/* Currently these are only used when PL_regkind[OP(rn)] == EXACT so
we don't need this definition. */
#define IS_TEXT(rn) ( OP(rn)==EXACT || OP(rn)==REF || OP(rn)==NREF )
-#define IS_TEXTF(rn) ( OP(rn)==EXACTF || OP(rn)==REFF || OP(rn)==NREFF )
+#define IS_TEXTF(rn) ( (OP(rn)==EXACTFU || OP(rn)==EXACTF) || OP(rn)==REFF || OP(rn)==NREFF )
#define IS_TEXTFL(rn) ( OP(rn)==EXACTFL || OP(rn)==REFFL || OP(rn)==NREFFL )
#else
/* ... so we use this as its faster. */
#define IS_TEXT(rn) ( OP(rn)==EXACT )
+#define IS_TEXTFU(rn) ( OP(rn)==EXACTFU )
#define IS_TEXTF(rn) ( OP(rn)==EXACTF )
#define IS_TEXTFL(rn) ( OP(rn)==EXACTFL )
a) Anchored substring;
b) Fixed substring;
c) Whether we are anchored (beginning-of-line or \G);
- d) First node (of those at offset 0) which may distingush positions;
+ d) First node (of those at offset 0) which may distinguish positions;
We use a)b)d) and multiline-part of c), and try to find a position in the
string which does not contradict any of them.
*/
prog->float_substr = prog->float_utf8 = NULL; /* clear */
check = NULL; /* abort */
s = strpos;
- /* XXXX If the check string was an implicit check MBOL, then we need to unset the relevent flag
+ /* XXXX If the check string was an implicit check MBOL, then we need to unset the relevant flag
see http://bugs.activestate.com/show_bug.cgi?id=87173 */
if (prog->intflags & PREGf_IMPLICIT)
prog->extflags &= ~RXf_ANCH_MBOL;
even for \b or \B. But (minlen? 1 : 0) below assumes that
regstclass does not come from lookahead... */
/* If regstclass takes bytelength more than 1: If charlength==1, OK.
- This leaves EXACTF only, which is dealt with in find_byclass(). */
+ This leaves EXACTF, EXACTFU only, which are dealt with in find_byclass(). */
const U8* const str = (U8*)STRING(progi->regstclass);
const int cl_l = (PL_regkind[OP(progi->regstclass)] == EXACT
? CHR_DIST(str+STR_LEN(progi->regstclass), str)
}
if (!(utf8_target ? prog->float_utf8 : prog->float_substr)) /* Could have been deleted */
goto fail;
- /* Check is floating subtring. */
+ /* Check is floating substring. */
retry_floating_check:
t = check_at - start_shift;
DEBUG_EXECUTE_r( what = "floating" );
#define REXEC_FBC_EXACTISH_SCAN(CoNd) \
STMT_START { \
+ re_fold_t folder; \
+ switch (OP(c)) { \
+ case EXACTFU: folder = foldEQ_latin1; break; \
+ case EXACTFL: folder = foldEQ_locale; break; \
+ case EXACTF: folder = foldEQ; break; \
+ default: \
+ Perl_croak(aTHX_ "panic: Unexpected op %u", OP(c)); \
+ } \
while (s <= e) { \
if ( (CoNd) \
- && (ln == 1 || (OP(c) == EXACTF \
- ? foldEQ(s, m, ln) \
- : foldEQ_locale(s, m, ln))) \
- && (!reginfo || regtry(reginfo, &s)) ) \
+ && (ln == 1 || folder(s, m, ln)) \
+ && (!reginfo || regtry(reginfo, &s)) ) \
goto got_it; \
s++; \
} \
/* We know what class it must start with. */
switch (OP(c)) {
+ case ANYOFV:
case ANYOF:
- if (utf8_target) {
- REXEC_FBC_UTF8_CLASS_SCAN((ANYOF_FLAGS(c) & ANYOF_UNICODE) ||
+ if (utf8_target || OP(c) == ANYOFV) {
+ REXEC_FBC_UTF8_CLASS_SCAN((ANYOF_FLAGS(c) & ANYOF_NONBITMAP) ||
!UTF8_IS_INVARIANT((U8)s[0]) ?
