCHARAT(locinput);
- break;
- }
- if (!(OP(scan) == SPACE
- ? isSPACE(nextchr) : isSPACE_LC(nextchr)))
+ case ANYOF:
+ if (utf8_target) {
+ STRLEN inclasslen = PL_regeol - locinput;
+
+ if (!reginclass(rex, scan, (U8*)locinput, &inclasslen, utf8_target))
+ goto anyof_fail;
+ if (locinput >= PL_regeol)
sayNO;
- nextchr = UCHARAT(++locinput);
+ locinput += inclasslen ? inclasslen : UTF8SKIP(locinput);
+ nextchr = UCHARAT(locinput);
+ break;
}
else {
- if (!(OP(scan) == SPACE
- ? isSPACE(nextchr) : isSPACE_LC(nextchr)))
+ if (nextchr < 0)
+ nextchr = UCHARAT(locinput);
+ if (!REGINCLASS(rex, scan, (U8*)locinput))
+ goto anyof_fail;
+ if (!nextchr && locinput >= PL_regeol)
sayNO;
nextchr = UCHARAT(++locinput);
- }
- break;
- case NSPACEL:
- PL_reg_flags |= RF_tainted;
- /* FALL THROUGH */
- case NSPACE:
- if (!nextchr && locinput >= PL_regeol)
- sayNO;
- if (do_utf8) {
- LOAD_UTF8_CHARCLASS_SPACE();
- if (OP(scan) == NSPACE
- ? (bool)swash_fetch(PL_utf8_space, (U8*)locinput, do_utf8)
- : isSPACE_LC_utf8((U8*)locinput))
- {
- sayNO;
- }
- locinput += PL_utf8skip[nextchr];
- nextchr = UCHARAT(locinput);
break;
}
- if (OP(scan) == NSPACE
- ? isSPACE(nextchr) : isSPACE_LC(nextchr))
- sayNO;
- nextchr = UCHARAT(++locinput);
- break;
- case DIGITL:
- PL_reg_flags |= RF_tainted;
- /* FALL THROUGH */
- case DIGIT:
- if (!nextchr)
- sayNO;
- if (do_utf8) {
- LOAD_UTF8_CHARCLASS_DIGIT();
- if (!(OP(scan) == DIGIT
- ? (bool)swash_fetch(PL_utf8_digit, (U8*)locinput, do_utf8)
- : isDIGIT_LC_utf8((U8*)locinput)))
- {
- sayNO;
- }
- locinput += PL_utf8skip[nextchr];
- nextchr = UCHARAT(locinput);
- break;
+ anyof_fail:
+ /* If we might have the case of the German sharp s
+ * in a casefolding Unicode character class. */
+
+ if (ANYOF_FOLD_SHARP_S(scan, locinput, PL_regeol)) {
+ locinput += SHARP_S_SKIP;
+ nextchr = UCHARAT(locinput);
}
- if (!(OP(scan) == DIGIT
- ? isDIGIT(nextchr) : isDIGIT_LC(nextchr)))
- sayNO;
- nextchr = UCHARAT(++locinput);
+ else
+ sayNO;
break;
- case NDIGITL:
- PL_reg_flags |= RF_tainted;
- /* FALL THROUGH */
- case NDIGIT:
- if (!nextchr && locinput >= PL_regeol)
+ /* Special char classes - The defines start on line 129 or so */
+ CCC_TRY_AFF( ALNUM, ALNUML, perl_word, "a", isALNUM_LC_utf8, isALNUM, isALNUM_LC);
+ CCC_TRY_NEG(NALNUM, NALNUML, perl_word, "a", isALNUM_LC_utf8, isALNUM, isALNUM_LC);
+
+ CCC_TRY_AFF( SPACE, SPACEL, perl_space, " ", isSPACE_LC_utf8, isSPACE, isSPACE_LC);
+ CCC_TRY_NEG(NSPACE, NSPACEL, perl_space, " ", isSPACE_LC_utf8, isSPACE, isSPACE_LC);
+
+ CCC_TRY_AFF( DIGIT, DIGITL, posix_digit, "0", isDIGIT_LC_utf8, isDIGIT, isDIGIT_LC);
+ CCC_TRY_NEG(NDIGIT, NDIGITL, posix_digit, "0", isDIGIT_LC_utf8, isDIGIT, isDIGIT_LC);
+
+ case CLUMP: /* Match \X: logical Unicode character. This is defined as
+ a Unicode extended Grapheme Cluster */
+ /* From http://www.unicode.org/reports/tr29 (5.2 version). An
+ extended Grapheme Cluster is:
+
+ CR LF
+ | Prepend* Begin Extend*
+ | .
+
+ Begin is (Hangul-syllable | ! Control)
+ Extend is (Grapheme_Extend | Spacing_Mark)
+ Control is [ GCB_Control CR LF ]
+
+ The discussion below shows how the code for CLUMP is derived
+ from this regex. Note that most of these concepts are from
+ property values of the Grapheme Cluster Boundary (GCB) property.
+ No code point can have multiple property values for a given
+ property. Thus a code point in Prepend can't be in Control, but
+ it must be in !Control. This is why Control above includes
+ GCB_Control plus CR plus LF. The latter two are used in the GCB
+ property separately, and so can't be in GCB_Control, even though
+ they logically are controls. Control is not the same as gc=cc,
+ but includes format and other characters as well.
+
+ The Unicode definition of Hangul-syllable is:
+ L+
+ | (L* ( ( V | LV ) V* | LVT ) T*)
+ | T+
+ )
+ Each of these is a value for the GCB property, and hence must be
+ disjoint, so the order they are tested is immaterial, so the
+ above can safely be changed to
+ T+
+ | L+
+ | (L* ( LVT | ( V | LV ) V*) T*)
+
+ The last two terms can be combined like this:
+ L* ( L
+ | (( LVT | ( V | LV ) V*) T*))
+
+ And refactored into this:
+ L* (L | LVT T* | V V* T* | LV V* T*)
+
+ That means that if we have seen any L's at all we can quit
+ there, but if the next character is a LVT, a V or and LV we
+ should keep going.
