*/
/*
- * "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
#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 (do_utf8 && 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 \
+ ? (bool)swash_fetch(CAT2(PL_utf8_,CLASS), (U8*)locinput, do_utf8) \
+ : 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 (do_utf8 && 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 \
+ ? (bool)swash_fetch(CAT2(PL_utf8_,CLASS), (U8*)locinput, do_utf8) \
+ : 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) */
{
/* 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"));
+ /* XXX Does the destruction order has to change with do_utf8? */
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);
+ SvREFCNT_dec(do_utf8 ? 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 */
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 ); \
uvc = (UV)*uc; \
len = 1; \
} \
- \
if (uvc < 256) { \
charid = trie->charmap[ uvc ]; \
} \
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, do_utf8),
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, do_utf8),
!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, do_utf8),
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, do_utf8)),
!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, do_utf8),
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, do_utf8),
!isDIGIT(*s)
);
case NDIGITL:
= (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
return s;
}
-static void
-S_swap_match_buff (pTHX_ regexp *prog)
-{
- regexp_paren_pair *t;
-
- PERL_ARGS_ASSERT_SWAP_MATCH_BUFF;
-
- if (!prog->swap) {
- /* 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.
- 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(prog->swap, (prog->nparens + 1), regexp_paren_pair);
- }
- t = prog->swap;
- prog->swap = prog->offs;
- prog->offs = t;
-}
-
/*
- regexec_flags - match a regexp against a string
I32 multiline;
RXi_GET_DECL(prog,progi);
regmatch_info reginfo; /* create some info to pass to regtry etc */
- bool swap_on_fail = 0;
+ regexp_paren_pair *swap = NULL;
GET_RE_DEBUG_FLAGS_DECL;
PERL_ARGS_ASSERT_REGEXEC_FLAGS;
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) == rx)) {
- swap_on_fail = 1;
- swap_match_buff(prog); /* do we need a save destructor here for
- eval dies? */
+ /* We have to be careful. If the previous successful match
+ 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.
+ */
+ 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;
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;
}
goto phooey;
got_it:
+ Safefree(swap);
RX_MATCH_TAINTED_set(rx, PL_reg_flags & RF_tainted);
if (PL_reg_eval_set)
PL_colors[4], PL_colors[5]));
if (PL_reg_eval_set)
restore_pos(aTHX_ prog);
- if (swap_on_fail)
+ if (swap) {
/* we failed :-( roll it back */
- swap_match_buff(prog);
-
+ Safefree(prog->offs);
+ prog->offs = swap;
+ }
+
return 0;
}
/* 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)
reginitcolors();
{
RE_PV_QUOTED_DECL(s0, utf8_pat, PERL_DEBUG_PAD_ZERO(0),
- RX_PRECOMP(prog), RX_PRELEN(prog), 60);
+ RX_PRECOMP_const(prog), RX_PRELEN(prog), 60);
RE_PV_QUOTED_DECL(s1, do_utf8, PERL_DEBUG_PAD_ZERO(1),
start, end - start, 60);
{
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;
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)
/* 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 &&
if ( got_wordnum ) {
if ( ! ST.accepted ) {
ENTER;
- /* SAVETMPS; */ /* XXX is this necessary? dmq */
+ SAVETMPS; /* XXX is this necessary? dmq */
bufflen = TRIE_INITAL_ACCEPT_BUFFLEN;
sv_accept_buff=newSV(bufflen *
sizeof(reg_trie_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];
+ = MUTABLE_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;
DEBUG_EXECUTE_r({
AV *const trie_words
- = (AV *) rexi->data->data[ARG(ST.me)+TRIE_WORDS_OFFSET];
+ = MUTABLE_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]) ?
}
/* NOTREACHED */
case TRIE_next:
+ /* we dont want to throw this away, see bug 57042*/
+ if (oreplsv != GvSV(PL_replgv))
+ sv_setsv(oreplsv, GvSV(PL_replgv));
FREETMPS;
LEAVE;
sayYES;
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;
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) {
OP(scan) == BOUNDL))
sayNO;
break;
- case SPACEL:
- PL_reg_flags |= RF_tainted;
- /* FALL THROUGH */
- case SPACE:
- if (!nextchr)
- sayNO;
+ case ANYOF:
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)))
+ STRLEN inclasslen = PL_regeol - locinput;
+
+ if (!reginclass(rex, scan, (U8*)locinput, &inclasslen, do_utf8))
+ 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 (! do_utf8) {
+
+ /* 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, do_utf8))
+ {
+ 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, do_utf8)))
+ {
+ 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, do_utf8)) {
+
+ /* 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, do_utf8))
+ {
+
+ /* 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, do_utf8))
+ {
+ while (locinput < PL_regeol
+ && swash_fetch(PL_utf8_X_T,
+ (U8*)locinput, do_utf8))
+ {
+ 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, do_utf8))
+ {
+ 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, do_utf8))
+ {
+
+ /* Otherwise keep going. Must be LV, LVT
+ * or V. See if LVT */
+ if (swash_fetch(PL_utf8_X_LVT,
+ (U8*)locinput, do_utf8))
+ {
+ 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, do_utf8))
+ {
+ 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,
+ do_utf8))
+ {
+ locinput += UTF8SKIP(locinput);
+ }
+ }
+ }
+ }
+
+ /* Match any extender */
+ while (locinput < PL_regeol
+ && swash_fetch(PL_utf8_X_extend,
+ (U8*)locinput, do_utf8))
+ {
+ 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;
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];
PL_op = oop;
PAD_RESTORE_LOCAL(old_comppad);
PL_curcop = ocurcop;
+ PL_regeol = saved_regeol;
if (!logical) {
/* /(?{...})/ */
sv_setsv(save_scalar(PL_replgv), ret);
assert(rx);
}
if (rx) {
- rx = reg_temp_copy(rx);
+ rx = reg_temp_copy(NULL, rx);
}
else {
U32 pm_flags = 0;
/* This isn't a first class regexp. Instead, it's
caching a regexp onto an existing, Perl visible
scalar. */
- sv_magic(ret, (SV*) rx, PERL_MAGIC_qr, 0, 0);
+ sv_magic(ret, MUTABLE_SV(rx), PERL_MAGIC_qr, 0, 0);
}
PL_regsize = osize;
}
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;
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;
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:
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 */
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 */
}
cur_curlyx = cur_eval->u.eval.prev_curlyx;
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;
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 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,
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;
if (do_utf8 && !UTF8_IS_INVARIANT(c)) {
c = utf8n_to_uvchr(p, UTF8_MAXBYTES, &len,
- (UTF8_ALLOW_DEFAULT & UTF8_ALLOW_ANYUV) | UTF8_CHECK_ONLY);
- /* see [perl #37836] for UTF8_ALLOW_ANYUV */
+ (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)");
}
SV * const sw = regclass_swash(prog, n, TRUE, 0, (SV**)&av);
if (sw) {
- if (swash_fetch(sw, p, do_utf8))
+ U8 * utf8_p;
+ if (do_utf8) {
+ 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 (! do_utf8) Safefree(utf8_p);
}
}
if (match && lenp && *lenp == 0)
https://perl5.git.perl.org / perl5.git / blobdiff