# include "regcomp.h"
#endif
-#define RF_tainted 1 /* tainted information used? */
+#define RF_tainted 1 /* tainted information used? e.g. locale */
#define RF_warned 2 /* warned about big count? */
#define RF_utf8 8 /* Pattern contains multibyte chars? */
#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
+#ifdef EBCDIC
+ /* Often 'str' is a hard-coded utf8 string instead of utfebcdic. so just
+ * skip the check on EBCDIC platforms */
+# define LOAD_UTF8_CHARCLASS(class,str) LOAD_UTF8_CHARCLASS_NO_CHECK(class)
+#else
+# 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); assert(CAT2(PL_utf8_,class)); LEAVE; } } STMT_END
+#endif
/* 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 throw_away; ENTER; save_re_context(); throw_away = CAT2(is_utf8_,class)((const U8*)" "); LEAVE; } } STMT_END
+ if (!CAT2(PL_utf8_,class)) { \
+ bool throw_away __attribute__unused__; \
+ ENTER; save_re_context(); \
+ throw_away = 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")
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 PLACEHOLDER /* Something for the preprocessor to grab onto */
/* The actual code for CCC_TRY, which uses several variables from the routine
(IV)prog->check_end_shift);
});
- if (flags & REXEC_SCREAM) {
+ if ((flags & REXEC_SCREAM) && SvSCREAM(sv)) {
I32 p = -1; /* Internal iterator of scream. */
I32 * const pp = data ? data->scream_pos : &p;
+ const MAGIC *mg;
+ bool found = FALSE;
+
+ assert(SvMAGICAL(sv));
+ mg = mg_find(sv, PERL_MAGIC_study);
+ assert(mg);
- if (PL_screamfirst[BmRARE(check)] >= 0
+ if (mg->mg_private == 1) {
+ found = ((U8 *)mg->mg_ptr)[BmRARE(check)] != (U8)~0;
+ } else if (mg->mg_private == 2) {
+ found = ((U16 *)mg->mg_ptr)[BmRARE(check)] != (U16)~0;
+ } else {
+ assert (mg->mg_private == 4);
+ found = ((U32 *)mg->mg_ptr)[BmRARE(check)] != (U32)~0;
+ }
+
+ if (found
|| ( BmRARE(check) == '\n'
&& (BmPREVIOUS(check) == SvCUR(check) - 1)
&& SvTAIL(check) ))
reginclass(prog, c, (U8*)s, &inclasslen, utf8_target));
}
else {
- while (s < strend) {
- STRLEN skip = 1;
-
- if (REGINCLASS(prog, c, (U8*)s) ||
- (ANYOF_FOLD_SHARP_S(c, s, strend) &&
- /* The assignment of 2 is intentional:
- * for the folded sharp s, the skip is 2. */
- (skip = SHARP_S_SKIP))) {
- if (tmp && (!reginfo || regtry(reginfo, &s)))
- goto got_it;
- else
- tmp = doevery;
- }
- else
- tmp = 1;
- s += skip;
- }
+ REXEC_FBC_CLASS_SCAN(REGINCLASS(prog, c, (U8*)s));
}
break;
case CANY:
folder = foldEQ_latin1; /* /a, except the sharp s one which */
goto do_exactf_non_utf8; /* isn't dealt with by these */
- case EXACTFU:
- if (UTF_PATTERN || utf8_target) {
- utf8_fold_flags = 0;
- goto do_exactf_utf8;
- }
- fold_array = PL_fold_latin1;
- folder = foldEQ_latin1;
- /* XXX This uses the full utf8 fold because if the pattern contains
- * 'ss' it could match LATIN_SMALL_LETTER SHARP_S in the string.
