static const char malformed_text[] = "Malformed UTF-8 character";
static const char unees[] =
"Malformed UTF-8 character (unexpected end of string)";
+
+/* Be sure to synchronize this message with the similar one in regcomp.c */
static const char cp_above_legal_max[] =
"Use of code point 0x%" UVXf " is not allowed; the"
" permissible max is 0x%" UVXf;
-#define MAX_EXTERNALLY_LEGAL_CP ((UV) (IV_MAX))
-
/*
=head1 Unicode Support
These are various utility functions for manipulating UTF8-encoded
}
}
+STATIC HV *
+S_new_msg_hv(pTHX_ const char * const message, /* The message text */
+ U32 categories, /* Packed warning categories */
+ U32 flag) /* Flag associated with this message */
+{
+ /* Creates, populates, and returns an HV* that describes an error message
+ * for the translators between UTF8 and code point */
+
+ SV* msg_sv = newSVpv(message, 0);
+ SV* category_sv = newSVuv(categories);
+ SV* flag_bit_sv = newSVuv(flag);
+
+ HV* msg_hv = newHV();
+
+ PERL_ARGS_ASSERT_NEW_MSG_HV;
+
+ (void) hv_stores(msg_hv, "text", msg_sv);
+ (void) hv_stores(msg_hv, "warn_categories", category_sv);
+ (void) hv_stores(msg_hv, "flag_bit", flag_bit_sv);
+
+ return msg_hv;
+}
+
/*
=for apidoc uvoffuni_to_utf8_flags
=cut
*/
+U8 *
+Perl_uvoffuni_to_utf8_flags(pTHX_ U8 *d, UV uv, const UV flags)
+{
+ PERL_ARGS_ASSERT_UVOFFUNI_TO_UTF8_FLAGS;
+
+ return uvoffuni_to_utf8_flags_msgs(d, uv, flags, NULL);
+}
+
/* All these formats take a single UV code point argument */
const char surrogate_cp_format[] = "UTF-16 surrogate U+%04" UVXf;
const char nonchar_cp_format[] = "Unicode non-character U+%04" UVXf
" Unicode, requires a Perl extension," \
" and so is not portable";
-#define HANDLE_UNICODE_SURROGATE(uv, flags) \
+#define HANDLE_UNICODE_SURROGATE(uv, flags, msgs) \
STMT_START { \
if (flags & UNICODE_WARN_SURROGATE) { \
- Perl_ck_warner_d(aTHX_ packWARN(WARN_SURROGATE), \
- surrogate_cp_format, uv); \
+ U32 category = packWARN(WARN_SURROGATE); \
+ const char * format = surrogate_cp_format; \
+ if (msgs) { \
+ *msgs = new_msg_hv(Perl_form(aTHX_ format, uv), \
+ category, \
+ UNICODE_GOT_SURROGATE); \
+ } \
+ else { \
+ Perl_ck_warner_d(aTHX_ category, format, uv); \
+ } \
} \
if (flags & UNICODE_DISALLOW_SURROGATE) { \
return NULL; \
} \
} STMT_END;
-#define HANDLE_UNICODE_NONCHAR(uv, flags) \
+#define HANDLE_UNICODE_NONCHAR(uv, flags, msgs) \
STMT_START { \
if (flags & UNICODE_WARN_NONCHAR) { \
- Perl_ck_warner_d(aTHX_ packWARN(WARN_NONCHAR), \
- nonchar_cp_format, uv); \
+ U32 category = packWARN(WARN_NONCHAR); \
+ const char * format = nonchar_cp_format; \
+ if (msgs) { \
+ *msgs = new_msg_hv(Perl_form(aTHX_ format, uv), \
+ category, \
+ UNICODE_GOT_NONCHAR); \
+ } \
+ else { \
+ Perl_ck_warner_d(aTHX_ category, format, uv); \
+ } \
} \
if (flags & UNICODE_DISALLOW_NONCHAR) { \
return NULL; \
#define MARK UTF_CONTINUATION_MARK
#define MASK UTF_CONTINUATION_MASK
+/*
+=for apidoc uvchr_to_utf8_flags_msgs
+
+THIS FUNCTION SHOULD BE USED IN ONLY VERY SPECIALIZED CIRCUMSTANCES.
+
+Most code should use C<L</uvchr_to_utf8_flags>()> rather than call this directly.
+
+This function is for code that wants any warning and/or error messages to be
+returned to the caller rather than be displayed. All messages that would have
+been displayed if all lexical warnings are enabled will be returned.
+
+It is just like C<L</uvchr_to_utf8_flags>> but it takes an extra parameter
+placed after all the others, C<msgs>. If this parameter is 0, this function
+behaves identically to C<L</uvchr_to_utf8_flags>>. Otherwise, C<msgs> should
+be a pointer to an C<HV *> variable, in which this function creates a new HV to
+contain any appropriate messages. The hash has three key-value pairs, as
+follows:
+
+=over 4
+
+=item C<text>
+
+The text of the message as a C<SVpv>.
+
+=item C<warn_categories>
+
+The warning category (or categories) packed into a C<SVuv>.
+
+=item C<flag>
+
+A single flag bit associated with this message, in a C<SVuv>.
+The bit corresponds to some bit in the C<*errors> return value,
+such as C<UNICODE_GOT_SURROGATE>.
+
+=back
+
+It's important to note that specifying this parameter as non-null will cause
+any warnings this function would otherwise generate to be suppressed, and
+instead be placed in C<*msgs>. The caller can check the lexical warnings state
+(or not) when choosing what to do with the returned messages.
+
+The caller, of course, is responsible for freeing any returned HV.
+
+=cut
+*/
+
+/* Undocumented; we don't want people using this. Instead they should use
+ * uvchr_to_utf8_flags_msgs() */
U8 *
-Perl_uvoffuni_to_utf8_flags(pTHX_ U8 *d, UV uv, const UV flags)
+Perl_uvoffuni_to_utf8_flags_msgs(pTHX_ U8 *d, UV uv, const UV flags, HV** msgs)
{
- PERL_ARGS_ASSERT_UVOFFUNI_TO_UTF8_FLAGS;
+ PERL_ARGS_ASSERT_UVOFFUNI_TO_UTF8_FLAGS_MSGS;
+
+ if (msgs) {
+ *msgs = NULL;
+ }
if (OFFUNI_IS_INVARIANT(uv)) {
*d++ = LATIN1_TO_NATIVE(uv);
if (UNLIKELY( UNICODE_IS_32_CONTIGUOUS_NONCHARS(uv)
|| UNICODE_IS_END_PLANE_NONCHAR_GIVEN_NOT_SUPER(uv)))
{
- HANDLE_UNICODE_NONCHAR(uv, flags);
+ HANDLE_UNICODE_NONCHAR(uv, flags, msgs);
}
else if (UNLIKELY(UNICODE_IS_SURROGATE(uv))) {
- HANDLE_UNICODE_SURROGATE(uv, flags);
+ HANDLE_UNICODE_SURROGATE(uv, flags, msgs);
}
}
#endif
* performance hit on these high EBCDIC code points. */
if (UNLIKELY(UNICODE_IS_SUPER(uv))) {
- if (UNLIKELY(uv > MAX_EXTERNALLY_LEGAL_CP)) {
- Perl_croak(aTHX_ cp_above_legal_max, uv, MAX_EXTERNALLY_LEGAL_CP);
+ if (UNLIKELY(uv > MAX_LEGAL_CP)) {
+ Perl_croak(aTHX_ cp_above_legal_max, uv, MAX_LEGAL_CP);
}
- if ( (flags & UNICODE_WARN_SUPER)
- || ( (flags & UNICODE_WARN_PERL_EXTENDED)
+ if ( (flags & UNICODE_WARN_SUPER)
+ || ( (flags & UNICODE_WARN_PERL_EXTENDED)
&& UNICODE_IS_PERL_EXTENDED(uv)))
{
- Perl_ck_warner_d(aTHX_ packWARN(WARN_NON_UNICODE),
+ const char * format = super_cp_format;
+ U32 category = packWARN(WARN_NON_UNICODE);
+ U32 flag = UNICODE_GOT_SUPER;
+
+ /* Choose the more dire applicable warning */
+ if (UNICODE_IS_PERL_EXTENDED(uv)) {
+ format = perl_extended_cp_format;
+ if (flags & (UNICODE_WARN_PERL_EXTENDED
+ |UNICODE_DISALLOW_PERL_EXTENDED))
+ {
+ flag = UNICODE_GOT_PERL_EXTENDED;
+ }
+ }