+
+ There is a subtlety with Prepend* which showed up in testing.
+ Note that the Begin, and only the Begin is required in:
+ | Prepend* Begin Extend*
+ Also, Begin contains '! Control'. A Prepend must be a '!
+ Control', which means it must be a Begin. What it comes down to
+ is that if we match Prepend* and then find no suitable Begin
+ afterwards, that if we backtrack the last Prepend, that one will
+ be a suitable Begin.
+ */
+
+ if (locinput >= PL_regeol)
sayNO;
- if (do_utf8) {
- LOAD_UTF8_CHARCLASS_DIGIT();
- if (OP(scan) == NDIGIT
- ? (bool)swash_fetch(PL_utf8_digit, (U8*)locinput, do_utf8)
- : isDIGIT_LC_utf8((U8*)locinput))
- {
- sayNO;
+ if (! utf8_target) {
+
+ /* Match either CR LF or '.', as all the other possibilities
+ * require utf8 */
+ locinput++; /* Match the . or CR */
+ if (nextchr == '\r'
+ && locinput < PL_regeol
+ && UCHARAT(locinput) == '\n') locinput++;
+ }
+ else {
+
+ /* Utf8: See if is ( CR LF ); already know that locinput <
+ * PL_regeol, so locinput+1 is in bounds */
+ if (nextchr == '\r' && UCHARAT(locinput + 1) == '\n') {
+ locinput += 2;
}
- locinput += PL_utf8skip[nextchr];
- nextchr = UCHARAT(locinput);
- break;
+ else {
+ /* In case have to backtrack to beginning, then match '.' */
+ char *starting = locinput;
+
+ /* In case have to backtrack the last prepend */
+ char *previous_prepend = 0;
+
+ LOAD_UTF8_CHARCLASS_GCB();
+
+ /* Match (prepend)* */
+ while (locinput < PL_regeol
+ && swash_fetch(PL_utf8_X_prepend,
+ (U8*)locinput, utf8_target))
+ {
+ previous_prepend = locinput;
+ locinput += UTF8SKIP(locinput);
+ }
+
+ /* As noted above, if we matched a prepend character, but
+ * the next thing won't match, back off the last prepend we
+ * matched, as it is guaranteed to match the begin */
+ if (previous_prepend
+ && (locinput >= PL_regeol
+ || ! swash_fetch(PL_utf8_X_begin,
+ (U8*)locinput, utf8_target)))
+ {
+ locinput = previous_prepend;
+ }
+
+ /* Note that here we know PL_regeol > locinput, as we
+ * tested that upon input to this switch case, and if we
+ * moved locinput forward, we tested the result just above
+ * and it either passed, or we backed off so that it will
+ * now pass */
+ if (! swash_fetch(PL_utf8_X_begin, (U8*)locinput, utf8_target)) {
+
+ /* Here did not match the required 'Begin' in the
+ * second term. So just match the very first
+ * character, the '.' of the final term of the regex */
+ locinput = starting + UTF8SKIP(starting);
+ } else {
+
+ /* Here is the beginning of a character that can have
+ * an extender. It is either a hangul syllable, or a
+ * non-control */
+ if (swash_fetch(PL_utf8_X_non_hangul,
+ (U8*)locinput, utf8_target))
+ {
+
+ /* Here not a Hangul syllable, must be a
+ * ('! * Control') */
+ locinput += UTF8SKIP(locinput);
+ } else {
+
+ /* Here is a Hangul syllable. It can be composed
+ * of several individual characters. One
+ * possibility is T+ */
+ if (swash_fetch(PL_utf8_X_T,
+ (U8*)locinput, utf8_target))
+ {
+ while (locinput < PL_regeol
+ && swash_fetch(PL_utf8_X_T,
+ (U8*)locinput, utf8_target))
+ {
+ locinput += UTF8SKIP(locinput);
+ }
+ } else {
+
+ /* Here, not T+, but is a Hangul. That means
+ * it is one of the others: L, LV, LVT or V,
+ * and matches:
+ * L* (L | LVT T* | V V* T* | LV V* T*) */
+
+ /* Match L* */
+ while (locinput < PL_regeol
+ && swash_fetch(PL_utf8_X_L,
+ (U8*)locinput, utf8_target))
+ {
+ locinput += UTF8SKIP(locinput);
+ }
+
+ /* Here, have exhausted L*. If the next
+ * character is not an LV, LVT nor V, it means
+ * we had to have at least one L, so matches L+
+ * in the original equation, we have a complete
+ * hangul syllable. Are done. */
+
+ if (locinput < PL_regeol
+ && swash_fetch(PL_utf8_X_LV_LVT_V,
+ (U8*)locinput, utf8_target))
+ {
+
+ /* Otherwise keep going. Must be LV, LVT
+ * or V. See if LVT */
+ if (swash_fetch(PL_utf8_X_LVT,
+ (U8*)locinput, utf8_target))
+ {
+ locinput += UTF8SKIP(locinput);
+ } else {
+
+ /* Must be V or LV. Take it, then
+ * match V* */
+ locinput += UTF8SKIP(locinput);