- * There could be a new node type, say EXACTFU_SS, which is
- * generated by regcomp only if there is an 'ss', and then every
- * other case could goto do_exactf_non_utf8;*/
- goto do_exactf_utf8;
-
case EXACTF:
if (UTF_PATTERN || utf8_target) {
- utf8_fold_flags = 0;
+
+ /* regcomp.c already folded this if pattern is in UTF-8 */
+ utf8_fold_flags = (UTF_PATTERN) ? FOLDEQ_S2_ALREADY_FOLDED : 0;
goto do_exactf_utf8;
}
fold_array = PL_fold;
case EXACTFL:
if (UTF_PATTERN || utf8_target) {
- utf8_fold_flags = 0; /* XXX, add new flag for locale */
+ utf8_fold_flags = FOLDEQ_UTF8_LOCALE;
goto do_exactf_utf8;
}
fold_array = PL_fold_locale;
folder = foldEQ_locale;
+ goto do_exactf_non_utf8;
+
+ case EXACTFU:
+ if (UTF_PATTERN || utf8_target) {
+ utf8_fold_flags = (UTF_PATTERN) ? FOLDEQ_S2_ALREADY_FOLDED : 0;
+ goto do_exactf_utf8;
+ }
+
+ /* Any 'ss' in the pattern should have been replaced by regcomp,
+ * so we don't have to worry here about this single special case
+ * in the Latin1 range */
+ fold_array = PL_fold_latin1;
+ folder = foldEQ_latin1;
/* FALL THROUGH */
pat_string = STRING(c);
ln = STR_LEN(c); /* length to match in octets/bytes */
+ /* We know that we have to match at least 'ln' bytes (which is the
+ * same as characters, since not utf8). If we have to match 3
+ * characters, and there are only 2 availabe, we know without
+ * trying that it will fail; so don't start a match past the
+ * required minimum number from the far end */
e = HOP3c(strend, -((I32)ln), s);
if (!reginfo && e < s) {
break;
do_exactf_utf8:
+ {
+ unsigned expansion;
+
/* If one of the operands is in utf8, we can't use the simpler
* folding above, due to the fact that many different characters
? utf8_length((U8 *) pat_string, (U8 *) pat_end)
: ln;
+ /* We have 'lnc' characters to match in the pattern, but because of
+ * multi-character folding, each character in the target can match
+ * up to 3 characters (Unicode guarantees it will never exceed
+ * this) if it is utf8-encoded; and up to 2 if not (based on the
+ * fact that the Latin 1 folds are already determined, and the
+ * only multi-char fold in that range is the sharp-s folding to
+ * 'ss'. Thus, a pattern character can match as little as 1/3 of a
+ * string character. Adjust lnc accordingly, always matching at
+ * least 1 */
+ expansion = (utf8_target) ? UTF8_MAX_FOLD_CHAR_EXPAND : 2;
+ lnc = (lnc < expansion) ? 1 : lnc / expansion;
+
+ /* As in the non-UTF8 case, if we have to match 3 characters, and
+ * only 2 are left, it's guaranteed to fail, so don't start a
+ * match that would require us to go beyond the end of the string
+ */
e = HOP3c(strend, -((I32)lnc), s);
if (!reginfo && e < s) {
e = s; /* Due to minlen logic of intuit() */
}
+ /* XXX Note that we could recalculate e every so-often through the
+ * loop to stop earlier, as the worst case expansion above will
+ * rarely be met, and as we go along we would usually find that e
+ * moves further to the left. Unclear if worth the expense */