- /* Choose the more dire applicable warning */
- (UNICODE_IS_PERL_EXTENDED(uv))
- ? perl_extended_cp_format
- : super_cp_format,
- uv);
+ if (msgs) {
+ *msgs = new_msg_hv(Perl_form(aTHX_ format, uv),
+ category, flag);
+ }
+ else {
+ Perl_ck_warner_d(aTHX_ packWARN(WARN_NON_UNICODE), format, uv);
+ }
}
if ( (flags & UNICODE_DISALLOW_SUPER)
|| ( (flags & UNICODE_DISALLOW_PERL_EXTENDED)
}
}
else if (UNLIKELY(UNICODE_IS_END_PLANE_NONCHAR_GIVEN_NOT_SUPER(uv))) {
- HANDLE_UNICODE_NONCHAR(uv, flags);
+ HANDLE_UNICODE_NONCHAR(uv, flags, msgs);
}
/* Test for and handle 4-byte result. In the test immediately below, the
characters. The end-plane non-characters for EBCDIC were
handled just above */
if (UNLIKELY(UNICODE_IS_32_CONTIGUOUS_NONCHARS(uv))) {
- HANDLE_UNICODE_NONCHAR(uv, flags);
+ HANDLE_UNICODE_NONCHAR(uv, flags, msgs);
}
else if (UNLIKELY(UNICODE_IS_SURROGATE(uv))) {
- HANDLE_UNICODE_SURROGATE(uv, flags);
+ HANDLE_UNICODE_SURROGATE(uv, flags, msgs);
}
#endif
STRLEN len = OFFUNISKIP(uv);
U8 *p = d+len-1;
while (p > d) {
- *p-- = I8_TO_NATIVE_UTF8((uv & UTF_CONTINUATION_MASK) | UTF_CONTINUATION_MARK);
- uv >>= UTF_ACCUMULATION_SHIFT;
+ *p-- = I8_TO_NATIVE_UTF8((uv & MASK) | MARK);
+ uv >>= SHIFT;
}
*p = I8_TO_NATIVE_UTF8((uv & UTF_START_MASK(len)) | UTF_START_MARK(len));
return d+len;
A somewhat misleadingly named synonym for C<UNICODE_WARN_PERL_EXTENDED> is
retained for backward compatibility: C<UNICODE_WARN_ABOVE_31_BIT>. Similarly,
C<UNICODE_DISALLOW_ABOVE_31_BIT> is usable instead of the more accurately named
-C<UNICODE_DISALLOW_PERL_EXTENDED>. The names are misleading because these
-flags can apply to code points that actually do fit in 31 bits. This happens
-on EBCDIC platforms, and sometimes when the L<overlong
-malformation|/C<UTF8_GOT_LONG>> is also present. The new names accurately
-describe the situation in all cases.
+C<UNICODE_DISALLOW_PERL_EXTENDED>. The names are misleading because on EBCDIC
+platforms,these flags can apply to code points that actually do fit in 31 bits.
+The new names accurately describe the situation in all cases.
=cut
*/
PERL_STATIC_INLINE char *
S_unexpected_non_continuation_text(pTHX_ const U8 * const s,
- /* How many bytes to print */
+ /* Max number of bytes to print */
STRLEN print_len,
/* Which one is the non-continuation */
? "immediately"
: Perl_form(aTHX_ "%d bytes",
(int) non_cont_byte_pos);
+ const U8 * x = s + non_cont_byte_pos;
+ const U8 * e = s + print_len;
PERL_ARGS_ASSERT_UNEXPECTED_NON_CONTINUATION_TEXT;
* calculated, it's likely faster to pass it; verify under DEBUGGING */
assert(expect_len == UTF8SKIP(s));
+ /* As a defensive coding measure, don't output anything past a NUL. Such
+ * bytes shouldn't be in the middle of a malformation, and could mark the
+ * end of the allocated string, and what comes after is undefined */
+ for (; x < e; x++) {
+ if (*x == '\0') {
+ x++; /* Output this particular NUL */
+ break;
+ }
+ }
+
return Perl_form(aTHX_ "%s: %s (unexpected non-continuation byte 0x%02x,"
" %s after start byte 0x%02x; need %d bytes, got %d)",
malformed_text,
- _byte_dump_string(s, print_len, 0),
+ _byte_dump_string(s, x - s, 0),
*(s + non_cont_byte_pos),
where,
*s,
*/
UV
-Perl_utf8n_to_uvchr(pTHX_ const U8 *s,
- STRLEN curlen,
- STRLEN *retlen,
- const U32 flags)
+Perl_utf8n_to_uvchr(const U8 *s,
+ STRLEN curlen,
+ STRLEN *retlen,
+ const U32 flags)
{
PERL_ARGS_ASSERT_UTF8N_TO_UVCHR;
Most code should use L</utf8_to_uvchr_buf>() rather than call this directly.
This function is for code that needs to know what the precise malformation(s)
-are when an error is found.
+are when an error is found. If you also need to know the generated warning
+messages, use L</utf8n_to_uvchr_msgs>() instead.
It is like C<L</utf8n_to_uvchr>> but it takes an extra parameter placed after
all the others, C<errors>. If this parameter is 0, this function behaves
=item C<UTF8_GOT_NON_CONTINUATION>
The input sequence was malformed in that a non-continuation type byte was found
-in a position where only a continuation type one should be.
+in a position where only a continuation type one should be. See also
+L</C<UTF8_GOT_SHORT>>.
=item C<UTF8_GOT_OVERFLOW>
a complete sequence. In other words, the input is for a partial character
sequence.
+
+C<UTF8_GOT_SHORT> and C<UTF8_GOT_NON_CONTINUATION> both indicate a too short
+sequence. The difference is that C<UTF8_GOT_NON_CONTINUATION> indicates always
+that there is an error, while C<UTF8_GOT_SHORT> means that an incomplete
+sequence was looked at. If no other flags are present, it means that the
+sequence was valid as far as it went. Depending on the application, this could
+mean one of three things:
+
+=over
+
+=item *
+
+The C<curlen> length parameter passed in was too small, and the function was
+prevented from examining all the necessary bytes.
+
+=item *
+
+The buffer being looked at is based on reading data, and the data received so
+far stopped in the middle of a character, so that the next read will
+read the remainder of this character. (It is up to the caller to deal with the
+split bytes somehow.)
+
+=item *
+
+This is a real error, and the partial sequence is all we're going to get.
+
+=back
+
=item C<UTF8_GOT_SUPER>
The input sequence was malformed in that it is for a non-Unicode code point;
flag to suppress any warnings, and then examine the C<*errors> return.
=cut
+
+Also implemented as a macro in utf8.h
+*/
+
+UV
+Perl_utf8n_to_uvchr_error(const U8 *s,
+ STRLEN curlen,
+ STRLEN *retlen,
+ const U32 flags,
+ U32 * errors)
+{
+ PERL_ARGS_ASSERT_UTF8N_TO_UVCHR_ERROR;
+
+ return utf8n_to_uvchr_msgs(s, curlen, retlen, flags, errors, NULL);
+}
+
+/*
+
+=for apidoc utf8n_to_uvchr_msgs
+
+THIS FUNCTION SHOULD BE USED IN ONLY VERY SPECIALIZED CIRCUMSTANCES.
+Most code should use L</utf8_to_uvchr_buf>() rather than call this directly.
+
+This function is for code that needs to know what the precise malformation(s)
+are when an error is found, and wants the corresponding warning and/or error
+messages to be returned to the caller rather than be displayed. All messages
+that would have been displayed if all lexcial warnings are enabled will be
+returned.