reginclass(prog, c, (U8*)s, 0, utf8_target) :
REGINCLASS(prog, c, (U8*)s));
tmp = doevery;
);
break;
+ case EXACTFU:
case EXACTF:
m = STRING(c);
ln = STR_LEN(c); /* length to match in octets/bytes */
}
else {
c1 = *(U8*)m;
- c2 = PL_fold[c1];
+ if (utf8_target || OP(c) == EXACTFU) {
+
+ /* Micro sign folds to GREEK SMALL LETTER MU;
+ LATIN_SMALL_LETTER_SHARP_S folds to 'ss', and this sets
+ c2 to the first 's' of the pair, and the code below will
+ look for others */
+ c2 = (c1 == MICRO_SIGN)
+ ? GREEK_SMALL_LETTER_MU
+ : (c1 == LATIN_SMALL_LETTER_SHARP_S)
+ ? 's'
+ : PL_fold_latin1[c1];
+ } else c2 = PL_fold[c1];
}
goto do_exactf;
case EXACTFL:
);
case SPACEL:
REXEC_FBC_CSCAN_TAINT(
- *s == ' ' || isSPACE_LC_utf8((U8*)s),
+ isSPACE_LC_utf8((U8*)s),
isSPACE_LC(*s)
);
case NSPACE:
);
case NSPACEL:
REXEC_FBC_CSCAN_TAINT(
- !(*s == ' ' || isSPACE_LC_utf8((U8*)s)),
+ !isSPACE_LC_utf8((U8*)s),
!isSPACE_LC(*s)
);
case DIGIT:
PerlIO_printf( Perl_debug_log,
" Scanning for legal start char...\n");
}
- );
- while ( uc <= (U8*)last_start && !BITMAP_TEST(bitmap,*uc) ) {
- uc++;
- }
+ );
+ if (utf8_target) {
+ while ( uc <= (U8*)last_start && !BITMAP_TEST(bitmap,*uc) ) {
+ uc += UTF8SKIP(uc);
+ }
+ } else {
+ while ( uc <= (U8*)last_start && !BITMAP_TEST(bitmap,*uc) ) {
+ uc++;
+ }
+ }
s= (char *)uc;
}
if (uc >(U8*)last_start) break;
goto phooey;
} else if (RXf_GPOS_CHECK == (prog->extflags & RXf_GPOS_CHECK))
{
- /* the warning about reginfo.ganch being used without intialization
+ /* the warning about reginfo.ganch being used without initialization
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;
(?>A), which pop all the backtrack states associated with A before
continuing.
-Bascktrack state structs are allocated in slabs of about 4K in size.
+Backtrack state structs are allocated in slabs of about 4K in size.
PL_regmatch_state and st always point to the currently active state,
and PL_regmatch_slab points to the slab currently containing
PL_regmatch_state. The first time regmatch() is called, the first slab is
SV *popmark = NULL; /* are we looking for a mark? */
SV *sv_commit = NULL; /* last mark name seen in failure */
SV *sv_yes_mark = NULL; /* last mark name we have seen
- during a successfull match */
+ during a successful match */
U32 lastopen = 0; /* last open we saw */
bool has_cutgroup = RX_HAS_CUTGROUP(rex) ? 1 : 0;
SV* const oreplsv = GvSV(PL_replgv);
"%*s %smatched empty string...%s\n",
REPORT_CODE_OFF+depth*2, "", PL_colors[4], PL_colors[5])
);
- break;
+ if (!trie->jump)
+ break;
} else {
DEBUG_EXECUTE_r(
PerlIO_printf(Perl_debug_log,
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)
+ (SvUTF8(*tmp) ? PERL_PV_ESCAPE_UNI : 0)|PERL_PV_ESCAPE_NONASCII
)
: "not compiled under -Dr",
PL_colors[5] );
nextchr = UCHARAT(locinput);
break;
}
- case EXACTFL:
+ case EXACTFL: {
+ re_fold_t folder;
+ const U8 * fold_array;
+ const char * s;
+
PL_reg_flags |= RF_tainted;
- /* FALL THROUGH */
- case EXACTF: {
- char * const s = STRING(scan);
+ folder = foldEQ_locale;
+ fold_array = PL_fold_locale;
+ goto do_exactf;
+
+ case EXACTFU:
+ folder = foldEQ_latin1;
+ fold_array = PL_fold_latin1;
+ goto do_exactf;
+
+ case EXACTF:
+ folder = foldEQ;
+ fold_array = PL_fold;
+
+ do_exactf:
+ s = STRING(scan);
ln = STR_LEN(scan);
if (utf8_target || UTF_PATTERN) {
/* Inline the first character, for speed. */
if (UCHARAT(s) != nextchr &&
- UCHARAT(s) != ((OP(scan) == EXACTF)
- ? PL_fold : PL_fold_locale)[nextchr])
+ UCHARAT(s) != fold_array[nextchr])
+ {
sayNO;
+ }
if (PL_regeol - locinput < ln)
sayNO;
- if (ln > 1 && (OP(scan) == EXACTF
- ? ! foldEQ(s, locinput, ln)
- : ! foldEQ_locale(s, locinput, ln)))
+ if (ln > 1 && ! folder(s, locinput, ln))
sayNO;
locinput += ln;
nextchr = UCHARAT(locinput);
break;
- }
+ }
case BOUNDL:
case NBOUNDL:
PL_reg_flags |= RF_tainted;
OP(scan) == BOUNDL))
sayNO;
break;
+ case ANYOFV:
case ANYOF:
- if (utf8_target) {
+ if (utf8_target || state_num == ANYOFV) {
STRLEN inclasslen = PL_regeol - locinput;
if (locinput >= PL_regeol)
sayNO;
if (!reginclass(rex, scan, (U8*)locinput, &inclasslen, utf8_target))
- goto anyof_fail;
+ sayNO;
locinput += inclasslen;
nextchr = UCHARAT(locinput);
break;
if (!nextchr && locinput >= PL_regeol)
sayNO;
if (!REGINCLASS(rex, scan, (U8*)locinput))
- goto anyof_fail;
+ 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;
/* Special char classes - The defines start on line 129 or so */
CCC_TRY_AFF_U( ALNUM, ALNUML, perl_word, "a", isALNUM_LC_utf8, isWORDCHAR_L1, isALNUM_LC);
break;
case NREFFL:
- {
+ { /* The capture buffer cases. The ones beginning with N for the
+ named buffers just convert to the equivalent numbered and
+ pretend they were called as the corresponding numbered buffer
+ op. */
+ /* don't initialize these, it makes C++ unhappy */
char *s;
char type;
+ re_fold_t folder;
+ const U8 *fold_array;
+
PL_reg_flags |= RF_tainted;
- /* FALL THROUGH */
- case NREF:
+ folder = foldEQ_locale;
+ fold_array = PL_fold_locale;
+ type = REFFL;
+ goto do_nref;
+
+ case NREFFU:
+ folder = foldEQ_latin1;
+ fold_array = PL_fold_latin1;
+ type = REFFU;
+ goto do_nref;
+
case NREFF:
- type = OP(scan);
+ folder = foldEQ;
+ fold_array = PL_fold;
+ type = REFF;
+ goto do_nref;
+
+ case NREF:
+ type = REF;
+ folder = NULL;
+ fold_array = NULL;
+ do_nref:
+
+ /* For the named back references, find the corresponding buffer
+ * number */
n = reg_check_named_buff_matched(rex,scan);
- if ( n ) {
- type = REF + ( type - NREF );
- goto do_ref;
- } else {
+ if ( ! n ) {
sayNO;
- }
- /* unreached */
+ }
+ goto do_nref_ref_common;
+
case REFFL:
PL_reg_flags |= RF_tainted;
- /* FALL THROUGH */
+ folder = foldEQ_locale;
+ fold_array = PL_fold_locale;
+ goto do_ref;
+
+ case REFFU:
+ folder = foldEQ_latin1;
+ fold_array = PL_fold_latin1;
+ goto do_ref;
+
+ case REFF:
+ folder = foldEQ;
+ fold_array = PL_fold;
+ goto do_ref;
+
case REF:
- case REFF:
- n = ARG(scan); /* which paren pair */
+ folder = NULL;
+ fold_array = NULL;
+
+ do_ref:
type = OP(scan);
- do_ref:
+ n = ARG(scan); /* which paren pair */
+
+ do_nref_ref_common:
ln = PL_regoffs[n].start;
PL_reg_leftiter = PL_reg_maxiter; /* Void cache */
if (*PL_reglastparen < n || ln == -1)
break;
s = PL_bostr + ln;
- if (utf8_target && type != REF) { /* REF can do byte comparison */
- char *l = locinput;
- 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
- * have to map both upper and title case to lower case.
- */
- if (type == REFF) {
- while (s < e) {
- STRLEN ulen1, ulen2;
- U8 tmpbuf1[UTF8_MAXBYTES_CASE+1];
- U8 tmpbuf2[UTF8_MAXBYTES_CASE+1];
-
- if (l >= PL_regeol)
- sayNO;
- toLOWER_utf8((U8*)s, tmpbuf1, &ulen1);
- toLOWER_utf8((U8*)l, tmpbuf2, &ulen2);
- if (ulen1 != ulen2 || memNE((char *)tmpbuf1, (char *)tmpbuf2, ulen1))
- sayNO;
- s += ulen1;
- l += ulen2;
- }
+ if (type != REF /* REF can do byte comparison */
+ && (utf8_target
+ || (type == REFFU
+ && (*s == (char) LATIN_SMALL_LETTER_SHARP_S
+ || *locinput == (char) LATIN_SMALL_LETTER_SHARP_S))))
+ { /* XXX handle REFFL better */
+ char * limit = PL_regeol;
+
+ /* This call case insensitively compares the entire buffer
+ * at s, with the current input starting at locinput, but
+ * not going off the end given by PL_regeol, and returns in
+ * limit upon success, how much of the current input was
+ * matched */
+ if (! foldEQ_utf8(s, NULL, PL_regoffs[n].end - ln, utf8_target,
+ locinput, &limit, 0, utf8_target))
+ {
+ sayNO;
}
- locinput = l;
+ locinput = limit;
nextchr = UCHARAT(locinput);
break;
}
- /* Inline the first character, for speed. */
+ /* Not utf8: Inline the first character, for speed. */