+ while (locinput < PL_regeol
+ && swash_fetch(PL_utf8_X_V,
+ (U8*)locinput, utf8_target))
+ {
+ locinput += UTF8SKIP(locinput);
+ }
+ }
+
+ /* And any of LV, LVT, or V can be followed
+ * by T* */
+ while (locinput < PL_regeol
+ && swash_fetch(PL_utf8_X_T,
+ (U8*)locinput,
+ utf8_target))
+ {
+ locinput += UTF8SKIP(locinput);
+ }
+ }
+ }
+ }
+
+ /* Match any extender */
+ while (locinput < PL_regeol
+ && swash_fetch(PL_utf8_X_extend,
+ (U8*)locinput, utf8_target))
+ {
+ locinput += UTF8SKIP(locinput);
+ }
+ }
+ }
+ if (locinput > PL_regeol) sayNO;
}
- if (OP(scan) == NDIGIT
- ? isDIGIT(nextchr) : isDIGIT_LC(nextchr))
- sayNO;
- nextchr = UCHARAT(++locinput);
- break;
- case CLUMP:
- if (locinput >= PL_regeol)
- sayNO;
- if (do_utf8) {
- LOAD_UTF8_CHARCLASS_MARK();
- if (swash_fetch(PL_utf8_mark,(U8*)locinput, do_utf8))
- sayNO;
- locinput += PL_utf8skip[nextchr];
- while (locinput < PL_regeol &&
- swash_fetch(PL_utf8_mark,(U8*)locinput, do_utf8))
- locinput += UTF8SKIP(locinput);
- if (locinput > PL_regeol)
- sayNO;
- }
- else
- locinput++;
nextchr = UCHARAT(locinput);
break;
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 */
+ if (utf8_target && 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
? memNE(s, locinput, ln)
: (type == REFF
- ? ibcmp(s, locinput, ln)
- : ibcmp_locale(s, locinput, ln))))
+ ? ! foldEQ(s, locinput, ln)
+ : ! foldEQ_locale(s, locinput, ln))))
sayNO;
locinput += ln;
nextchr = UCHARAT(locinput);
#define ST st->u.eval
{
SV *ret;
+ REGEXP *re_sv;
regexp *re;
regexp_internal *rei;
regnode *startpoint;
case GOSTART:
- case GOSUB: /* /(...(?1))/ */
- if (cur_eval && cur_eval->locinput==locinput) {
+ case GOSUB: /* /(...(?1))/ /(...(?&foo))/ */
+ if (cur_eval && cur_eval->locinput==locinput) {
if (cur_eval->u.eval.close_paren == (U32)ARG(scan))
Perl_croak(aTHX_ "Infinite recursion in regex");
- if ( ++nochange_depth > MAX_RECURSE_EVAL_NOCHANGE_DEPTH )
+ if ( ++nochange_depth > max_nochange_depth )
Perl_croak(aTHX_
"Pattern subroutine nesting without pos change"
" exceeded limit in regex");
} else {
nochange_depth = 0;
}
+ re_sv = rex_sv;
re = rex;
rei = rexi;
- (void)ReREFCNT_inc(rex);
+ (void)ReREFCNT_inc(rex_sv);
if (OP(scan)==GOSUB) {
startpoint = scan + ARG2L(scan);
ST.close_paren = ARG(scan);
/* NOTREACHED */
case EVAL: /* /(?{A})B/ /(??{A})B/ and /(?(?{A})X|Y)B/ */
if (cur_eval && cur_eval->locinput==locinput) {
- if ( ++nochange_depth > MAX_RECURSE_EVAL_NOCHANGE_DEPTH )
+ if ( ++nochange_depth > max_nochange_depth )
Perl_croak(aTHX_ "EVAL without pos change exceeded limit in regex");
} else {
nochange_depth = 0;
}
- { regexp *ocurpm = PM_GETRE(PL_curpm);
- char *osubbeg = rex->subbeg;
- STRLEN osublen = rex->sublen;
{
/* execute the code in the {...} */
dSP;
OP_4tree * const oop = PL_op;
COP * const ocurcop = PL_curcop;
PAD *old_comppad;
-
+ char *saved_regeol = PL_regeol;
n = ARG(scan);
PL_op = (OP_4tree*)rexi->data->data[n];
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);
sv_setsv(sv_mrk, sv_yes_mark);
}
- /* make sure that $1 and friends are available with nested eval */
- PM_SETRE(PL_curpm,rex);
- rex->subbeg = ocurpm->subbeg;
- rex->sublen = ocurpm->sublen;
CALLRUNOPS(aTHX); /* Scalar context. */
SPAGAIN;
PL_op = oop;
PAD_RESTORE_LOCAL(old_comppad);
PL_curcop = ocurcop;
-
+ PL_regeol = saved_regeol;
if (!logical) {
/* /(?{...})/ */
sv_setsv(save_scalar(PL_replgv), ret);
{
/* extract RE object from returned value; compiling if
* necessary */
-
MAGIC *mg = NULL;
- const SV *sv;
- if(SvROK(ret) && SvSMAGICAL(sv = SvRV(ret)))
- mg = mg_find(sv, PERL_MAGIC_qr);
- else if (SvSMAGICAL(ret)) {
- if (SvGMAGICAL(ret))
+ REGEXP *rx = NULL;
+
+ if (SvROK(ret)) {
+ SV *const sv = SvRV(ret);
+
+ if (SvTYPE(sv) == SVt_REGEXP) {
+ rx = (REGEXP*) sv;
+ } else if (SvSMAGICAL(sv)) {
+ mg = mg_find(sv, PERL_MAGIC_qr);
+ assert(mg);
+ }
+ } else if (SvTYPE(ret) == SVt_REGEXP) {
+ rx = (REGEXP*) ret;
+ } else if (SvSMAGICAL(ret)) {
+ if (SvGMAGICAL(ret)) {