+
+It is just like C<L</utf8n_to_uvchr_error>> but it takes an extra parameter
+placed after all the others, C<msgs>. If this parameter is 0, this function
+behaves identically to C<L</utf8n_to_uvchr_error>>. Otherwise, C<msgs> should
+be a pointer to an C<AV *> variable, in which this function creates a new AV to
+contain any appropriate messages. The elements of the array are ordered so
+that the first message that would have been displayed is in the 0th element,
+and so on. Each element is a hash with three key-value pairs, as follows:
+
+=over 4
+
+=item C<text>
+
+The text of the message as a C<SVpv>.
+
+=item C<warn_categories>
+
+The warning category (or categories) packed into a C<SVuv>.
+
+=item C<flag>
+
+A single flag bit associated with this message, in a C<SVuv>.
+The bit corresponds to some bit in the C<*errors> return value,
+such as C<UTF8_GOT_LONG>.
+
+=back
+
+It's important to note that specifying this parameter as non-null will cause
+any warnings this function would otherwise generate to be suppressed, and
+instead be placed in C<*msgs>. The caller can check the lexical warnings state
+(or not) when choosing what to do with the returned messages.
+
+If the flag C<UTF8_CHECK_ONLY> is passed, no warnings are generated, and hence
+no AV is created.
+
+The caller, of course, is responsible for freeing any returned AV.
+
+=cut
*/
UV
-Perl_utf8n_to_uvchr_error(pTHX_ const U8 *s,
- STRLEN curlen,
- STRLEN *retlen,
- const U32 flags,
- U32 * errors)
+Perl__utf8n_to_uvchr_msgs_helper(const U8 *s,
+ STRLEN curlen,
+ STRLEN *retlen,
+ const U32 flags,
+ U32 * errors,
+ AV ** msgs)
{
const U8 * const s0 = s;
- U8 * send = NULL; /* (initialized to silence compilers' wrong
- warning) */
- U32 possible_problems = 0; /* A bit is set here for each potential problem
- found as we go along */
- UV uv = *s;
- STRLEN expectlen = 0; /* How long should this sequence be?
- (initialized to silence compilers' wrong
- warning) */
- STRLEN avail_len = 0; /* When input is too short, gives what that is */
- U32 discard_errors = 0; /* Used to save branches when 'errors' is NULL;
- this gets set and discarded */
+ const U8 * send = s0 + curlen;
+ U32 possible_problems; /* A bit is set here for each potential problem
+ found as we go along */
+ UV uv;
+ STRLEN expectlen; /* How long should this sequence be? */
+ STRLEN avail_len; /* When input is too short, gives what that is */
+ U32 discard_errors; /* Used to save branches when 'errors' is NULL; this
+ gets set and discarded */
/* The below are used only if there is both an overlong malformation and a
* too short one. Otherwise the first two are set to 's0' and 'send', and
* the third not used at all */
- U8 * adjusted_s0 = (U8 *) s0;
+ U8 * adjusted_s0;
U8 temp_char_buf[UTF8_MAXBYTES + 1]; /* Used to avoid a Newx in this
routine; see [perl #130921] */
- UV uv_so_far = 0; /* (Initialized to silence compilers' wrong warning) */
+ UV uv_so_far;
+ dTHX;
- PERL_ARGS_ASSERT_UTF8N_TO_UVCHR_ERROR;
+ PERL_ARGS_ASSERT__UTF8N_TO_UVCHR_MSGS_HELPER;
+
+ /* Here, is one of: a) malformed; b) a problematic code point (surrogate,
+ * non-unicode, or nonchar); or c) on ASCII platforms, one of the Hangul
+ * syllables that the dfa doesn't properly handle. Quickly dispose of the
+ * final case. */
+
+#ifndef EBCDIC
+
+ /* Each of the affected Hanguls starts with \xED */
+
+ if (is_HANGUL_ED_utf8_safe(s0, send)) {
+ if (retlen) {
+ *retlen = 3;
+ }
+ if (errors) {
+ *errors = 0;
+ }
+ if (msgs) {
+ *msgs = NULL;
+ }
+
+ return ((0xED & UTF_START_MASK(3)) << (2 * UTF_ACCUMULATION_SHIFT))
+ | ((s0[1] & UTF_CONTINUATION_MASK) << UTF_ACCUMULATION_SHIFT)
+ | (s0[2] & UTF_CONTINUATION_MASK);
+ }
+
+#endif
+
+ /* In conjunction with the exhaustive tests that can be enabled in
+ * APItest/t/utf8_warn_base.pl, this can make sure the dfa does precisely
+ * what it is intended to do, and that no flaws in it are masked by
+ * dropping down and executing the code below
+ assert(! isUTF8_CHAR(s0, send)
+ || UTF8_IS_SURROGATE(s0, send)
+ || UTF8_IS_SUPER(s0, send)
+ || UTF8_IS_NONCHAR(s0,send));
+ */
+
+ s = s0;
+ uv = *s0;
+ possible_problems = 0;
+ expectlen = 0;
+ avail_len = 0;
+ discard_errors = 0;
+ adjusted_s0 = (U8 *) s0;
+ uv_so_far = 0;
if (errors) {
*errors = 0;
*retlen = expectlen;
}
- /* An invariant is trivially well-formed */
- if (UTF8_IS_INVARIANT(uv)) {
- return uv;
- }
-
/* A continuation character can't start a valid sequence */
if (UNLIKELY(UTF8_IS_CONTINUATION(uv))) {
possible_problems |= UTF8_GOT_CONTINUATION;
/* Setup the loop end point, making sure to not look past the end of the
* input string, and flag it as too short if the size isn't big enough. */
- send = (U8*) s0;
if (UNLIKELY(curlen < expectlen)) {
possible_problems |= UTF8_GOT_SHORT;
avail_len = curlen;
- send += curlen;
}
else {
- send += expectlen;
+ send = (U8*) s0 + expectlen;
}
/* Now, loop through the remaining bytes in the character's sequence,
bool disallowed = FALSE;
const U32 orig_problems = possible_problems;
+ if (msgs) {
+ *msgs = NULL;
+ }
+
while (possible_problems) { /* Handle each possible problem */
UV pack_warn = 0;
char * message = NULL;
+ U32 this_flag_bit = 0;
/* Each 'if' clause handles one problem. They are ordered so that
* the first ones' messages will be displayed before the later
* necessarily do so in the future. We output (only) the
* most dire warning */
if (! (flags & UTF8_CHECK_ONLY)) {
- if (ckWARN_d(WARN_UTF8)) {
+ if (msgs || ckWARN_d(WARN_UTF8)) {
pack_warn = packWARN(WARN_UTF8);
}
- else if (ckWARN_d(WARN_NON_UNICODE)) {
+ else if (msgs || ckWARN_d(WARN_NON_UNICODE)) {
pack_warn = packWARN(WARN_NON_UNICODE);
}
if (pack_warn) {
message = Perl_form(aTHX_ "%s: %s (overflows)",
malformed_text,
_byte_dump_string(s0, curlen, 0));
+ this_flag_bit = UTF8_GOT_OVERFLOW;
}
}
}
assert(0);
disallowed = TRUE;
- if (ckWARN_d(WARN_UTF8) && ! (flags & UTF8_CHECK_ONLY)) {
+ if ( (msgs
+ || ckWARN_d(WARN_UTF8)) && ! (flags & UTF8_CHECK_ONLY))
+ {
pack_warn = packWARN(WARN_UTF8);
message = Perl_form(aTHX_ "%s (empty string)",
malformed_text);
+ this_flag_bit = UTF8_GOT_EMPTY;
}
}
}
if (! (flags & UTF8_ALLOW_CONTINUATION)) {
disallowed = TRUE;
- if (ckWARN_d(WARN_UTF8) && ! (flags & UTF8_CHECK_ONLY)) {
+ if (( msgs
+ || ckWARN_d(WARN_UTF8)) && ! (flags & UTF8_CHECK_ONLY))
+ {
pack_warn = packWARN(WARN_UTF8);
message = Perl_form(aTHX_
"%s: %s (unexpected continuation byte 0x%02x,"
" with no preceding start byte)",
malformed_text,
_byte_dump_string(s0, 1, 0), *s0);
+ this_flag_bit = UTF8_GOT_CONTINUATION;
}
}
}
if (! (flags & UTF8_ALLOW_SHORT)) {
disallowed = TRUE;
- if (ckWARN_d(WARN_UTF8) && ! (flags & UTF8_CHECK_ONLY)) {
+ if (( msgs
+ || ckWARN_d(WARN_UTF8)) && ! (flags & UTF8_CHECK_ONLY))
+ {
pack_warn = packWARN(WARN_UTF8);
message = Perl_form(aTHX_
"%s: %s (too short; %d byte%s available, need %d)",