+
while (s <= e) {
char *my_strend= (char *)strend;
if (foldEQ_utf8_flags(s, &my_strend, 0, utf8_target,
s += UTF8SKIP(s);
}
break;
+ }
case BOUNDL:
PL_reg_flags |= RF_tainted;
FBC_BOUND(isALNUM_LC,
break;
case ALNUMU:
REXEC_FBC_CSCAN_PRELOAD(
- LOAD_UTF8_CHARCLASS_PERL_WORD(),
- swash_fetch(RE_utf8_perl_word,(U8*)s, utf8_target),
+ LOAD_UTF8_CHARCLASS_ALNUM(),
+ swash_fetch(PL_utf8_alnum,(U8*)s, utf8_target),
isWORDCHAR_L1((U8) *s)
);
break;
case ALNUM:
REXEC_FBC_CSCAN_PRELOAD(
- LOAD_UTF8_CHARCLASS_PERL_WORD(),
- swash_fetch(RE_utf8_perl_word,(U8*)s, utf8_target),
+ LOAD_UTF8_CHARCLASS_ALNUM(),
+ swash_fetch(PL_utf8_alnum,(U8*)s, utf8_target),
isWORDCHAR((U8) *s)
);
break;
break;
case NALNUMU:
REXEC_FBC_CSCAN_PRELOAD(
- LOAD_UTF8_CHARCLASS_PERL_WORD(),
- swash_fetch(RE_utf8_perl_word,(U8*)s, utf8_target),
+ LOAD_UTF8_CHARCLASS_ALNUM(),
+ !swash_fetch(PL_utf8_alnum,(U8*)s, utf8_target),
! isWORDCHAR_L1((U8) *s)
);
break;
case NALNUM:
REXEC_FBC_CSCAN_PRELOAD(
- LOAD_UTF8_CHARCLASS_PERL_WORD(),
- !swash_fetch(RE_utf8_perl_word, (U8*)s, utf8_target),
+ LOAD_UTF8_CHARCLASS_ALNUM(),
+ !swash_fetch(PL_utf8_alnum, (U8*)s, utf8_target),
! isALNUM(*s)
);
break;
break;
case SPACEU:
REXEC_FBC_CSCAN_PRELOAD(
- LOAD_UTF8_CHARCLASS_PERL_SPACE(),
- *s == ' ' || swash_fetch(RE_utf8_perl_space,(U8*)s, utf8_target),
+ LOAD_UTF8_CHARCLASS_SPACE(),
+ *s == ' ' || swash_fetch(PL_utf8_space,(U8*)s, utf8_target),
isSPACE_L1((U8) *s)
);
break;
case SPACE:
REXEC_FBC_CSCAN_PRELOAD(
- LOAD_UTF8_CHARCLASS_PERL_SPACE(),
- *s == ' ' || swash_fetch(RE_utf8_perl_space,(U8*)s, utf8_target),
+ LOAD_UTF8_CHARCLASS_SPACE(),
+ *s == ' ' || swash_fetch(PL_utf8_space,(U8*)s, utf8_target),
isSPACE((U8) *s)
);
break;
break;
case NSPACEU:
REXEC_FBC_CSCAN_PRELOAD(
- LOAD_UTF8_CHARCLASS_PERL_SPACE(),
- !( *s == ' ' || swash_fetch(RE_utf8_perl_space,(U8*)s, utf8_target)),
+ LOAD_UTF8_CHARCLASS_SPACE(),
+ !( *s == ' ' || swash_fetch(PL_utf8_space,(U8*)s, utf8_target)),
! isSPACE_L1((U8) *s)
);
break;
case NSPACE:
REXEC_FBC_CSCAN_PRELOAD(
- LOAD_UTF8_CHARCLASS_PERL_SPACE(),
- !(*s == ' ' || swash_fetch(RE_utf8_perl_space,(U8*)s, utf8_target)),
+ LOAD_UTF8_CHARCLASS_SPACE(),
+ !(*s == ' ' || swash_fetch(PL_utf8_space,(U8*)s, utf8_target)),
! isSPACE((U8) *s)
);
break;
break;
case DIGIT:
REXEC_FBC_CSCAN_PRELOAD(
- LOAD_UTF8_CHARCLASS_POSIX_DIGIT(),
- swash_fetch(RE_utf8_posix_digit,(U8*)s, utf8_target),
+ LOAD_UTF8_CHARCLASS_DIGIT(),
+ swash_fetch(PL_utf8_digit,(U8*)s, utf8_target),
isDIGIT(*s)
);
break;
break;
case NDIGIT:
REXEC_FBC_CSCAN_PRELOAD(
- LOAD_UTF8_CHARCLASS_POSIX_DIGIT(),
- !swash_fetch(RE_utf8_posix_digit,(U8*)s, utf8_target),
+ LOAD_UTF8_CHARCLASS_DIGIT(),
+ !swash_fetch(PL_utf8_digit,(U8*)s, utf8_target),
!isDIGIT(*s)
);
break;
dontbother = end_shift;
strend = HOPc(strend, -dontbother);
while ( (s <= last) &&
- ((flags & REXEC_SCREAM)
+ ((flags & REXEC_SCREAM) && SvSCREAM(sv)
? (s = screaminstr(sv, must, HOP3c(s, back_min, (back_min<0 ? strbeg : strend)) - strbeg,