+ /* I don't believe that there is ever qr magic
+ here. */
+ assert(!mg_find(ret, PERL_MAGIC_qr));
sv_unmagic(ret, PERL_MAGIC_qr);
- else
+ }
+ else {
mg = mg_find(ret, PERL_MAGIC_qr);
+ /* testing suggests mg only ends up non-NULL for
+ scalars who were upgraded and compiled in the
+ else block below. In turn, this is only
+ triggered in the "postponed utf8 string" tests
+ in t/op/pat.t */
+ }
}
if (mg) {
- re = (regexp *)mg->mg_obj;
- (void)ReREFCNT_inc(re);
+ rx = (REGEXP *) mg->mg_obj; /*XXX:dmq*/
+ assert(rx);
+ }
+ if (rx) {
+ rx = reg_temp_copy(NULL, rx);
}
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)) {
+ assert (SvUTF8(ret));
+ } else if (SvUTF8(ret)) {
+ /* Not doing UTF-8, despite what the SV says. Is
+ this only if we're trapped in use 'bytes'? */
+ /* Make a copy of the octet sequence, but without
+ the flag on, as the compiler now honours the
+ SvUTF8 flag on ret. */
+ STRLEN len;
+ const char *const p = SvPV(ret, len);
+ ret = newSVpvn_flags(p, len, SVs_TEMP);
+ }
+ rx = CALLREGCOMP(ret, pm_flags);
if (!(SvFLAGS(ret)
& (SVs_TEMP | SVs_PADTMP | SVf_READONLY
- | SVs_GMG)))
- sv_magic(ret,(SV*)ReREFCNT_inc(re),
- PERL_MAGIC_qr,0,0);
+ | SVs_GMG))) {
+ /* This isn't a first class regexp. Instead, it's
+ caching a regexp onto an existing, Perl visible
+ scalar. */
+ sv_magic(ret, MUTABLE_SV(rx), PERL_MAGIC_qr, 0, 0);
+ }
PL_regsize = osize;
}
+ re_sv = rx;
+ re = (struct regexp *)SvANY(rx);
}
+ RXp_MATCH_COPIED_off(re);
+ re->subbeg = rex->subbeg;
+ re->sublen = rex->sublen;
rei = RXi_GET(re);
-
- /* restore PL_curpm after the eval */
- PM_SETRE(PL_curpm,ocurpm);
- rex->sublen = osublen;
- rex->subbeg = osubbeg;
-
DEBUG_EXECUTE_r(
- debug_start_match(re, do_utf8, locinput, PL_regeol,
+ debug_start_match(re_sv, utf8_target, locinput, PL_regeol,
"Matching embedded");
);
startpoint = rei->program + 1;
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. */
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;
+ /* 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;
+
PL_reginput = locinput;
PL_regsize = 0;
PL_reg_maxiter = 0;
ST.toggle_reg_flags = PL_reg_flags;
- if (re->extflags & RXf_UTF8)
+ if (RX_UTF8(re_sv))
PL_reg_flags |= RF_utf8;
else
PL_reg_flags &= ~RF_utf8;
ST.toggle_reg_flags ^= PL_reg_flags; /* diff of old and new */
- ST.prev_rex = rex;
+ ST.prev_rex = rex_sv;
ST.prev_curlyx = cur_curlyx;
+ SETREX(rex_sv,re_sv);
rex = re;
rexi = rei;
cur_curlyx = NULL;
PUSH_YES_STATE_GOTO(EVAL_AB, startpoint);
/* NOTREACHED */
}
- /* restore PL_curpm after the eval */
- PM_SETRE(PL_curpm,ocurpm);
- rex->sublen = osublen;
- rex->subbeg = osubbeg;
- }
/* logical is 1, /(?(?{...})X|Y)/ */
- sw = (bool)SvTRUE(ret);
+ sw = cBOOL(SvTRUE(ret));
logical = 0;
break;
}
case EVAL_AB: /* cleanup after a successful (??{A})B */
/* note: this is called twice; first after popping B, then A */
PL_reg_flags ^= ST.toggle_reg_flags;
- ReREFCNT_dec(rex);
- rex = ST.prev_rex;
+ ReREFCNT_dec(rex_sv);
+ SETREX(rex_sv,ST.prev_rex);
+ rex = (struct regexp *)SvANY(rex_sv);
rexi = RXi_GET(rex);
regcpblow(ST.cp);
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 )
+ nochange_depth--;
sayYES;
case EVAL_AB_fail: /* unsuccessfully ran A or B in (??{A})B */
/* note: this is called twice; first after popping B, then A */
PL_reg_flags ^= ST.toggle_reg_flags;
- ReREFCNT_dec(rex);
- rex = ST.prev_rex;
+ ReREFCNT_dec(rex_sv);
+ 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);
regcppop(rex);
cur_curlyx = ST.prev_curlyx;
/* XXXX This is too dramatic a measure... */
PL_reg_maxiter = 0;
+ if ( nochange_depth )
+ nochange_depth--;
sayNO_SILENT;
#undef ST
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 = cBOOL(*PL_reglastparen >= n && PL_regoffs[n].end != -1);
break;
case NGROUPP:
/* reg_check_named_buff_matched returns 0 for no match */
- sw = (bool)(0 < reg_check_named_buff_matched(rex,scan));
+ sw = cBOOL(0 < reg_check_named_buff_matched(rex,scan));
break;
case INSUBP:
n = ARG(scan);
/* these fields contain the state of the current curly.