if (UCHARAT(s) != nextchr &&
(type == REF ||
- (UCHARAT(s) != (type == REFF
- ? PL_fold : PL_fold_locale)[nextchr])))
+ UCHARAT(s) != fold_array[nextchr]))
sayNO;
ln = PL_regoffs[n].end - ln;
if (locinput + ln > PL_regeol)
sayNO;
if (ln > 1 && (type == REF
? memNE(s, locinput, ln)
- : (type == REFF
- ? ! foldEQ(s, locinput, ln)
- : ! foldEQ_locale(s, locinput, ln))))
+ : ! folder(s, locinput, ln)))
sayNO;
locinput += ln;
nextchr = UCHARAT(locinput);
{
ST.c1 = (U8)*STRING(text_node);
- ST.c2 =
- (IS_TEXTF(text_node))
- ? PL_fold[ST.c1]
- : (IS_TEXTFL(text_node))
- ? PL_fold_locale[ST.c1]
- : ST.c1;
+ switch (OP(text_node)) {
+ case EXACTF: ST.c2 = PL_fold[ST.c1]; break;
+ case EXACTFU: ST.c2 = PL_fold_latin1[ST.c1]; break;
+ case EXACTFL: ST.c2 = PL_fold_locale[ST.c1]; break;
+ default: ST.c2 = ST.c1;
+ }
}
}
}
if this changes back then the macro for IS_TEXT and
friends need to change. */
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];
+ ST.c1 = *s;
+ switch (OP(text_node)) {
+ case EXACTF: ST.c2 = PL_fold[ST.c1]; break;
+ case EXACTFU: ST.c2 = PL_fold_latin1[ST.c1]; break;
+ case EXACTFL: ST.c2 = PL_fold_locale[ST.c1]; break;
+ default: ST.c2 = ST.c1; break;
+ }
}
else { /* UTF_PATTERN */
- if (IS_TEXTF(text_node)) {
+ if (IS_TEXTFU(text_node) || IS_TEXTF(text_node)) {
STRLEN ulen1, ulen2;
U8 tmpbuf1[UTF8_MAXBYTES_CASE+1];
U8 tmpbuf2[UTF8_MAXBYTES_CASE+1];
n = ARG(scan);
if ( n == (U32)what_len_TRICKYFOLD(locinput,utf8_target,ln) ) {
locinput += ln;
- } else if ( 0xDF == n && !utf8_target && !UTF_PATTERN ) {
+ } else if ( LATIN_SMALL_LETTER_SHARP_S == n && !utf8_target && !UTF_PATTERN ) {
sayNO;
} else {
U8 folded[UTF8_MAXBYTES_CASE+1];
case CANY:
scan = loceol;
break;
+ case EXACT:
+ /* To get here, EXACTish nodes must have *byte* length == 1. That
+ * means they match only characters in the string that can be expressed
+ * as a single byte. For non-utf8 strings, that means a simple match.
+ * For utf8 strings, the character matched must be an invariant, or
+ * downgradable to a single byte. The pattern's utf8ness is
+ * irrelevant, as since it's a single byte, it either isn't utf8, or if
+ * it is, it's an invariant */
+
+ c = (U8)*STRING(p);
+ assert(! UTF_PATTERN || UNI_IS_INVARIANT(c));
+
+ if (! utf8_target || UNI_IS_INVARIANT(c)) {
+ while (scan < loceol && UCHARAT(scan) == c) {
+ scan++;
+ }
+ }
+ else {
+
+ /* Here, the string is utf8, and the pattern char is different
+ * in utf8 than not, so can't compare them directly. Outside the
+ * loop, find find the two utf8 bytes that represent c, and then
+ * look for those in sequence in the utf8 string */
+ U8 high = UTF8_TWO_BYTE_HI(c);
+ U8 low = UTF8_TWO_BYTE_LO(c);
+ loceol = PL_regeol;
+
+ while (hardcount < max
+ && scan + 1 < loceol
+ && UCHARAT(scan) == high
+ && UCHARAT(scan + 1) == low)
+ {
+ scan += 2;
+ hardcount++;
+ }
+ }
+ break;
case EXACTFL:
PL_reg_flags |= RF_tainted;
/* FALL THROUGH */
- case EXACT:
case EXACTF:
- /* To get here, EXACTish nodes must have *byte* length == 1. That means
- * they match only characters in the string that can be expressed as a
- * single byte. For non-utf8 strings, that means a simple match. For
- * utf8 strings, the character matched must be an invariant, or
- * downgradable to a single byte. The pattern's utf8ness is
- * irrelevant, as it must be a single byte, so either it isn't utf8, or
- * if it is it's an invariant */
+ case EXACTFU:
+
+ /* The comments for the EXACT case above apply as well to these fold
+ * ones */
c = (U8)*STRING(p);
assert(! UTF_PATTERN || UNI_IS_INVARIANT(c));
- if ((! utf8_target) || UNI_IS_INVARIANT(c)) {
+ if (utf8_target) { /* Use full Unicode fold matching */
- /* Here, the string isn't utf8, or the character in the EXACT
- * node is the same in utf8 as not, so can just do equality.