end_shift, &scream_pos, 0))
: (s = fbm_instr((unsigned char*)HOP3(s, back_min, (back_min<0 ? strbeg : strend)),
utf8_target ? to_utf8_substr(prog) : to_byte_substr(prog);
float_real = utf8_target ? prog->float_utf8 : prog->float_substr;
- if (flags & REXEC_SCREAM) {
+ if ((flags & REXEC_SCREAM) && SvSCREAM(sv)) {
last = screaminstr(sv, float_real, s - strbeg,
end_shift, &scream_pos, 1); /* last one */
if (!last)
/* This is safe against NULLs: */
ReREFCNT_dec(PM_GETRE(PL_reg_curpm));
/* PM_reg_curpm owns a reference to this regexp. */
- ReREFCNT_inc(rx);
+ (void)ReREFCNT_inc(rx);
#endif
PM_SETRE(PL_reg_curpm, rx);
PL_reg_oldcurpm = PL_curpm;
PL_reg_flags |= RF_tainted;
folder = foldEQ_locale;
fold_array = PL_fold_locale;
- fold_utf8_flags = 0;
+ fold_utf8_flags = FOLDEQ_UTF8_LOCALE;
goto do_exactf;
case EXACTFU:
folder = foldEQ_latin1;
fold_array = PL_fold_latin1;
- fold_utf8_flags = 0;
+ fold_utf8_flags = (UTF_PATTERN) ? FOLDEQ_S1_ALREADY_FOLDED : 0;
goto do_exactf;
case EXACTFA:
folder = foldEQ_latin1;
fold_array = PL_fold_latin1;
- fold_utf8_flags = 1;
+ fold_utf8_flags = FOLDEQ_UTF8_NOMIX_ASCII;
goto do_exactf;
case EXACTF:
folder = foldEQ;
fold_array = PL_fold;
- fold_utf8_flags = 0;
+ fold_utf8_flags = (UTF_PATTERN) ? FOLDEQ_S1_ALREADY_FOLDED : 0;
do_exactf:
s = STRING(scan);
char *e = PL_regeol;
if (! foldEQ_utf8_flags(s, 0, ln, cBOOL(UTF_PATTERN),
- l, &e, 0, utf8_target, fold_utf8_flags)) {
- /* 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 (!(utf8_target &&
- toLOWER(s[0]) == 's' &&
- ln >= 2 &&
- toLOWER(s[1]) == 's' &&
- (U8)l[0] == 0xC3 &&
- e - l >= 2 &&
- (U8)l[1] == 0x9F))
- sayNO;
+ l, &e, 0, utf8_target, fold_utf8_flags))
+ {
+ sayNO;
}
locinput = e;
nextchr = UCHARAT(locinput);
case NBOUNDU:
case NBOUNDA:
/* was last char in word? */
- if (utf8_target && FLAGS(scan) != REGEX_ASCII_RESTRICTED_CHARSET) {
+ if (utf8_target
+ && FLAGS(scan) != REGEX_ASCII_RESTRICTED_CHARSET
+ && FLAGS(scan) != REGEX_ASCII_MORE_RESTRICTED_CHARSET)
+ {
if (locinput == PL_bostr)
ln = '\n';
else {
n = isALNUM(nextchr);
break;
case REGEX_ASCII_RESTRICTED_CHARSET:
+ case REGEX_ASCII_MORE_RESTRICTED_CHARSET:
ln = isWORDCHAR_A(ln);
n = isWORDCHAR_A(nextchr);
break;
ALNUML, NALNUML, isALNUM_LC, isALNUM_LC_utf8,
ALNUMU, NALNUMU, isWORDCHAR_L1,
ALNUMA, NALNUMA, isWORDCHAR_A,
- perl_word, "a");
+ alnum, "a");
CCC_TRY_U(SPACE, NSPACE, isSPACE,
SPACEL, NSPACEL, isSPACE_LC, isSPACE_LC_utf8,
SPACEU, NSPACEU, isSPACE_L1,
SPACEA, NSPACEA, isSPACE_A,
- perl_space, " ");
+ space, " ");
CCC_TRY(DIGIT, NDIGIT, isDIGIT,
DIGITL, NDIGITL, isDIGIT_LC, isDIGIT_LC_utf8,
DIGITA, NDIGITA, isDIGIT_A,
- posix_digit, "0");
+ digit, "0");
case CLUMP: /* Match \X: logical Unicode character. This is defined as
a Unicode extended Grapheme Cluster */
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
+ there, but if the next character is an LVT, a V, or an 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.