* they are accessed by subsequent WHILEMs */
ST.parenfloor = parenfloor;
- ST.min = ARG1(scan);
- ST.max = ARG2(scan);
- ST.A = NEXTOPER(scan) + EXTRA_STEP_2ARGS;
+ ST.me = scan;
ST.B = next;
ST.minmod = minmod;
minmod = 0;
}
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 */
{
/* see the discussion above about CURLYX/WHILEM */
I32 n;
+ int min = ARG1(cur_curlyx->u.curlyx.me);
+ int max = ARG2(cur_curlyx->u.curlyx.me);
+ regnode *A = NEXTOPER(cur_curlyx->u.curlyx.me) + EXTRA_STEP_2ARGS;
+
assert(cur_curlyx); /* keep Coverity happy */
n = ++cur_curlyx->u.curlyx.count; /* how many A's matched */
ST.save_lastloc = cur_curlyx->u.curlyx.lastloc;
PL_reginput = locinput;
DEBUG_EXECUTE_r( PerlIO_printf(Perl_debug_log,
- "%*s whilem: matched %ld out of %ld..%ld\n",
- REPORT_CODE_OFF+depth*2, "", (long)n,
- (long)cur_curlyx->u.curlyx.min,
- (long)cur_curlyx->u.curlyx.max)
+ "%*s whilem: matched %ld out of %d..%d\n",
+ REPORT_CODE_OFF+depth*2, "", (long)n, min, max)
);
/* First just match a string of min A's. */
- if (n < cur_curlyx->u.curlyx.min) {
+ if (n < min) {
cur_curlyx->u.curlyx.lastloc = locinput;
- PUSH_STATE_GOTO(WHILEM_A_pre, cur_curlyx->u.curlyx.A);
+ PUSH_STATE_GOTO(WHILEM_A_pre, A);
/* NOTREACHED */
}
/* Prefer A over B for maximal matching. */
- if (n < cur_curlyx->u.curlyx.max) { /* More greed allowed? */
+ if (n < max) { /* More greed allowed? */
ST.cp = regcppush(cur_curlyx->u.curlyx.parenfloor);
cur_curlyx->u.curlyx.lastloc = locinput;
REGCP_SET(ST.lastcp);
- PUSH_STATE_GOTO(WHILEM_A_max, cur_curlyx->u.curlyx.A);
+ PUSH_STATE_GOTO(WHILEM_A_max, A);
/* NOTREACHED */
}
goto do_whilem_B_max;
REGCP_UNWIND(ST.lastcp);
regcppop(rex);
- if (cur_curlyx->u.curlyx.count >= cur_curlyx->u.curlyx.max) {
+ if (cur_curlyx->u.curlyx.count >= /*max*/ARG2(cur_curlyx->u.curlyx.me)) {
/* Maximum greed exceeded */
if (cur_curlyx->u.curlyx.count >= REG_INFTY
&& ckWARN(WARN_REGEXP)
cur_curlyx->u.curlyx.lastloc = locinput;
ST.cp = regcppush(cur_curlyx->u.curlyx.parenfloor);
REGCP_SET(ST.lastcp);
- PUSH_STATE_GOTO(WHILEM_A_min, ST.save_curlyx->u.curlyx.A);
+ PUSH_STATE_GOTO(WHILEM_A_min,
+ /*A*/ NEXTOPER(ST.save_curlyx->u.curlyx.me) + EXTRA_STEP_2ARGS);
/* NOTREACHED */
#undef ST
case BRANCH: /* /(...|A|...)/ */
scan = NEXTOPER(scan); /* scan now points to inner node */
- if ((!next || (OP(next) != BRANCH && OP(next) != BRANCHJ))
- && !has_cutgroup)
- {
- /* last branch; skip state push and jump direct to node */
- continue;
- }
ST.lastparen = *PL_reglastparen;
ST.next_branch = next;
REGCP_SET(ST.cp);
case CUTGROUP:
PL_reginput = locinput;
sv_yes_mark = st->u.mark.mark_name = scan->flags ? NULL :
- (SV*)rexi->data->data[ ARG( scan ) ];
+ MUTABLE_SV(rexi->data->data[ ARG( scan ) ]);
PUSH_STATE_GOTO(CUTGROUP_next,next);
/* NOTREACHED */
case CUTGROUP_next_fail:
}
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;
case CURLYM: /* /A{m,n}B/ where A is fixed-length */
/* This is an optimisation of CURLYX that enables us to push
- * only a single backtracking state, no matter now many matches
+ * only a single backtracking state, no matter how many matches
* there are in {m,n}. It relies on the pattern being constant
* length, with no parens to influence future backrefs
*/
cur_eval->u.eval.close_paren == (U32)ST.me->flags)
goto fake_end;
- if ( ST.count < (ST.minmod ? ARG1(ST.me) : ARG2(ST.me)) )
- goto curlym_do_A; /* try to match another A */
+ {
+ I32 max = (ST.minmod ? ARG1(ST.me) : ARG2(ST.me));
+ if ( max == REG_INFTY || ST.count < max )
+ goto curlym_do_A; /* try to match another A */
+ }
goto curlym_do_B; /* try to match B */
case CURLYM_A_fail: /* just failed to match an A */
&& UCHARAT(PL_reginput) != ST.c2)
{
/* simulate B failing */
+ DEBUG_OPTIMISE_r(
+ PerlIO_printf(Perl_debug_log,
+ "%*s CURLYM Fast bail c1=%"IVdf" c2=%"IVdf"\n",
+ (int)(REPORT_CODE_OFF+(depth*2)),"",
+ (IV)ST.c1,(IV)ST.c2
+ ));
state_num = CURLYM_B_fail;
goto reenter_switch;
}
/* 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)
{
case CURLYM_B_fail: /* just failed to match a B */
REGCP_UNWIND(ST.cp);
if (ST.minmod) {
- if (ST.count == ARG2(ST.me) /* max */)
+ I32 max = ARG2(ST.me);
+ if (max != REG_INFTY && ST.count == max)
sayNO;
goto curlym_do_A; /* try to match a further A */
}
#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 */
if this changes back then the macro for IS_TEXT and
friends need to change. */
- if (!UTF) {
+ if (!UTF_PATTERN) {
ST.c2 = ST.c1 = *s;
if (IS_TEXTF(text_node))
ST.c2 = PL_fold[ST.c1];
else if (IS_TEXTFL(text_node))