(int)avail_len,
avail_len == 1 ? "" : "s",
(int)expectlen);
+ this_flag_bit = UTF8_GOT_SHORT;
}
}
if (! (flags & UTF8_ALLOW_NON_CONTINUATION)) {
disallowed = TRUE;
- if (ckWARN_d(WARN_UTF8) && ! (flags & UTF8_CHECK_ONLY)) {
+ if (( msgs
+ || ckWARN_d(WARN_UTF8)) && ! (flags & UTF8_CHECK_ONLY))
+ {
/* If we don't know for sure that the input length is
* valid, avoid as much as possible reading past the
printlen,
s - s0,
(int) expectlen));
+ this_flag_bit = UTF8_GOT_NON_CONTINUATION;
}
}
}
*errors |= UTF8_GOT_SURROGATE;
if ( ! (flags & UTF8_CHECK_ONLY)
- && ckWARN_d(WARN_SURROGATE))
+ && (msgs || ckWARN_d(WARN_SURROGATE)))
{
pack_warn = packWARN(WARN_SURROGATE);
else {
message = Perl_form(aTHX_ surrogate_cp_format, uv);
}
+ this_flag_bit = UTF8_GOT_SURROGATE;
}
}
*errors |= UTF8_GOT_SUPER;
if ( ! (flags & UTF8_CHECK_ONLY)
- && ckWARN_d(WARN_NON_UNICODE))
+ && (msgs || ckWARN_d(WARN_NON_UNICODE)))
{
pack_warn = packWARN(WARN_NON_UNICODE);
else {
message = Perl_form(aTHX_ super_cp_format, uv);
}
+ this_flag_bit = UTF8_GOT_SUPER;
}
}
if (UNLIKELY(isUTF8_PERL_EXTENDED(s0))) {
if ( ! (flags & UTF8_CHECK_ONLY)
&& (flags & (UTF8_WARN_PERL_EXTENDED|UTF8_WARN_SUPER))
- && ckWARN_d(WARN_NON_UNICODE))
+ && (msgs || ckWARN_d(WARN_NON_UNICODE)))
{
pack_warn = packWARN(WARN_NON_UNICODE);
" so is not portable",
_byte_dump_string(s0, curlen, 0));
}
+ this_flag_bit = UTF8_GOT_PERL_EXTENDED;
}
if (flags & ( UTF8_WARN_PERL_EXTENDED
*errors |= UTF8_GOT_NONCHAR;
if ( ! (flags & UTF8_CHECK_ONLY)
- && ckWARN_d(WARN_NONCHAR))
+ && (msgs || ckWARN_d(WARN_NONCHAR)))
{
/* The code above should have guaranteed that we don't
* get here with errors other than overlong */
pack_warn = packWARN(WARN_NONCHAR);
message = Perl_form(aTHX_ nonchar_cp_format, uv);
+ this_flag_bit = UTF8_GOT_NONCHAR;
}
}
else {
disallowed = TRUE;
- if (ckWARN_d(WARN_UTF8) && ! (flags & UTF8_CHECK_ONLY)) {
+ if (( msgs
+ || ckWARN_d(WARN_UTF8)) && ! (flags & UTF8_CHECK_ONLY))
+ {
pack_warn = packWARN(WARN_UTF8);
/* These error types cause 'uv' to be something that
small code points */
UNI_TO_NATIVE(uv));
}
+ this_flag_bit = UTF8_GOT_LONG;
}
}
} /* End of looking through the possible flags */
/* Display the message (if any) for the problem being handled in
* this iteration of the loop */
if (message) {
- if (PL_op)
+ if (msgs) {
+ assert(this_flag_bit);
+
+ if (*msgs == NULL) {
+ *msgs = newAV();
+ }
+
+ av_push(*msgs, newRV_noinc((SV*) new_msg_hv(message,
+ pack_warn,
+ this_flag_bit)));
+ }
+ else if (PL_op)
Perl_warner(aTHX_ pack_warn, "%s in %s", message,
OP_DESC(PL_op));
else
Only in very rare circumstances should code need to be dealing in Unicode
(as opposed to native) code points. In those few cases, use
-C<L<NATIVE_TO_UNI(utf8_to_uvchr_buf(...))|/utf8_to_uvchr_buf>> instead.
+C<L<NATIVE_TO_UNI(utf8_to_uvchr_buf(...))|/utf8_to_uvchr_buf>> instead. If you
+are not absolutely sure this is one of those cases, then assume it isn't and
+use plain C<utf8_to_uvchr_buf> instead.
Returns the Unicode (not-native) code point of the first character in the
string C<s> which
/*
=for apidoc utf8_length
-Return the length of the UTF-8 char encoded string C<s> in characters.
-Stops at C<e> (i.e. the C<*e> byte does not form part of the character).
-If C<e E<lt> s> or if the scan would end up past C<e>, it croaks.
+Returns the number of characters in the sequence of UTF-8-encoded bytes starting
+at C<s> and ending at the byte just before C<e>. If <s> and <e> point to the
+same place, it returns 0 with no warning raised.
+
+If C<e E<lt> s> or if the scan would end up past C<e>, it raises a UTF8 warning
+and returns the number of valid characters.
=cut
*/
Otherwise, C<*is_utf8p> is set to 0, and the return value is a pointer to a
newly created string containing a downgraded copy of C<s>, and whose length is
-returned in C<*lenp>, updated. The new string is C<NUL>-terminated.
+returned in C<*lenp>, updated. The new string is C<NUL>-terminated. The
+caller is responsible for arranging for the memory used by this string to get
+freed.
Upon successful return, the number of variants in the string can be computed by
having saved the value of C<*lenp> before the call, and subtracting the
}
finish_and_return:
- *d = '\0';
- *lenp = d - converted_start;
+ *d = '\0';
+ *lenp = d - converted_start;
/* Trim unused space */
Renew(converted_start, *lenp + 1, U8);
Converts a string C<s> of length C<*lenp> bytes from the native encoding into
UTF-8.
Returns a pointer to the newly-created string, and sets C<*lenp> to
-reflect the new length in bytes.
+reflect the new length in bytes. The caller is responsible for arranging for
+the memory used by this string to get freed.
Upon successful return, the number of variants in the string can be computed by
having saved the value of C<*lenp> before the call, and subtracting it from the
PERL_ARGS_ASSERT_BYTES_TO_UTF8;
PERL_UNUSED_CONTEXT;
- Newx(d, (*lenp) * 2 + 1, U8);
+ /* 1 for each byte + 1 for each byte that expands to two, + trailing NUL */
+ Newx(d, (*lenp) + variant_under_utf8_count(s, send) + 1, U8);
dst = d;
while (s < send) {
*d = '\0';
*lenp = d-dst;
- /* Trim unused space */
- Renew(dst, *lenp + 1, U8);
-
return dst;
}
bool
Perl__is_uni_FOO(pTHX_ const U8 classnum, const UV c)
{
- U8 tmpbuf[UTF8_MAXBYTES+1];
- uvchr_to_utf8(tmpbuf, c);
- return _is_utf8_FOO_with_len(classnum, tmpbuf, tmpbuf + sizeof(tmpbuf));
+ return _invlist_contains_cp(PL_XPosix_ptrs[classnum], c);
}
/* Internal function so we can deprecate the external one, and call
if (*p == '_')
return TRUE;
- return is_utf8_common(p, &PL_utf8_idstart, "IdStart", NULL);
+ return is_utf8_common(p, PL_utf8_idstart);
}
bool
Perl__is_uni_perl_idcont(pTHX_ UV c)
{
- U8 tmpbuf[UTF8_MAXBYTES+1];
- uvchr_to_utf8(tmpbuf, c);
- return _is_utf8_perl_idcont_with_len(tmpbuf, tmpbuf + sizeof(tmpbuf));
+ return _invlist_contains_cp(PL_utf8_perl_idcont, c);
}
bool
Perl__is_uni_perl_idstart(pTHX_ UV c)
{
- U8 tmpbuf[UTF8_MAXBYTES+1];
- uvchr_to_utf8(tmpbuf, c);
- return _is_utf8_perl_idstart_with_len(tmpbuf, tmpbuf + sizeof(tmpbuf));
+ return _invlist_contains_cp(PL_utf8_perl_idstart, c);
}
UV
return converted;
}
+/* If compiled on an early Unicode version, there may not be auxiliary tables
+ * */
+#ifndef HAS_UC_AUX_TABLES
+# define UC_AUX_TABLE_ptrs NULL
+# define UC_AUX_TABLE_lengths NULL
+#endif
+#ifndef HAS_TC_AUX_TABLES
+# define TC_AUX_TABLE_ptrs NULL
+# define TC_AUX_TABLE_lengths NULL
+#endif
+#ifndef HAS_LC_AUX_TABLES
+# define LC_AUX_TABLE_ptrs NULL
+# define LC_AUX_TABLE_lengths NULL
+#endif
+#ifndef HAS_CF_AUX_TABLES
+# define CF_AUX_TABLE_ptrs NULL
+# define CF_AUX_TABLE_lengths NULL
+#endif
+#ifndef HAS_UC_AUX_TABLES
+# define UC_AUX_TABLE_ptrs NULL
+# define UC_AUX_TABLE_lengths NULL
+#endif
+
/* Call the function to convert a UTF-8 encoded character to the specified case.