+ Also, Begin contains '! Control'. A Prepend must be a
+ '! Control', which means it must also 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)
/* Match either CR LF or '.', as all the other possibilities
* require utf8 */
locinput++; /* Match the . or CR */
- if (nextchr == '\r'
+ if (nextchr == '\r' /* And if it was CR, and the next is LF,
+ match the LF */
&& locinput < PL_regeol
&& UCHARAT(locinput) == '\n') locinput++;
}
folder = foldEQ_locale;
fold_array = PL_fold_locale;
type = REFFL;
- utf8_fold_flags = 0;
+ utf8_fold_flags = FOLDEQ_UTF8_LOCALE;
goto do_nref;
case NREFFA:
PL_reg_flags |= RF_tainted;
folder = foldEQ_locale;
fold_array = PL_fold_locale;
- utf8_fold_flags = 0;
+ utf8_fold_flags = FOLDEQ_UTF8_LOCALE;
goto do_ref;
case REFFA:
PL_op = (OP_4tree*)rexi->data->data[n];
DEBUG_STATE_r( PerlIO_printf(Perl_debug_log,
" re_eval 0x%"UVxf"\n", PTR2UV(PL_op)) );
+ /* wrap the call in two SAVECOMPPADs. This ensures that
+ * when the save stack is eventually unwound, all the
+ * accumulated SAVEt_CLEARSV's will be processed with
+ * interspersed SAVEt_COMPPAD's to ensure that lexicals
+ * are cleared in the right pad */
+ SAVECOMPPAD();
PAD_SAVE_LOCAL(old_comppad, (PAD*)rexi->data->data[n + 2]);
PL_regoffs[0].end = PL_reg_magic->mg_len = locinput - PL_bostr;
Copy(&saved_state, &PL_reg_state, 1, struct re_save_state);
PL_op = oop;
+ SAVECOMPPAD();
PAD_RESTORE_LOCAL(old_comppad);
PL_curcop = ocurcop;
PL_regeol = saved_regeol;
/* First just match a string of min A's. */
if (n < min) {
+ ST.cp = regcppush(cur_curlyx->u.curlyx.parenfloor);
cur_curlyx->u.curlyx.lastloc = locinput;
+ REGCP_SET(ST.lastcp);
+
PUSH_STATE_GOTO(WHILEM_A_pre, A);
/* NOTREACHED */
}
/* NOTREACHED */
case WHILEM_A_min_fail: /* just failed to match A in a minimal match */
- REGCP_UNWIND(ST.lastcp);
- regcppop(rex);
/* FALL THROUGH */
case WHILEM_A_pre_fail: /* just failed to match even minimal A */
+ REGCP_UNWIND(ST.lastcp);
+ regcppop(rex);
cur_curlyx->u.curlyx.lastloc = ST.save_lastloc;
cur_curlyx->u.curlyx.count--;
CACHEsayNO;
rex = (struct regexp *)SvANY(rex_sv);
rexi = RXi_GET(rex);
cur_curlyx = cur_eval->u.eval.prev_curlyx;
- ReREFCNT_inc(rex_sv);
+ (void)ReREFCNT_inc(rex_sv);
st->u.eval.cp = regcppush(0); /* Save *all* the positions. */
/* rex was changed so update the pointer in PL_reglastparen and PL_reglastcloseparen */
#define CASE_CLASS(nAmE) \
case nAmE: \
+ if (locinput >= PL_regeol) \
+ sayNO; \
if ((n=is_##nAmE(locinput,utf8_target))) { \
locinput += n; \
nextchr = UCHARAT(locinput); \
sayNO; \
break; \
case N##nAmE: \
+ if (locinput >= PL_regeol) \
+ sayNO; \
if ((n=is_##nAmE(locinput,utf8_target))) { \
sayNO; \
} else { \
case EXACTFL:
PL_reg_flags |= RF_tainted;
- /* FALL THROUGH */
+ utf8_flags = FOLDEQ_UTF8_LOCALE;
+ goto do_exactf;
+
case EXACTF:
case EXACTFU:
- utf8_flags = 0;
+ utf8_flags = (UTF_PATTERN) ? FOLDEQ_S2_ALREADY_FOLDED : 0;
/* The comments for the EXACT case above apply as well to these fold
* ones */
assert(! UTF_PATTERN || UNI_IS_INVARIANT(c));
if (utf8_target) { /* Use full Unicode fold matching */
-
- /* For the EXACTFL case, It doesn't really make sense to compare
- * locale and utf8, but it is best we can do. The documents warn
- * against mixing them */
-
char *tmpeol = loceol;
while (hardcount < max
&& foldEQ_utf8_flags(scan, &tmpeol, 0, utf8_target,
PERL_ARGS_ASSERT_REGCLASS_SWASH;
+ assert(ANYOF_NONBITMAP(node));
+
if (data && data->count) {
const U32 n = ARG(node);
}
/* If the bitmap didn't (or couldn't) match, and something outside the
- * bitmap could match, try that */
+ * bitmap could match, try that. Locale nodes specifiy completely the
+ * behavior of code points in the bit map (otherwise, a utf8 target would