ST.c2 = PL_fold_locale[ST.c1];
}
- else { /* UTF */
+ else { /* UTF_PATTERN */
if (IS_TEXTF(text_node)) {
STRLEN ulen1, ulen2;
U8 tmpbuf1[UTF8_MAXBYTES_CASE+1];
* string that could possibly match */
if (ST.max == REG_INFTY) {
ST.maxpos = PL_regeol - 1;
- if (do_utf8)
+ if (utf8_target)
while (UTF8_IS_CONTINUATION(*(U8*)ST.maxpos))
ST.maxpos--;
}
- else if (do_utf8) {
+ else if (utf8_target) {
int m = ST.max - ST.min;
for (ST.maxpos = locinput;
m >0 && ST.maxpos + UTF8SKIP(ST.maxpos) <= PL_regeol; m--)
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);
/* Couldn't or didn't -- move forward. */
ST.oldloc = locinput;
- if (do_utf8)
+ if (utf8_target)
locinput += UTF8SKIP(locinput);
else
locinput++;
/* find the next place where 'B' could work, then call B */
{
int n;
- if (do_utf8) {
+ if (utf8_target) {
n = (ST.oldloc == locinput) ? 0 : 1;
if (ST.c1 == ST.c2) {
STRLEN len;
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 */
{
UV c = 0;
if (ST.c1 != CHRTEST_VOID)
- c = do_utf8 ? utf8n_to_uvchr((U8*)PL_reginput,
+ c = utf8_target ? utf8n_to_uvchr((U8*)PL_reginput,
UTF8_MAXBYTES, 0, uniflags)
: (UV) UCHARAT(PL_reginput);
/* If it could work, try it. */
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. */
if (cur_eval) {
/* we've just finished A in /(??{A})B/; now continue with B */
I32 tmpix;
-
-
st->u.eval.toggle_reg_flags
= cur_eval->u.eval.toggle_reg_flags;
PL_reg_flags ^= st->u.eval.toggle_reg_flags;
- st->u.eval.prev_rex = rex; /* inner */
- rex = cur_eval->u.eval.prev_rex; /* outer */
+ st->u.eval.prev_rex = rex_sv; /* inner */
+ 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;
- ReREFCNT_inc(rex);
+ 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;
DEBUG_EXECUTE_r(
PerlIO_printf(Perl_debug_log, "%*s EVAL trying tail ... %"UVxf"\n",
REPORT_CODE_OFF+depth*2, "",PTR2UV(cur_eval)););
- PUSH_YES_STATE_GOTO(EVAL_AB,
+ if ( nochange_depth )
+ nochange_depth--;
+
+ PUSH_YES_STATE_GOTO(EVAL_AB,
st->u.eval.prev_eval->u.eval.B); /* match B */
}
/* trivial fail */
if (logical) {
logical = 0;
- sw = 1 - (bool)ST.wanted;
+ sw = 1 - cBOOL(ST.wanted);
}
else if (ST.wanted)
sayNO;
do_ifmatch:
ST.me = scan;
ST.logical = logical;
+ logical = 0; /* XXX: reset state of logical once it has been saved into ST */
+
/* execute body of (?...A) */
PUSH_YES_STATE_GOTO(IFMATCH_A, NEXTOPER(NEXTOPER(scan)));
/* NOTREACHED */
case IFMATCH_A: /* body of (?...A) succeeded */
if (ST.logical) {
- sw = (bool)ST.wanted;
+ sw = cBOOL(ST.wanted);
}
else if (!ST.wanted)
sayNO;
case PRUNE:
PL_reginput = locinput;
if (!scan->flags)
- sv_yes_mark = sv_commit = (SV*)rexi->data->data[ ARG( scan ) ];
+ sv_yes_mark = sv_commit = MUTABLE_SV(rexi->data->data[ ARG( scan ) ]);
PUSH_STATE_GOTO(COMMIT_next,next);
/* NOTREACHED */
case COMMIT_next_fail:
case MARKPOINT:
ST.prev_mark = mark_state;
ST.mark_name = sv_commit = sv_yes_mark
- = (SV*)rexi->data->data[ ARG( scan ) ];
+ = MUTABLE_SV(rexi->data->data[ ARG( scan ) ]);
mark_state = st;
ST.mark_loc = PL_reginput = locinput;
PUSH_YES_STATE_GOTO(MARKPOINT_next,next);
otherwise do nothing. Meaning we need to scan
*/
regmatch_state *cur = mark_state;
- SV *find = (SV*)rexi->data->data[ ARG( scan ) ];
+ SV *find = MUTABLE_SV(rexi->data->data[ ARG( scan ) ]);
while (cur) {
if ( sv_eq( cur->u.mark.mark_name,
sayNO;
/* NOTREACHED */
#undef ST
+ case FOLDCHAR:
+ n = ARG(scan);
+ if ( n == (U32)what_len_TRICKYFOLD(locinput,utf8_target,ln) ) {
+ locinput += ln;
+ } else if ( 0xDF == 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;
+ nextchr = UCHARAT(locinput);
+ } else
+ sayNO;
+ break;
+
+#define CASE_CLASS(nAmE) \
+ case nAmE: \
+ if ((n=is_##nAmE(locinput,utf8_target))) { \
+ locinput += n; \
+ nextchr = UCHARAT(locinput); \
+ } else \
+ sayNO; \
+ break; \
+ case N##nAmE: \
+ if ((n=is_##nAmE(locinput,utf8_target))) { \
+ 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",
}
PerlIO_printf(Perl_error_log, "%*s#%-3d %-10s %s\n",
REPORT_CODE_OFF + 2 + depth * 2,"",
- curd, reg_name[cur->resume_state],
+ curd, PL_reg_name[cur->resume_state],
(curyes == cur) ? "yes" : ""
);
if (curyes == cur)
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;
}
register I32 c;
register char *loceol = PL_regeol;
register I32 hardcount = 0;
- register bool do_utf8 = PL_reg_match_utf8;
+ register bool utf8_target = PL_reg_match_utf8;
+#ifndef DEBUGGING
+ PERL_UNUSED_ARG(depth);
+#endif
+
+ PERL_ARGS_ASSERT_REGREPEAT;
scan = PL_reginput;
if (max == REG_INFTY)
loceol = scan + max;
switch (OP(p)) {
case REG_ANY:
- if (do_utf8) {