* Note that there may be more than one character in the result.
- * INP is a pointer to the first byte of the input character
- * OUTP will be set to the first byte of the string of changed characters. It
+ * 's' is a pointer to the first byte of the input character
+ * 'd' will be set to the first byte of the string of changed characters. It
* needs to have space for UTF8_MAXBYTES_CASE+1 bytes
- * LENP will be set to the length in bytes of the string of changed characters
+ * 'lenp' will be set to the length in bytes of the string of changed characters
*
* The functions return the ordinal of the first character in the string of
- * OUTP */
+ * 'd' */
#define CALL_UPPER_CASE(uv, s, d, lenp) \
- _to_utf8_case(uv, s, d, lenp, &PL_utf8_toupper, "ToUc", "")
+ _to_utf8_case(uv, s, d, lenp, PL_utf8_toupper, \
+ Uppercase_Mapping_invmap, \
+ UC_AUX_TABLE_ptrs, \
+ UC_AUX_TABLE_lengths, \
+ "uppercase")
#define CALL_TITLE_CASE(uv, s, d, lenp) \
- _to_utf8_case(uv, s, d, lenp, &PL_utf8_totitle, "ToTc", "")
+ _to_utf8_case(uv, s, d, lenp, PL_utf8_totitle, \
+ Titlecase_Mapping_invmap, \
+ TC_AUX_TABLE_ptrs, \
+ TC_AUX_TABLE_lengths, \
+ "titlecase")
#define CALL_LOWER_CASE(uv, s, d, lenp) \
- _to_utf8_case(uv, s, d, lenp, &PL_utf8_tolower, "ToLc", "")
+ _to_utf8_case(uv, s, d, lenp, PL_utf8_tolower, \
+ Lowercase_Mapping_invmap, \
+ LC_AUX_TABLE_ptrs, \
+ LC_AUX_TABLE_lengths, \
+ "lowercase")
+
/* This additionally has the input parameter 'specials', which if non-zero will
* cause this to use the specials hash for folding (meaning get full case
* folding); otherwise, when zero, this implies a simple case fold */
#define CALL_FOLD_CASE(uv, s, d, lenp, specials) \
-_to_utf8_case(uv, s, d, lenp, &PL_utf8_tofold, "ToCf", (specials) ? "" : NULL)
+ (specials) \
+ ? _to_utf8_case(uv, s, d, lenp, PL_utf8_tofold, \
+ Case_Folding_invmap, \
+ CF_AUX_TABLE_ptrs, \
+ CF_AUX_TABLE_lengths, \
+ "foldcase") \
+ : _to_utf8_case(uv, s, d, lenp, PL_utf8_tosimplefold, \
+ Simple_Case_Folding_invmap, \
+ NULL, NULL, \
+ "foldcase")
UV
Perl_to_uni_upper(pTHX_ UV c, U8* p, STRLEN *lenp)
return _to_upper_title_latin1((U8) c, p, lenp, 'S');
}
- uvchr_to_utf8(p, c);
- return CALL_UPPER_CASE(c, p, p, lenp);
+ return CALL_UPPER_CASE(c, NULL, p, lenp);
}
UV
return _to_upper_title_latin1((U8) c, p, lenp, 's');
}
- uvchr_to_utf8(p, c);
- return CALL_TITLE_CASE(c, p, p, lenp);
+ return CALL_TITLE_CASE(c, NULL, p, lenp);
}
STATIC U8
return to_lower_latin1((U8) c, p, lenp, 0 /* 0 is a dummy arg */ );
}
- uvchr_to_utf8(p, c);
- return CALL_LOWER_CASE(c, p, p, lenp);
+ return CALL_LOWER_CASE(c, NULL, p, lenp);
}
UV
-Perl__to_fold_latin1(pTHX_ const U8 c, U8* p, STRLEN *lenp,
- const unsigned int flags)
+Perl__to_fold_latin1(const U8 c, U8* p, STRLEN *lenp, const unsigned int flags)
{
/* Corresponds to to_lower_latin1(); <flags> bits meanings:
* FOLD_FLAGS_NOMIX_ASCII iff non-ASCII to ASCII folds are prohibited
UV converted;
PERL_ARGS_ASSERT__TO_FOLD_LATIN1;
- PERL_UNUSED_CONTEXT;
assert (! (flags & FOLD_FLAGS_LOCALE));
PERL_ARGS_ASSERT__TO_UNI_FOLD_FLAGS;
if (flags & FOLD_FLAGS_LOCALE) {
- /* Treat a UTF-8 locale as not being in locale at all */
+ /* Treat a UTF-8 locale as not being in locale at all, except for
+ * potentially warning */
+ _CHECK_AND_WARN_PROBLEMATIC_LOCALE;
if (IN_UTF8_CTYPE_LOCALE) {
flags &= ~FOLD_FLAGS_LOCALE;
}
else {
- _CHECK_AND_WARN_PROBLEMATIC_LOCALE;
goto needs_full_generality;
}
}
/* Here, above 255. If no special needs, just use the macro */
if ( ! (flags & (FOLD_FLAGS_LOCALE|FOLD_FLAGS_NOMIX_ASCII))) {
- uvchr_to_utf8(p, c);
- return CALL_FOLD_CASE(c, p, p, lenp, flags & FOLD_FLAGS_FULL);
+ return CALL_FOLD_CASE(c, NULL, p, lenp, flags & FOLD_FLAGS_FULL);
}
else { /* Otherwise, _toFOLD_utf8_flags has the intelligence to deal with
the special flags. */
}
PERL_STATIC_INLINE bool
-S_is_utf8_common(pTHX_ const U8 *const p, SV **swash,
- const char *const swashname, SV* const invlist)
+S_is_utf8_common(pTHX_ const U8 *const p, SV* const invlist)
{
/* returns a boolean giving whether or not the UTF8-encoded character that
- * starts at <p> is in the swash indicated by <swashname>. <swash>
- * contains a pointer to where the swash indicated by <swashname>
- * is to be stored; which this routine will do, so that future calls will
- * look at <*swash> and only generate a swash if it is not null. <invlist>
- * is NULL or an inversion list that defines the swash. If not null, it
- * saves time during initialization of the swash.
+ * starts at <p> is in the inversion list indicated by <invlist>.