- * Each matching char must be 1 byte long */
- switch (OP(p)) {
- case EXACT:
- while (scan < loceol && UCHARAT(scan) == c) {
- scan++;
- }
- break;
- case EXACTF:
- while (scan < loceol &&
- (UCHARAT(scan) == c || UCHARAT(scan) == PL_fold[c]))
- {
- scan++;
- }
- break;
- case EXACTFL:
- while (scan < loceol &&
- (UCHARAT(scan) == c || UCHARAT(scan) == PL_fold_locale[c]))
- {
- scan++;
- }
- break;
- default:
- Perl_croak(aTHX_ "panic: Unexpected op %u", OP(p));
+ /* 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(scan, &tmpeol, 0, utf8_target,
+ STRING(p), NULL, 1, cBOOL(UTF_PATTERN)))
+ {
+ scan = tmpeol;
+ tmpeol = loceol;
+ hardcount++;
}
+
+ /* XXX Note that the above handles properly the German sharp s in
+ * the pattern matching ss in the string. But it doesn't handle
+ * properly cases where the string contains say 'LIGATURE ff' and
+ * the pattern is 'f+'. This would require, say, a new function or
+ * revised interface to foldEQ_utf8(), in which the maximum number
+ * of characters to match could be passed and it would return how
+ * many actually did. This is just one of many cases where
+ * multi-char folds don't work properly, and so the fix is being
+ * deferred */
}
else {
+ U8 folded;
- /* Here, the string is utf8, and the pattern char is different
- * in utf8 than not. */
-
+ /* Here, the string isn't utf8 and c is a single byte; and either
+ * the pattern isn't utf8 or c is an invariant, so its utf8ness
+ * doesn't affect c. Can just do simple comparisons for exact or
+ * fold matching. */
switch (OP(p)) {
- case EXACT:
- {
- /* Fastest to find the two utf8 bytes that represent c, and
- * then look for those in sequence in the utf8 string */
- U8 high = UTF8_TWO_BYTE_HI(c);
- U8 low = UTF8_TWO_BYTE_LO(c);
- loceol = PL_regeol;
-
- while (hardcount < max
- && scan + 1 < loceol
- && UCHARAT(scan) == high
- && UCHARAT(scan + 1) == low)
- {
- scan += 2;
- hardcount++;
- }
- }
- break;
- case EXACTFL: /* Doesn't really make sense, but is best we can
- do. The documents warn against mixing locale
- and utf8 */
- case EXACTF:
- { /* utf8 string, so use utf8 foldEQ */
- char *tmpeol = loceol;
- while (hardcount < max
- && foldEQ_utf8(scan, &tmpeol, 0, utf8_target,
- STRING(p), NULL, 1, UTF_PATTERN))
- {
- scan = tmpeol;
- tmpeol = loceol;
- hardcount++;
- }
-
- /* XXX Note that the above handles properly the German
- * sharp ss in the pattern matching ss in the string. But
- * it doesn't handle properly cases where the string
- * contains say 'LIGATURE ff' and the pattern is 'f+'.
- * This would require, say, a new function or revised
- * interface to foldEQ_utf8(), in which the maximum number
- * of characters to match could be passed and it would
- * return how many actually did. This is just one of many
- * cases where multi-char folds don't work properly, and so
- * the fix is being deferred */
- }
- break;
- default:
- Perl_croak(aTHX_ "panic: Unexpected op %u", OP(p));
+ case EXACTF: folded = PL_fold[c]; break;
+ case EXACTFU: folded = PL_fold_latin1[c]; break;
+ case EXACTFL: folded = PL_fold_locale[c]; break;
+ default: Perl_croak(aTHX_ "panic: Unexpected op %u", OP(p));
+ }
+ while (scan < loceol &&
+ (UCHARAT(scan) == c || UCHARAT(scan) == folded))
+ {
+ scan++;
}
}
break;
+ case ANYOFV:
case ANYOF:
if (utf8_target) {
loceol = PL_regeol;
if (utf8_target) {
loceol = PL_regeol;
while (hardcount < max && scan < loceol &&
- (*scan == ' ' || isSPACE_LC_utf8((U8*)scan))) {
+ isSPACE_LC_utf8((U8*)scan)) {
scan += UTF8SKIP(scan);
hardcount++;
}
if (utf8_target) {
loceol = PL_regeol;
while (hardcount < max && scan < loceol &&
- !(*scan == ' ' || isSPACE_LC_utf8((U8*)scan))) {
+ !isSPACE_LC_utf8((U8*)scan)) {
scan += UTF8SKIP(scan);
hardcount++;
}
scan++;
}
break;
+ case DIGITL:
+ PL_reg_flags |= RF_tainted;
+ if (utf8_target) {
+ loceol = PL_regeol;
+ while (hardcount < max && scan < loceol &&
+ isDIGIT_LC_utf8((U8*)scan)) {
+ scan += UTF8SKIP(scan);
+ hardcount++;
+ }
+ } else {
+ while (scan < loceol && isDIGIT_LC(*scan))
+ scan++;
+ }
+ break;
case NDIGIT:
if (utf8_target) {
loceol = PL_regeol;
while (scan < loceol && !isDIGIT(*scan))
scan++;
}
+ case NDIGITL:
+ PL_reg_flags |= RF_tainted;
+ if (utf8_target) {
+ loceol = PL_regeol;
+ while (hardcount < max && scan < loceol &&
+ !isDIGIT_LC_utf8((U8*)scan)) {
+ scan += UTF8SKIP(scan);
+ hardcount++;
+ }
+ } else {
+ while (scan < loceol && !isDIGIT_LC(*scan))
+ scan++;
+ }
+ break;
case LNBREAK:
if (utf8_target) {
loceol = PL_regeol;
bytes in p were matched. If there was no match, the value is undefined,
possibly changed from the input.