+ if (utf8_target) {
loceol = PL_regeol;
while (scan < loceol && hardcount < max && *scan != '\n') {
scan += UTF8SKIP(scan);
}
break;
case SANY:
- if (do_utf8) {
+ if (utf8_target) {
loceol = PL_regeol;
while (scan < loceol && hardcount < max) {
scan += UTF8SKIP(scan);
scan++;
break;
case ANYOF:
- if (do_utf8) {
+ if (utf8_target) {
loceol = PL_regeol;
while (hardcount < max && scan < loceol &&
- reginclass(prog, p, (U8*)scan, 0, do_utf8)) {
+ reginclass(prog, p, (U8*)scan, 0, utf8_target)) {
scan += UTF8SKIP(scan);
hardcount++;
}
}
break;
case ALNUM:
- if (do_utf8) {
+ if (utf8_target) {
loceol = PL_regeol;
LOAD_UTF8_CHARCLASS_ALNUM();
while (hardcount < max && scan < loceol &&
- swash_fetch(PL_utf8_alnum, (U8*)scan, do_utf8)) {
+ swash_fetch(PL_utf8_alnum, (U8*)scan, utf8_target)) {
scan += UTF8SKIP(scan);
hardcount++;
}
break;
case ALNUML:
PL_reg_flags |= RF_tainted;
- if (do_utf8) {
+ if (utf8_target) {
loceol = PL_regeol;
while (hardcount < max && scan < loceol &&
isALNUM_LC_utf8((U8*)scan)) {
}
break;
case NALNUM:
- if (do_utf8) {
+ if (utf8_target) {
loceol = PL_regeol;
LOAD_UTF8_CHARCLASS_ALNUM();
while (hardcount < max && scan < loceol &&
- !swash_fetch(PL_utf8_alnum, (U8*)scan, do_utf8)) {
+ !swash_fetch(PL_utf8_alnum, (U8*)scan, utf8_target)) {
scan += UTF8SKIP(scan);
hardcount++;
}
break;
case NALNUML:
PL_reg_flags |= RF_tainted;
- if (do_utf8) {
+ if (utf8_target) {
loceol = PL_regeol;
while (hardcount < max && scan < loceol &&
!isALNUM_LC_utf8((U8*)scan)) {
}
break;
case SPACE:
- if (do_utf8) {
+ if (utf8_target) {
loceol = PL_regeol;
LOAD_UTF8_CHARCLASS_SPACE();
while (hardcount < max && scan < loceol &&
(*scan == ' ' ||
- swash_fetch(PL_utf8_space,(U8*)scan, do_utf8))) {
+ swash_fetch(PL_utf8_space,(U8*)scan, utf8_target))) {
scan += UTF8SKIP(scan);
hardcount++;
}
break;
case SPACEL:
PL_reg_flags |= RF_tainted;
- if (do_utf8) {
+ if (utf8_target) {
loceol = PL_regeol;
while (hardcount < max && scan < loceol &&
(*scan == ' ' || isSPACE_LC_utf8((U8*)scan))) {
}
break;
case NSPACE:
- if (do_utf8) {
+ if (utf8_target) {
loceol = PL_regeol;
LOAD_UTF8_CHARCLASS_SPACE();
while (hardcount < max && scan < loceol &&
!(*scan == ' ' ||
- swash_fetch(PL_utf8_space,(U8*)scan, do_utf8))) {
+ swash_fetch(PL_utf8_space,(U8*)scan, utf8_target))) {
scan += UTF8SKIP(scan);
hardcount++;
}
} else {
while (scan < loceol && !isSPACE(*scan))
scan++;
- break;
}
+ break;
case NSPACEL:
PL_reg_flags |= RF_tainted;
- if (do_utf8) {
+ if (utf8_target) {
loceol = PL_regeol;
while (hardcount < max && scan < loceol &&
!(*scan == ' ' || isSPACE_LC_utf8((U8*)scan))) {
}
break;
case DIGIT:
- if (do_utf8) {
+ if (utf8_target) {
loceol = PL_regeol;
LOAD_UTF8_CHARCLASS_DIGIT();
while (hardcount < max && scan < loceol &&
- swash_fetch(PL_utf8_digit, (U8*)scan, do_utf8)) {
+ swash_fetch(PL_utf8_digit, (U8*)scan, utf8_target)) {
scan += UTF8SKIP(scan);
hardcount++;
}
}
break;
case NDIGIT:
- if (do_utf8) {
+ if (utf8_target) {
loceol = PL_regeol;
LOAD_UTF8_CHARCLASS_DIGIT();
while (hardcount < max && scan < loceol &&
- !swash_fetch(PL_utf8_digit, (U8*)scan, do_utf8)) {
+ !swash_fetch(PL_utf8_digit, (U8*)scan, utf8_target)) {
scan += UTF8SKIP(scan);
hardcount++;
}
while (scan < loceol && !isDIGIT(*scan))
scan++;
}
+ case LNBREAK:
+ if (utf8_target) {
+ 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 (utf8_target) {
+ 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 (utf8_target) {
+ 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 (utf8_target) {
+ 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 (utf8_target) {
+ 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. */
}
RXi_GET_DECL(prog,progi);
const struct reg_data * const data = prog ? progi->data : NULL;
+ PERL_ARGS_ASSERT_REGCLASS_SWASH;
+
if (data && data->count) {
const U32 n = ARG(node);
if (data->what[n] == 's') {
- SV * const rv = (SV*)data->data[n];
- AV * const av = (AV*)SvRV((SV*)rv);
+ SV * const rv = MUTABLE_SV(data->data[n]);
+ AV * const av = MUTABLE_AV(SvRV(rv));
SV **const ary = AvARRAY(av);
SV **a, **b;
* documentation of these array elements. */
si = *ary;
- a = SvROK(ary[1]) ? &ary[1] : 0;
- b = SvTYPE(ary[2]) == SVt_PVAV ? &ary[2] : 0;
+ a = SvROK(ary[1]) ? &ary[1] : NULL;
+ b = SvTYPE(ary[2]) == SVt_PVAV ? &ary[2] : NULL;
if (a)
sw = *a;
The n is the ANYOF regnode, the p is the target string, lenp
is pointer to the maximum length of how far to go in the p
(if the lenp is zero, UTF8SKIP(p) is used),
- do_utf8 tells whether the target string is in UTF-8.