+ * cause them to be treated as Unicode and not locale), except in
+ * the very unlikely event when this node is a synthetic start class, which
+ * could be a combination of locale and non-locale nodes. So allow locale
+ * to match for the synthetic start class, which will give a false
+ * positive that will be resolved when the match is done again as not part
+ * of the synthetic start class */
if (!match) {
if (utf8_target && (flags & ANYOF_UNICODE_ALL) && c >= 256) {
match = TRUE; /* Everything above 255 matches */
}
- else if ((flags & ANYOF_NONBITMAP_NON_UTF8
- || (utf8_target && flags & ANYOF_UTF8)))
+ else if (ANYOF_NONBITMAP(n)
+ && ((flags & ANYOF_NONBITMAP_NON_UTF8)
+ || (utf8_target
+ && (c >=256
+ || (! (flags & ANYOF_LOCALE))
+ || (flags & ANYOF_IS_SYNTHETIC)))))
{
AV *av;
SV * const sw = regclass_swash(prog, n, TRUE, 0, (SV**)&av);
else if (flags & ANYOF_LOC_NONBITMAP_FOLD) {
/* Here, we need to test if the fold of the target string
- * matches. In the case of a multi-char fold that is
- * caught by regcomp.c, it has stored all such folds into
- * 'av'; we linearly check to see if any match the target
- * string (folded). We know that the originals were each
- * one character, but we don't currently know how many
- * characters/bytes each folded to, except we do know that
- * there are small limits imposed by Unicode. XXX A
- * performance enhancement would be to have regcomp.c store
- * the max number of chars/bytes that are in an av entry,
- * as, say the 0th element. Even better would be to have a
- * hash of the few characters that can start a multi-char
- * fold to the max number of chars of those folds.
- *
- * Further down, if there isn't a
- * match in the av, we will check if there is another
- * fold-type match. For that, we also need the fold, but
- * only the first character. No sense in folding it twice,
- * so we do it here, even if there isn't any multi-char
- * fold, so we always fold at least the first character.
- * If the node is a straight ANYOF node, or there is only
- * one character available in the string, or if there isn't
- * any av, that's all we have to fold. In the case of a
- * multi-char fold, we do have guarantees in Unicode that
- * it can only expand up to so many characters and so many
- * bytes. We keep track so don't exceed either.
+ * matches. The non-multi char folds have all been moved to
+ * the compilation phase, and the multi-char folds have
+ * been stored by regcomp into 'av'; we linearly check to
+ * see if any match the target string (folded). We know
+ * that the originals were each one character, but we don't
+ * currently know how many characters/bytes each folded to,
+ * except we do know that there are small limits imposed by
+ * Unicode. XXX A performance enhancement would be to have
+ * regcomp.c store the max number of chars/bytes that are
+ * in an av entry, as, say the 0th element. Even better
+ * would be to have a hash of the few characters that can
+ * start a multi-char fold to the max number of chars of
+ * those folds.
*
* If there is a match, we will need to advance (if lenp is
* specified) the match pointer in the target string. But
* create a map so that we know how many bytes in the
* source to advance given that we have matched a certain
* number of bytes in the fold. This map is stored in
- * 'map_fold_len_back'. The first character in the fold
- * has array element 1 contain the number of bytes in the
- * source that folded to it; the 2nd is the cumulative
- * number to match it; ... */
- U8 map_fold_len_back[UTF8_MAX_FOLD_CHAR_EXPAND] = { 0 };
+ * 'map_fold_len_back'. Let n mean the number of bytes in
+ * the fold of the first character that we are folding.