+ Note that this can be a synthetic start class, a combination of various
+ nodes, so things you think might be mutually exclusive, such as locale,
+ aren't. It can match both locale and non-locale
+
*/
STATIC bool
maxlen = c_len;
}
- if (utf8_target || (flags & ANYOF_UNICODE)) {
- if (utf8_target && !ANYOF_RUNTIME(n)) {
- if (c_len != (STRLEN)-1 && c < 256 && ANYOF_BITMAP_TEST(n, c))
+ /* If this character is potentially in the bitmap, check it */
+ if (c < 256) {
+ if (ANYOF_BITMAP_TEST(n, c))
+ match = TRUE;
+ else if (flags & ANYOF_NON_UTF8_LATIN1_ALL
+ && ! utf8_target
+ && ! isASCII(c))
+ {
+ match = TRUE;
+ }
+
+ else if (flags & ANYOF_LOCALE) {
+ PL_reg_flags |= RF_tainted;
+
+ if ((flags & ANYOF_LOC_NONBITMAP_FOLD)
+ && ANYOF_BITMAP_TEST(n, PL_fold_locale[c]))
+ {
match = TRUE;
+ }
+ else if (ANYOF_CLASS_TEST_ANY_SET(n) &&
+ ((ANYOF_CLASS_TEST(n, ANYOF_ALNUM) && isALNUM_LC(c)) ||
+ (ANYOF_CLASS_TEST(n, ANYOF_NALNUM) && !isALNUM_LC(c)) ||
+ (ANYOF_CLASS_TEST(n, ANYOF_SPACE) && isSPACE_LC(c)) ||
+ (ANYOF_CLASS_TEST(n, ANYOF_NSPACE) && !isSPACE_LC(c)) ||
+ (ANYOF_CLASS_TEST(n, ANYOF_DIGIT) && isDIGIT_LC(c)) ||
+ (ANYOF_CLASS_TEST(n, ANYOF_NDIGIT) && !isDIGIT_LC(c)) ||
+ (ANYOF_CLASS_TEST(n, ANYOF_ALNUMC) && isALNUMC_LC(c)) ||
+ (ANYOF_CLASS_TEST(n, ANYOF_NALNUMC) && !isALNUMC_LC(c)) ||
+ (ANYOF_CLASS_TEST(n, ANYOF_ALPHA) && isALPHA_LC(c)) ||
+ (ANYOF_CLASS_TEST(n, ANYOF_NALPHA) && !isALPHA_LC(c)) ||
+ (ANYOF_CLASS_TEST(n, ANYOF_ASCII) && isASCII(c)) ||
+ (ANYOF_CLASS_TEST(n, ANYOF_NASCII) && !isASCII(c)) ||
+ (ANYOF_CLASS_TEST(n, ANYOF_CNTRL) && isCNTRL_LC(c)) ||
+ (ANYOF_CLASS_TEST(n, ANYOF_NCNTRL) && !isCNTRL_LC(c)) ||
+ (ANYOF_CLASS_TEST(n, ANYOF_GRAPH) && isGRAPH_LC(c)) ||
+ (ANYOF_CLASS_TEST(n, ANYOF_NGRAPH) && !isGRAPH_LC(c)) ||
+ (ANYOF_CLASS_TEST(n, ANYOF_LOWER) && isLOWER_LC(c)) ||
+ (ANYOF_CLASS_TEST(n, ANYOF_NLOWER) && !isLOWER_LC(c)) ||
+ (ANYOF_CLASS_TEST(n, ANYOF_PRINT) && isPRINT_LC(c)) ||
+ (ANYOF_CLASS_TEST(n, ANYOF_NPRINT) && !isPRINT_LC(c)) ||
+ (ANYOF_CLASS_TEST(n, ANYOF_PUNCT) && isPUNCT_LC(c)) ||
+ (ANYOF_CLASS_TEST(n, ANYOF_NPUNCT) && !isPUNCT_LC(c)) ||
+ (ANYOF_CLASS_TEST(n, ANYOF_UPPER) && isUPPER_LC(c)) ||
+ (ANYOF_CLASS_TEST(n, ANYOF_NUPPER) && !isUPPER_LC(c)) ||
+ (ANYOF_CLASS_TEST(n, ANYOF_XDIGIT) && isXDIGIT(c)) ||