+ utf8_target tells whether the target string is in UTF-8.
*/
STATIC bool
-S_reginclass(pTHX_ const regexp *prog, register const regnode *n, register const U8* p, STRLEN* lenp, register bool do_utf8)
+S_reginclass(pTHX_ const regexp *prog, register const regnode *n, register const U8* p, STRLEN* lenp, register bool utf8_target)
{
dVAR;
const char flags = ANYOF_FLAGS(n);
STRLEN len = 0;
STRLEN plen;
- if (do_utf8 && !UTF8_IS_INVARIANT(c)) {
+ PERL_ARGS_ASSERT_REGINCLASS;
+
+ if (utf8_target && !UTF8_IS_INVARIANT(c)) {
c = utf8n_to_uvchr(p, UTF8_MAXBYTES, &len,
- (UTF8_ALLOW_DEFAULT & UTF8_ALLOW_ANYUV) | UTF8_CHECK_ONLY);
- /* see [perl #37836] for UTF8_ALLOW_ANYUV */
+ (UTF8_ALLOW_DEFAULT & UTF8_ALLOW_ANYUV)
+ | UTF8_ALLOW_FFFF | UTF8_CHECK_ONLY);
+ /* see [perl #37836] for UTF8_ALLOW_ANYUV; [perl #38293] for
+ * UTF8_ALLOW_FFFF */
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)) {
+ if (utf8_target || (flags & ANYOF_UNICODE)) {
if (lenp)
*lenp = 0;
- if (do_utf8 && !ANYOF_RUNTIME(n)) {
+ if (utf8_target && !ANYOF_RUNTIME(n)) {
if (len != (STRLEN)-1 && c < 256 && ANYOF_BITMAP_TEST(n, c))
match = TRUE;
}
- if (!match && do_utf8 && (flags & ANYOF_UNICODE_ALL) && c >= 256)
+ if (!match && utf8_target && (flags & ANYOF_UNICODE_ALL) && c >= 256)
match = TRUE;
if (!match) {
AV *av;
SV * const sw = regclass_swash(prog, n, TRUE, 0, (SV**)&av);
if (sw) {
- if (swash_fetch(sw, p, do_utf8))
+ U8 * utf8_p;
+ if (utf8_target) {
+ utf8_p = (U8 *) p;
+ } else {
+ STRLEN len = 1;
+ utf8_p = bytes_to_utf8(p, &len);
+ }
+ if (swash_fetch(sw, utf8_p, 1))
match = TRUE;
else if (flags & ANYOF_FOLD) {
if (!match && lenp && av) {
SV* const sv = *av_fetch(av, i, FALSE);
STRLEN len;
const char * const s = SvPV_const(sv, len);
-
- if (len <= plen && memEQ(s, (char*)p, len)) {
+ if (len <= plen && memEQ(s, (char*)utf8_p, len)) {
*lenp = len;
match = TRUE;
break;
}
if (!match) {
U8 tmpbuf[UTF8_MAXBYTES_CASE+1];
- STRLEN tmplen;
- to_utf8_fold(p, tmpbuf, &tmplen);
- if (swash_fetch(sw, tmpbuf, do_utf8))
+ STRLEN tmplen;
+ to_utf8_fold(utf8_p, tmpbuf, &tmplen);
+ if (swash_fetch(sw, tmpbuf, 1))
match = TRUE;
}
}
+
+ /* If we allocated a string above, free it */
+ if (! utf8_target) Safefree(utf8_p);
}
}
if (match && lenp && *lenp == 0)
S_reghop3(U8 *s, I32 off, const U8* lim)
{
dVAR;
+
+ PERL_ARGS_ASSERT_REGHOP3;
+
if (off >= 0) {
while (off-- && s < lim) {
/* XXX could check well-formedness here */
S_reghop4(U8 *s, I32 off, const U8* llim, const U8* rlim)
{
dVAR;
+
+ PERL_ARGS_ASSERT_REGHOP4;
+
if (off >= 0) {
while (off-- && s < rlim) {
/* XXX could check well-formedness here */
S_reghopmaybe3(U8* s, I32 off, const U8* lim)
{
dVAR;
+
+ PERL_ARGS_ASSERT_REGHOPMAYBE3;
+
if (off >= 0) {
while (off-- && s < lim) {
/* XXX could check well-formedness here */
#ifdef PERL_OLD_COPY_ON_WRITE
rex->saved_copy = PL_nrs;
#endif
- RX_MATCH_COPIED_on(rex);
+ RXp_MATCH_COPIED_on(rex);
}
PL_reg_magic->mg_len = PL_reg_oldpos;
PL_reg_eval_set = 0;
S_to_utf8_substr(pTHX_ register regexp *prog)
{
int i = 1;
+
+ PERL_ARGS_ASSERT_TO_UTF8_SUBSTR;
+
do {
if (prog->substrs->data[i].substr
&& !prog->substrs->data[i].utf8_substr) {
{
dVAR;
int i = 1;
+
+ PERL_ARGS_ASSERT_TO_BYTE_SUBSTR;
+
do {
if (prog->substrs->data[i].utf8_substr
&& !prog->substrs->data[i].substr) {