*
* Note that it is assumed that the buffer length of <p> is enough to
* contain all the bytes that comprise the character. Thus, <*p> should
* have been checked before this call for mal-formedness enough to assure
- * that. */
+ * that. This function, does make sure to not look past any NUL, so it is
+ * safe to use on C, NUL-terminated, strings */
+ STRLEN len = my_strnlen((char *) p, UTF8SKIP(p));
PERL_ARGS_ASSERT_IS_UTF8_COMMON;
* as far as there being enough bytes available in it to accommodate the
* character without reading beyond the end, and pass that number on to the
* validating routine */
- if (! isUTF8_CHAR(p, p + UTF8SKIP(p))) {
- _force_out_malformed_utf8_message(p, p + UTF8SKIP(p),
- _UTF8_NO_CONFIDENCE_IN_CURLEN,
+ if (! isUTF8_CHAR(p, p + len)) {
+ _force_out_malformed_utf8_message(p, p + len, _UTF8_NO_CONFIDENCE_IN_CURLEN,
1 /* Die */ );
NOT_REACHED; /* NOTREACHED */
}
- if (!*swash) {
- U8 flags = _CORE_SWASH_INIT_ACCEPT_INVLIST;
- *swash = _core_swash_init("utf8",
-
- /* Only use the name if there is no inversion
- * list; otherwise will go out to disk */
- (invlist) ? "" : swashname,
-
- &PL_sv_undef, 1, 0, invlist, &flags);
- }
-
- return swash_fetch(*swash, p, TRUE) != 0;
+ return is_utf8_common_with_len(p, p + len, invlist);
}
PERL_STATIC_INLINE bool
S_is_utf8_common_with_len(pTHX_ const U8 *const p, const U8 * const e,
- SV **swash, const char *const swashname,
SV* const invlist)
{
/* returns a boolean giving whether or not the UTF8-encoded character that
- * starts at <p>, and extending no further than <e - 1> is in the swash
- * indicated by <swashname>. <swash> contains a pointer to where the swash
- * indicated by <swashname> is to be stored; which this routine will do, so
- * that future calls will look at <*swash> and only generate a swash if it
- * is not null. <invlist> is NULL or an inversion list that defines the
- * swash. If not null, it saves time during initialization of the swash.
- */
+ * starts at <p>, and extending no further than <e - 1> is in the inversion
+ * list <invlist>. */
+
+ UV cp = utf8n_to_uvchr(p, e - p, NULL, 0);
PERL_ARGS_ASSERT_IS_UTF8_COMMON_WITH_LEN;
- if (! isUTF8_CHAR(p, e)) {
+ if (cp == 0 && (p >= e || *p != '\0')) {
_force_out_malformed_utf8_message(p, e, 0, 1);
NOT_REACHED; /* NOTREACHED */
}
- if (!*swash) {
- U8 flags = _CORE_SWASH_INIT_ACCEPT_INVLIST;
- *swash = _core_swash_init("utf8",
-
- /* Only use the name if there is no inversion
- * list; otherwise will go out to disk */
- (invlist) ? "" : swashname,
-
- &PL_sv_undef, 1, 0, invlist, &flags);
- }
-
- return swash_fetch(*swash, p, TRUE) != 0;
+ assert(invlist);
+ return _invlist_contains_cp(invlist, cp);
}
STATIC void
if (instr(file, "mathoms.c")) {
Perl_warner(aTHX_ WARN_DEPRECATED,
- "In %s, line %d, starting in Perl v5.30, %s()"
+ "In %s, line %d, starting in Perl v5.32, %s()"
" will be removed. Avoid this message by"
" converting to use %s().\n",
file, line, name, alternative);
}
else {
Perl_warner(aTHX_ WARN_DEPRECATED,
- "In %s, line %d, starting in Perl v5.30, %s() will"
+ "In %s, line %d, starting in Perl v5.32, %s() will"
" require an additional parameter. Avoid this"
" message by converting to use %s().\n",
file, line, name, alternative);
case _CC_GRAPH:
case _CC_CASED:
- return is_utf8_common(p,
- &PL_utf8_swash_ptrs[classnum],
- swash_property_names[classnum],
- PL_XPosix_ptrs[classnum]);
+ return is_utf8_common(p, PL_XPosix_ptrs[classnum]);
case _CC_SPACE:
return is_XPERLSPACE_high(p);
case _CC_VERTSPACE:
return is_VERTWS_high(p);
case _CC_IDFIRST:
- if (! PL_utf8_perl_idstart) {
- PL_utf8_perl_idstart
- = _new_invlist_C_array(_Perl_IDStart_invlist);
- }
- return is_utf8_common(p, &PL_utf8_perl_idstart,
- "_Perl_IDStart", NULL);
+ return is_utf8_common(p, PL_utf8_perl_idstart);
case _CC_IDCONT:
- if (! PL_utf8_perl_idcont) {
- PL_utf8_perl_idcont
- = _new_invlist_C_array(_Perl_IDCont_invlist);
- }
- return is_utf8_common(p, &PL_utf8_perl_idcont,
- "_Perl_IDCont", NULL);
+ return is_utf8_common(p, PL_utf8_perl_idcont);
}
}
{
PERL_ARGS_ASSERT__IS_UTF8_FOO_WITH_LEN;
- assert(classnum < _FIRST_NON_SWASH_CC);
-
- return is_utf8_common_with_len(p,
- e,
- &PL_utf8_swash_ptrs[classnum],
- swash_property_names[classnum],
- PL_XPosix_ptrs[classnum]);
+ return is_utf8_common_with_len(p, e, PL_XPosix_ptrs[classnum]);
}
bool
Perl__is_utf8_perl_idstart_with_len(pTHX_ const U8 *p, const U8 * const e)
{
- SV* invlist = NULL;
-
PERL_ARGS_ASSERT__IS_UTF8_PERL_IDSTART_WITH_LEN;
- if (! PL_utf8_perl_idstart) {
- invlist = _new_invlist_C_array(_Perl_IDStart_invlist);
- }
- return is_utf8_common_with_len(p, e, &PL_utf8_perl_idstart,
- "_Perl_IDStart", invlist);
+ return is_utf8_common_with_len(p, e, PL_utf8_perl_idstart);
}
bool
if (*p == '_')
return TRUE;
- return is_utf8_common(p, &PL_utf8_xidstart, "XIdStart", NULL);
+ return is_utf8_common(p, PL_utf8_xidstart);
}
bool
Perl__is_utf8_perl_idcont_with_len(pTHX_ const U8 *p, const U8 * const e)
{
- SV* invlist = NULL;
-
PERL_ARGS_ASSERT__IS_UTF8_PERL_IDCONT_WITH_LEN;
- if (! PL_utf8_perl_idcont) {
- invlist = _new_invlist_C_array(_Perl_IDCont_invlist);
- }
- return is_utf8_common_with_len(p, e, &PL_utf8_perl_idcont,
- "_Perl_IDCont", invlist);
+ return is_utf8_common_with_len(p, e, PL_utf8_perl_idcont);
}
bool
{
PERL_ARGS_ASSERT__IS_UTF8_IDCONT;
- return is_utf8_common(p, &PL_utf8_idcont, "IdContinue", NULL);
+ return is_utf8_common(p, PL_utf8_idcont);
}
bool
{
PERL_ARGS_ASSERT__IS_UTF8_XIDCONT;
- return is_utf8_common(p, &PL_utf8_idcont, "XIdContinue", NULL);
+ return is_utf8_common(p, PL_utf8_xidcont);
}
bool
{
PERL_ARGS_ASSERT__IS_UTF8_MARK;
- return is_utf8_common(p, &PL_utf8_mark, "IsM", NULL);
+ return is_utf8_common(p, PL_utf8_mark);
}
- /* change namve uv1 to 'from' */
STATIC UV
-S__to_utf8_case(pTHX_ const UV uv1, const U8 *p, U8* ustrp, STRLEN *lenp,
- SV **swashp, const char *normal, const char *special)
+S__to_utf8_case(pTHX_ const UV uv1, const U8 *p,
+ U8* ustrp, STRLEN *lenp,
+ SV *invlist, const int * const invmap,
+ const unsigned int * const * const aux_tables,
+ const U8 * const aux_table_lengths,
+ const char * const normal)
{
STRLEN len = 0;
+ /* Change the case of code point 'uv1' whose UTF-8 representation (assumed
+ * by this routine to be valid) begins at 'p'. 'normal' is a string to use
+ * to name the new case in any generated messages, as a fallback if the
+ * operation being used is not available. The new case is given by the
+ * data structures in the remaining arguments.