+ * Then map_fold_len_back[n] is set to the number of bytes
+ * in that first character. Similarly let m be the
+ * corresponding number for the second character to be
+ * folded. Then map_fold_len_back[n+m] is set to the
+ * number of bytes occupied by the first two source
+ * characters. ... */
+ U8 map_fold_len_back[UTF8_MAXBYTES_CASE+1] = { 0 };
U8 folded[UTF8_MAXBYTES_CASE+1];
STRLEN foldlen = 0; /* num bytes in fold of 1st char */
- STRLEN foldlen_for_av; /* num bytes in fold of all chars */
+ STRLEN total_foldlen = 0; /* num bytes in fold of all
+ chars */
if (OP(n) == ANYOF || maxlen == 1 || ! lenp || ! av) {
/* Here, only need to fold the first char of the target
- * string */
+ * string. It the source wasn't utf8, is 1 byte long */
to_utf8_fold(utf8_p, folded, &foldlen);
- foldlen_for_av = foldlen;
- map_fold_len_back[1] = UTF8SKIP(utf8_p);
+ total_foldlen = foldlen;
+ map_fold_len_back[foldlen] = (utf8_target)
+ ? UTF8SKIP(utf8_p)
+ : 1;
}
else {
/* Here, need to fold more than the first char. Do so
* up to the limits */
- UV which_char = 0;
U8* source_ptr = utf8_p; /* The source for the fold
is the regex target
string */
U8* e = utf8_p + maxlen; /* Can't go beyond last
available byte in the
target string */
- while (which_char < UTF8_MAX_FOLD_CHAR_EXPAND
- && source_ptr < e)
+ U8 i;
+ for (i = 0;
+ i < UTF8_MAX_FOLD_CHAR_EXPAND && source_ptr < e;
+ i++)
{
/* Fold the next character */
break;
}
- /* Save the first character's folded length, in
- * case we have to use it later */
- if (! foldlen) {
- foldlen = this_char_foldlen;
- }
-
- /* Here, add the fold of this character */
+ /* Add the fold of this character */
Copy(this_char_folded,
folded_ptr,
this_char_foldlen,
U8);
- which_char++;
- map_fold_len_back[which_char] =
- map_fold_len_back[which_char - 1]
- + UTF8SKIP(source_ptr);
- folded_ptr += this_char_foldlen;
source_ptr += UTF8SKIP(source_ptr);
+ folded_ptr += this_char_foldlen;
+ total_foldlen = folded_ptr - folded;
+
+ /* Create map from the number of bytes in the fold
+ * back to the number of bytes in the source. If
+ * the source isn't utf8, the byte count is just
+ * the number of characters so far */
+ map_fold_len_back[total_foldlen]
+ = (utf8_target)
+ ? source_ptr - utf8_p
+ : i + 1;
}
*folded_ptr = '\0';
- foldlen_for_av = folded_ptr - folded;
}
/* Do the linear search to see if the fold is in the list
- * of multi-char folds. (Useless to look if won't be able
- * to store that it is a multi-char fold in *lenp) */
- if (lenp && av) {
+ * of multi-char folds. */
+ if (av) {
I32 i;
for (i = 0; i <= av_len(av); i++) {
SV* const sv = *av_fetch(av, i, FALSE);
STRLEN len;
const char * const s = SvPV_const(sv, len);
- if (len <= foldlen_for_av && memEQ(s,
- (char*)folded,
- len))
+
+ if (len <= total_foldlen
+ && memEQ(s, (char*)folded, len)
+
+ /* If 0, means matched a partial char. See
+ * [perl #90536] */
+ && map_fold_len_back[len])
{
/* Advance the target string ptr to account for
* this fold, but have to translate from the
* folded length to the corresponding source
- * length. The array is indexed by how many
- * characters in the match */
- *lenp = map_fold_len_back[
- utf8_length(folded, folded + len)];
+ * length. */
+ if (lenp) {
+ *lenp = map_fold_len_back[len];
+ }
match = TRUE;
break;
}
}
}
-#if 0
- if (!match) { /* See if the folded version matches */
- SV** listp;
-