+ (ANYOF_CLASS_TEST(n, ANYOF_NXDIGIT) && !isXDIGIT(c)) ||
+ (ANYOF_CLASS_TEST(n, ANYOF_PSXSPC) && isPSXSPC(c)) ||
+ (ANYOF_CLASS_TEST(n, ANYOF_NPSXSPC) && !isPSXSPC(c)) ||
+ (ANYOF_CLASS_TEST(n, ANYOF_BLANK) && isBLANK(c)) ||
+ (ANYOF_CLASS_TEST(n, ANYOF_NBLANK) && !isBLANK(c))
+ ) /* How's that for a conditional? */
+ ) {
+ match = TRUE;
+ }
}
- if (!match && utf8_target && (flags & ANYOF_UNICODE_ALL) && c >= 256)
- match = TRUE;
- if (!match) {
+ }
+
+ /* If the bitmap didn't (or couldn't) match, and something outside the
+ * bitmap could match, try that */
+ if (!match) {
+ if (utf8_target && (flags & ANYOF_UNICODE_ALL)) {
+ if (c >= 256
+ || ((flags & ANYOF_LOC_NONBITMAP_FOLD) /* Latin1 1 that has a
+ non-Latin1 fold
+ should match */
+ && _HAS_NONLATIN1_FOLD_CLOSURE_ONLY_FOR_USE_BY_REGCOMP_DOT_C_AND_REGEXEC_DOT_C(c)))
+ {
+ match = TRUE;
+ }
+ }
+ if (!match && ((flags & ANYOF_NONBITMAP_NON_UTF8)
+ || (utf8_target && flags & ANYOF_UTF8)))
+ {
AV *av;
SV * const sw = regclass_swash(prog, n, TRUE, 0, (SV**)&av);
-
+
if (sw) {
U8 * utf8_p;
if (utf8_target) {
utf8_p = (U8 *) p;
} else {
- STRLEN len = 1;
+
+ /* Not utf8. Convert as much of the string as available up
+ * to the limit of how far the (single) character in the
+ * pattern can possibly match (no need to go further). If
+ * the node is a straight ANYOF or not folding, it can't
+ * match more than one. Otherwise, It can match up to how
+ * far a single char can fold to. Since not utf8, each
+ * character is a single byte, so the max it can be in
+ * bytes is the same as the max it can be in characters */
+ STRLEN len = (OP(n) == ANYOF
+ || ! (flags & ANYOF_LOC_NONBITMAP_FOLD))
+ ? 1
+ : (maxlen < UTF8_MAX_FOLD_CHAR_EXPAND)
+ ? maxlen
+ : UTF8_MAX_FOLD_CHAR_EXPAND;
utf8_p = bytes_to_utf8(p, &len);
}
- if (swash_fetch(sw, utf8_p, 1))
+
+ if (swash_fetch(sw, utf8_p, 1)) /* See if in the swash */
match = TRUE;
- else if (flags & ANYOF_FOLD) {
- if (!match && lenp && 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.
+ *
+ * If there is a match, we will need to advance (if lenp is
+ * specified) the match pointer in the target string. But
+ * what we are comparing here isn't that string directly,
+ * but its fold, whose length may differ from the original.