+ *
+ * On return 'ustrp' points to '*lenp' UTF-8 encoded bytes representing the
+ * entire changed case string, and the return value is the first code point
+ * in that string */
+
PERL_ARGS_ASSERT__TO_UTF8_CASE;
/* For code points that don't change case, we already know that the output
* some others */
if (uv1 < 0xFB00) {
goto cases_to_self;
-
}
if (UNLIKELY(UNICODE_IS_SUPER(uv1))) {
- if (UNLIKELY(uv1 > MAX_EXTERNALLY_LEGAL_CP)) {
+ if (UNLIKELY(uv1 > MAX_LEGAL_CP)) {
Perl_croak(aTHX_ cp_above_legal_max, uv1,
- MAX_EXTERNALLY_LEGAL_CP);
+ MAX_LEGAL_CP);
}
if (ckWARN_d(WARN_NON_UNICODE)) {
const char* desc = (PL_op) ? OP_DESC(PL_op) : normal;
}
/* Note that non-characters are perfectly legal, so no warning should
- * be given. There are so few of them, that it isn't worth the extra
- * tests to avoid swash creation */
+ * be given. */
}
- if (!*swashp) /* load on-demand */
- *swashp = _core_swash_init("utf8", normal, &PL_sv_undef,
- 4, 0, NULL, NULL);
+ {
+ unsigned int i;
+ const unsigned int * cp_list;
+ U8 * d;
- if (special) {
- /* It might be "special" (sometimes, but not always,
- * a multicharacter mapping) */
- HV *hv = NULL;
- SV **svp;
+ /* 'index' is guaranteed to be non-negative, as this is an inversion
+ * map that covers all possible inputs. See [perl #133365] */
+ SSize_t index = _invlist_search(invlist, uv1);
+ IV base = invmap[index];
- /* If passed in the specials name, use that; otherwise use any
- * given in the swash */
- if (*special != '\0') {
- hv = get_hv(special, 0);
- }
- else {
- svp = hv_fetchs(MUTABLE_HV(SvRV(*swashp)), "SPECIALS", 0);
- if (svp) {
- hv = MUTABLE_HV(SvRV(*svp));
- }
- }
+ /* The data structures are set up so that if 'base' is non-negative,
+ * the case change is 1-to-1; and if 0, the change is to itself */
+ if (base >= 0) {
+ IV lc;
- if (hv
- && (svp = hv_fetch(hv, (const char*)p, UVCHR_SKIP(uv1), FALSE))
- && (*svp))
- {
- const char *s;
-
- s = SvPV_const(*svp, len);
- if (len == 1)
- /* EIGHTBIT */
- len = uvchr_to_utf8(ustrp, *(U8*)s) - ustrp;
- else {
- Copy(s, ustrp, len, U8);
- }
- }
- }
+ if (base == 0) {
+ goto cases_to_self;
+ }
- if (!len && *swashp) {
- const UV uv2 = swash_fetch(*swashp, p, TRUE /* => is UTF-8 */);
+ /* This computes, e.g. lc(H) as 'H - A + a', using the lc table */
+ lc = base + uv1 - invlist_array(invlist)[index];
+ *lenp = uvchr_to_utf8(ustrp, lc) - ustrp;
+ return lc;
+ }
- if (uv2) {
- /* It was "normal" (a single character mapping). */
- len = uvchr_to_utf8(ustrp, uv2) - ustrp;
- }
- }
+ /* Here 'base' is negative. That means the mapping is 1-to-many, and
+ * requires an auxiliary table look up. abs(base) gives the index into
+ * a list of such tables which points to the proper aux table. And a
+ * parallel list gives the length of each corresponding aux table. */
+ cp_list = aux_tables[-base];
- if (len) {
- if (lenp) {
- *lenp = len;
+ /* Create the string of UTF-8 from the mapped-to code points */
+ d = ustrp;
+ for (i = 0; i < aux_table_lengths[-base]; i++) {
+ d = uvchr_to_utf8(d, cp_list[i]);
}
- return valid_utf8_to_uvchr(ustrp, 0);
+ *d = '\0';
+ *lenp = d - ustrp;
+
+ return cp_list[0];
}
/* Here, there was no mapping defined, which means that the code point maps
* to itself. Return the inputs */
cases_to_self:
- len = UTF8SKIP(p);
- if (p != ustrp) { /* Don't copy onto itself */
- Copy(p, ustrp, len, U8);
+ if (p) {
+ len = UTF8SKIP(p);
+ if (p != ustrp) { /* Don't copy onto itself */
+ Copy(p, ustrp, len, U8);
+ }
+ *lenp = len;
+ }
+ else {
+ *lenp = uvchr_to_utf8(ustrp, uv1) - ustrp;
}
-
- if (lenp)
- *lenp = len;
return uv1;
}
+Size_t
+Perl__inverse_folds(pTHX_ const UV cp, unsigned int * first_folds_to,
+ const unsigned int ** remaining_folds_to)
+{
+ /* Returns the count of the number of code points that fold to the input
+ * 'cp' (besides itself).
+ *
+ * If the return is 0, there is nothing else that folds to it, and
+ * '*first_folds_to' is set to 0, and '*remaining_folds_to' is set to NULL.
+ *
+ * If the return is 1, '*first_folds_to' is set to the single code point,
+ * and '*remaining_folds_to' is set to NULL.
+ *
+ * Otherwise, '*first_folds_to' is set to a code point, and
+ * '*remaining_fold_to' is set to an array that contains the others. The
+ * length of this array is the returned count minus 1.
+ *
+ * The reason for this convolution is to avoid having to deal with
+ * allocating and freeing memory. The lists are already constructed, so
+ * the return can point to them, but single code points aren't, so would
+ * need to be constructed if we didn't employ something like this API */
+
+ /* 'index' is guaranteed to be non-negative, as this is an inversion map
+ * that covers all possible inputs. See [perl #133365] */
+ SSize_t index = _invlist_search(PL_utf8_foldclosures, cp);
+ int base = _Perl_IVCF_invmap[index];
+
+ PERL_ARGS_ASSERT__INVERSE_FOLDS;
+
+ if (base == 0) { /* No fold */
+ *first_folds_to = 0;
+ *remaining_folds_to = NULL;
+ return 0;
+ }
+
+#ifndef HAS_IVCF_AUX_TABLES /* This Unicode version only has 1-1 folds */
+
+ assert(base > 0);
+
+#else
+
+ if (UNLIKELY(base < 0)) { /* Folds to more than one character */
+
+ /* The data structure is set up so that the absolute value of 'base' is
+ * an index into a table of pointers to arrays, with the array
+ * corresponding to the index being the list of code points that fold
+ * to 'cp', and the parallel array containing the length of the list
+ * array */
+ *first_folds_to = IVCF_AUX_TABLE_ptrs[-base][0];
+ *remaining_folds_to = IVCF_AUX_TABLE_ptrs[-base] + 1; /* +1 excludes
+ *first_folds_to
+ */
+ return IVCF_AUX_TABLE_lengths[-base];
+ }
+
+#endif
+
+ /* Only the single code point. This works like 'fc(G) = G - A + a' */
+ *first_folds_to = base + cp - invlist_array(PL_utf8_foldclosures)[index];
+ *remaining_folds_to = NULL;
+ return 1;
+}
+
STATIC UV
S_check_locale_boundary_crossing(pTHX_ const U8* const p, const UV result,
U8* const ustrp, STRLEN *lenp)
assert(UTF8_IS_ABOVE_LATIN1(*p));
/* We know immediately if the first character in the string crosses the
- * boundary, so can skip */
+ * boundary, so can skip testing */
if (result > 255) {
/* Look at every character in the result; if any cross the
if (*e == NULL) {
utf8n_flags = _UTF8_NO_CONFIDENCE_IN_CURLEN;
- *e = p + UTF8SKIP(p);
+
+ /* strnlen() makes this function safe for the common case of
+ * NUL-terminated strings */
+ *e = p + my_strnlen((char *) p, UTF8SKIP(p));
/* For mathoms.c calls, we use the function name we know is stored
* there. It could be part of a larger path */
L1_func_extra_param) \
\
if (flags & (locale_flags)) { \
+ _CHECK_AND_WARN_PROBLEMATIC_LOCALE; \
/* Treat a UTF-8 locale as not being in locale at all */ \
if (IN_UTF8_CTYPE_LOCALE) { \
flags &= ~(locale_flags); \
} \
- else { \
- _CHECK_AND_WARN_PROBLEMATIC_LOCALE; \
- } \