- /* Consider "k" =~ /[K]/i. The line above would have
- * just folded the 'k' to itself, and that isn't going
- * to match 'K'. So we look through the closure of
- * everything that folds to 'k'. That will find the
- * 'K'. Initialize the list, if necessary */
- if (! PL_utf8_foldclosures) {
-
- /* If the folds haven't been read in, call a fold
- * function to force that */
- if (! PL_utf8_tofold) {
- U8 dummy[UTF8_MAXBYTES+1];
- STRLEN dummy_len;
- to_utf8_fold((U8*) "A", dummy, &dummy_len);
- }
- PL_utf8_foldclosures =
- _swash_inversion_hash(PL_utf8_tofold);
- }
-
- /* The data structure is a hash with the keys every
- * character that is folded to, like 'k', and the
- * values each an array of everything that folds to its
- * key. e.g. [ 'k', 'K', KELVIN_SIGN ] */
- if ((listp = hv_fetch(PL_utf8_foldclosures,
- (char *) folded, foldlen, FALSE)))
- {
- AV* list = (AV*) *listp;
- IV i;
- for (i = 0; i <= av_len(list); i++) {
- SV** try_p = av_fetch(list, i, FALSE);
- char* try_c;
- if (try_p == NULL) {
- Perl_croak(aTHX_ "panic: invalid PL_utf8_foldclosures structure");
- }
- /* Don't have to worry about embedded nulls
- * since NULL isn't folded or foldable */
- try_c = SvPVX(*try_p);
-
- /* The fold in a few cases of an above Latin1
- * char is in the Latin1 range, and hence may
- * be in the bitmap */
- if (UTF8_IS_INVARIANT(*try_c)
- && ANYOF_BITMAP_TEST(n,
- UNI_TO_NATIVE(*try_c)))
- {
- match = TRUE;
- break;
- }
- else if
- (UTF8_IS_DOWNGRADEABLE_START(*try_c)
- && ANYOF_BITMAP_TEST(n, UNI_TO_NATIVE(
- TWO_BYTE_UTF8_TO_UNI(try_c[0],
- try_c[1]))))
- {
- /* Since the fold comes from internally
- * generated data, we can safely assume it
- * is valid utf8 in the test above */
- match = TRUE;
- break;
- } else if (swash_fetch(sw, (U8*) try_c, TRUE)) {
- match = TRUE;
- break;
- }
- }
- }
- }
-#endif
}
/* If we allocated a string above, free it */
STATIC U8 *
S_reghop3(U8 *s, I32 off, const U8* lim)
{
+ /* return the position 'off' UTF-8 characters away from 's', forward if
+ * 'off' >= 0, backwards if negative. But don't go outside of position
+ * 'lim', which better be < s if off < 0 */
+
dVAR;
PERL_ARGS_ASSERT_REGHOP3;
prog->substrs->data[i].utf8_substr = sv;
sv_utf8_upgrade(sv);
if (SvVALID(prog->substrs->data[i].substr)) {
- const U8 flags = BmFLAGS(prog->substrs->data[i].substr);
- if (flags & FBMcf_TAIL) {
+ if (SvTAIL(prog->substrs->data[i].substr)) {
/* Trim the trailing \n that fbm_compile added last
time. */
SvCUR_set(sv, SvCUR(sv) - 1);
/* Whilst this makes the SV technically "invalid" (as its
buffer is no longer followed by "\0") when fbm_compile()
adds the "\n" back, a "\0" is restored. */
- }
- fbm_compile(sv, flags);
+ fbm_compile(sv, FBMcf_TAIL);
+ } else
+ fbm_compile(sv, 0);
}
if (prog->substrs->data[i].substr == prog->check_substr)
prog->check_utf8 = sv;
SV* sv = newSVsv(prog->substrs->data[i].utf8_substr);
if (sv_utf8_downgrade(sv, TRUE)) {
if (SvVALID(prog->substrs->data[i].utf8_substr)) {
- const U8 flags
- = BmFLAGS(prog->substrs->data[i].utf8_substr);
- if (flags & FBMcf_TAIL) {
+ if (SvTAIL(prog->substrs->data[i].utf8_substr)) {
/* Trim the trailing \n that fbm_compile added last
time. */
SvCUR_set(sv, SvCUR(sv) - 1);
- }
- fbm_compile(sv, flags);
- }
+ fbm_compile(sv, FBMcf_TAIL);
+ } else
+ fbm_compile(sv, 0);
+ }
} else {
SvREFCNT_dec(sv);
sv = &PL_sv_undef;
https://perl5.git.perl.org / perl5.git / blobdiff