+ * As we go along in constructing the fold, therefore, we
+ * 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 };
+ 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 */
+
+ if (OP(n) == ANYOF || maxlen == 1 || ! lenp || ! av) {
+
+ /* Here, only need to fold the first char of the target
+ * string */
+ to_utf8_fold(utf8_p, folded, &foldlen);
+ foldlen_for_av = foldlen;
+ map_fold_len_back[1] = UTF8SKIP(utf8_p);
+ }
+ 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* folded_ptr = folded;
+ 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)
+ {
+
+ /* Fold the next character */
+ U8 this_char_folded[UTF8_MAXBYTES_CASE+1];
+ STRLEN this_char_foldlen;
+ to_utf8_fold(source_ptr,
+ this_char_folded,
+ &this_char_foldlen);
+
+ /* Bail if it would exceed the byte limit for
+ * folding a single char. */
+ if (this_char_foldlen + folded_ptr - folded >
+ UTF8_MAXBYTES_CASE)
+ {
+ 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 */
+ 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 = '\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) {
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 <= maxlen && memEQ(s, (char*)utf8_p, len)) {
- *lenp = len;
+ if (len <= foldlen_for_av && memEQ(s,
+ (char*)folded,
+ 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)];
match = TRUE;
break;
}
}
}
- if (!match) {
- U8 tmpbuf[UTF8_MAXBYTES_CASE+1];
+ 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);
+ }
- STRLEN tmplen;
- to_utf8_fold(utf8_p, tmpbuf, &tmplen);
- if (swash_fetch(sw, tmpbuf, 1))
- match = TRUE;
+ /* 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, 1)) {
+ match = TRUE;
+ break;
+ }
+ }
+ }
}
}
}
}
}
- if (!match && c < 256) {
- if (ANYOF_BITMAP_TEST(n, c))
- match = TRUE;
- else if (flags & ANYOF_FOLD) {
- U8 f;
-
- if (flags & ANYOF_LOCALE) {
- PL_reg_flags |= RF_tainted;
- f = PL_fold_locale[c];
- }
- else
- f = PL_fold[c];
- if (f != c && ANYOF_BITMAP_TEST(n, f))
- match = TRUE;
- }
-
- if (!match && (flags & ANYOF_CLASS)) {
- PL_reg_flags |= RF_tainted;
- if (
- (ANYOF_CLASS_TEST(n, ANYOF_ALNUM) && isALNUM_LC(c)) ||
- (ANYOF_CLASS_TEST(n, ANYOF_NALNUM) && !isALNUM_LC(c)) ||
- (ANYOF_CLASS_TEST(n, ANYOF_SPACE) && isSPACE_LC(c)) ||
- (ANYOF_CLASS_TEST(n, ANYOF_NSPACE) && !isSPACE_LC(c)) ||
- (ANYOF_CLASS_TEST(n, ANYOF_DIGIT) && isDIGIT_LC(c)) ||
- (ANYOF_CLASS_TEST(n, ANYOF_NDIGIT) && !isDIGIT_LC(c)) ||
- (ANYOF_CLASS_TEST(n, ANYOF_ALNUMC) && isALNUMC_LC(c)) ||
- (ANYOF_CLASS_TEST(n, ANYOF_NALNUMC) && !isALNUMC_LC(c)) ||
- (ANYOF_CLASS_TEST(n, ANYOF_ALPHA) && isALPHA_LC(c)) ||
- (ANYOF_CLASS_TEST(n, ANYOF_NALPHA) && !isALPHA_LC(c)) ||
- (ANYOF_CLASS_TEST(n, ANYOF_ASCII) && isASCII(c)) ||
- (ANYOF_CLASS_TEST(n, ANYOF_NASCII) && !isASCII(c)) ||
- (ANYOF_CLASS_TEST(n, ANYOF_CNTRL) && isCNTRL_LC(c)) ||
- (ANYOF_CLASS_TEST(n, ANYOF_NCNTRL) && !isCNTRL_LC(c)) ||
- (ANYOF_CLASS_TEST(n, ANYOF_GRAPH) && isGRAPH_LC(c)) ||
- (ANYOF_CLASS_TEST(n, ANYOF_NGRAPH) && !isGRAPH_LC(c)) ||
- (ANYOF_CLASS_TEST(n, ANYOF_LOWER) && isLOWER_LC(c)) ||
- (ANYOF_CLASS_TEST(n, ANYOF_NLOWER) && !isLOWER_LC(c)) ||
- (ANYOF_CLASS_TEST(n, ANYOF_PRINT) && isPRINT_LC(c)) ||
- (ANYOF_CLASS_TEST(n, ANYOF_NPRINT) && !isPRINT_LC(c)) ||
- (ANYOF_CLASS_TEST(n, ANYOF_PUNCT) && isPUNCT_LC(c)) ||
- (ANYOF_CLASS_TEST(n, ANYOF_NPUNCT) && !isPUNCT_LC(c)) ||
- (ANYOF_CLASS_TEST(n, ANYOF_UPPER) && isUPPER_LC(c)) ||
- (ANYOF_CLASS_TEST(n, ANYOF_NUPPER) && !isUPPER_LC(c)) ||
- (ANYOF_CLASS_TEST(n, ANYOF_XDIGIT) && isXDIGIT(c)) ||
- (ANYOF_CLASS_TEST(n, ANYOF_NXDIGIT) && !isXDIGIT(c)) ||
- (ANYOF_CLASS_TEST(n, ANYOF_PSXSPC) && isPSXSPC(c)) ||
- (ANYOF_CLASS_TEST(n, ANYOF_NPSXSPC) && !isPSXSPC(c)) ||
- (ANYOF_CLASS_TEST(n, ANYOF_BLANK) && isBLANK(c)) ||
- (ANYOF_CLASS_TEST(n, ANYOF_NBLANK) && !isBLANK(c))
- ) /* How's that for a conditional? */
- {
- match = TRUE;
- }
- }
- }
return (flags & ANYOF_INVERT) ? !match : match;
}