} \
\
if (UTF8_IS_INVARIANT(*p)) { \
} \
} \
else if UTF8_IS_NEXT_CHAR_DOWNGRADEABLE(p, e) { \
+ U8 c = EIGHT_BIT_UTF8_TO_NATIVE(*p, *(p+1)); \
if (flags & (locale_flags)) { \
- result = LC_L1_change_macro(EIGHT_BIT_UTF8_TO_NATIVE(*p, \
- *(p+1))); \
+ result = LC_L1_change_macro(c); \
} \
else { \
- return L1_func(EIGHT_BIT_UTF8_TO_NATIVE(*p, *(p+1)), \
- ustrp, lenp, L1_func_extra_param); \
+ return L1_func(c, ustrp, lenp, L1_func_extra_param); \
} \
} \
else { /* malformed UTF-8 or ord above 255 */ \
/* Special case these two characters, as what normally gets
* returned under locale doesn't work */
- if (memEQs((char *) p, UTF8SKIP(p), CAP_SHARP_S))
+ if (memBEGINs((char *) p, e - p, CAP_SHARP_S))
{
/* diag_listed_as: Can't do %s("%s") on non-UTF-8 locale; resolved to "%s". */
Perl_ck_warner(aTHX_ packWARN(WARN_LOCALE),
}
else
#endif
- if (memEQs((char *) p, UTF8SKIP(p), LONG_S_T))
+ if (memBEGINs((char *) p, e - p, LONG_S_T))
{
/* diag_listed_as: Can't do %s("%s") on non-UTF-8 locale; resolved to "%s". */
Perl_ck_warner(aTHX_ packWARN(WARN_LOCALE),
* 255/256 boundary which is forbidden under /l, and so the code
* wouldn't catch that they are equivalent (which they are only in
* this release) */
- else if (memEQs((char *) p, UTF8SKIP(p), DOTTED_I)) {
+ else if (memBEGINs((char *) p, e - p, DOTTED_I)) {
/* diag_listed_as: Can't do %s("%s") on non-UTF-8 locale; resolved to "%s". */
Perl_ck_warner(aTHX_ packWARN(WARN_LOCALE),
"Can't do fc(\"\\x{0130}\") on non-UTF-8 locale; "
* works. */
*lenp = 2 * sizeof(LATIN_SMALL_LETTER_LONG_S_UTF8) - 2;
- Copy(LATIN_SMALL_LETTER_LONG_S_UTF8 LATIN_SMALL_LETTER_LONG_S_UTF8,
+ Copy(LATIN_SMALL_LETTER_LONG_S_UTF8 LATIN_SMALL_LETTER_LONG_S_UTF8,
ustrp, *lenp, U8);
return LATIN_SMALL_LETTER_LONG_S;
SV* retval = &PL_sv_undef;
HV* swash_hv = NULL;
- const int invlist_swash_boundary =
- (flags_p && *flags_p & _CORE_SWASH_INIT_ACCEPT_INVLIST)
- ? 512 /* Based on some benchmarking, but not extensive, see commit
- message */
- : -1; /* Never return just an inversion list */
+ const bool use_invlist= (flags_p && *flags_p & _CORE_SWASH_INIT_ACCEPT_INVLIST);
assert(listsv != &PL_sv_undef || strNE(name, "") || invlist);
assert(! invlist || minbits == 1);
/* Here, there is no swash already. Set up a minimal one, if
* we are going to return a swash */
- if ((int) _invlist_len(invlist) > invlist_swash_boundary) {
+ if (! use_invlist) {
swash_hv = newHV();
retval = newRV_noinc(MUTABLE_SV(swash_hv));
}
/* Here, we have computed the union of all the passed-in data. It may
* be that there was an inversion list in the swash which didn't get
* touched; otherwise save the computed one */
- if (! invlist_in_swash_is_valid
- && (int) _invlist_len(swash_invlist) > invlist_swash_boundary)
- {
+ if (! invlist_in_swash_is_valid && ! use_invlist) {
if (! hv_stores(MUTABLE_HV(SvRV(retval)), "V", swash_invlist))
{
Perl_croak(aTHX_ "panic: hv_store() unexpectedly failed");
/* The result is immutable. Forbid attempts to change it. */
SvREADONLY_on(swash_invlist);
- /* Use the inversion list stand-alone if small enough */
- if ((int) _invlist_len(swash_invlist) <= invlist_swash_boundary) {
+ if (use_invlist) {
SvREFCNT_dec(retval);
if (!swash_invlist_unclaimed)
SvREFCNT_inc_simple_void_NN(swash_invlist);
return swatch;
}
-HV*
-Perl__swash_inversion_hash(pTHX_ SV* const swash)
-{
-
- /* Subject to change or removal. For use only in regcomp.c and regexec.c
- * Can't be used on a property that is subject to user override, as it
- * relies on the value of SPECIALS in the swash which would be set by
- * utf8_heavy.pl to the hash in the non-overriden file, and hence is not set
- * for overridden properties
- *
- * Returns a hash which is the inversion and closure of a swash mapping.
- * For example, consider the input lines:
- * 004B 006B
- * 004C 006C
- * 212A 006B
- *
- * The returned hash would have two keys, the UTF-8 for 006B and the UTF-8 for
- * 006C. The value for each key is an array. For 006C, the array would
- * have two elements, the UTF-8 for itself, and for 004C. For 006B, there
- * would be three elements in its array, the UTF-8 for 006B, 004B and 212A.
- *
- * Note that there are no elements in the hash for 004B, 004C, 212A. The
- * keys are only code points that are folded-to, so it isn't a full closure.
- *
- * Essentially, for any code point, it gives all the code points that map to
- * it, or the list of 'froms' for that point.
- *
- * Currently it ignores any additions or deletions from other swashes,
- * looking at just the main body of the swash, and if there are SPECIALS
- * in the swash, at that hash
- *
- * The specials hash can be extra code points, and most likely consists of
- * maps from single code points to multiple ones (each expressed as a string
- * of UTF-8 characters). This function currently returns only 1-1 mappings.
- * However consider this possible input in the specials hash:
- * "\xEF\xAC\x85" => "\x{0073}\x{0074}", # U+FB05 => 0073 0074
- * "\xEF\xAC\x86" => "\x{0073}\x{0074}", # U+FB06 => 0073 0074
- *
- * Both FB05 and FB06 map to the same multi-char sequence, which we don't
- * currently handle. But it also means that FB05 and FB06 are equivalent in
- * a 1-1 mapping which we should handle, and this relationship may not be in
- * the main table. Therefore this function examines all the multi-char
- * sequences and adds the 1-1 mappings that come out of that.
- *
- * XXX This function was originally intended to be multipurpose, but its
- * only use is quite likely to remain for constructing the inversion of
- * the CaseFolding (//i) property. If it were more general purpose for
- * regex patterns, it would have to do the FB05/FB06 game for simple folds,
- * because certain folds are prohibited under /iaa and /il. As an example,
- * in Unicode 3.0.1 both U+0130 and U+0131 fold to 'i', and hence are both
- * equivalent under /i. But under /iaa and /il, the folds to 'i' are
- * prohibited, so we would not figure out that they fold to each other.
- * Code could be written to automatically figure this out, similar to the
- * code that does this for multi-character folds, but this is the only case
- * where something like this is ever likely to happen, as all the single
- * char folds to the 0-255 range are now quite settled. Instead there is a
- * little special code that is compiled only for this Unicode version. This
- * is smaller and didn't require much coding time to do. But this makes
- * this routine strongly tied to being used just for CaseFolding. If ever
- * it should be generalized, this would have to be fixed */
-
- U8 *l, *lend;
- STRLEN lcur;
- HV *const hv = MUTABLE_HV(SvRV(swash));
-
- /* The string containing the main body of the table. This will have its
- * assertion fail if the swash has been converted to its inversion list */
- SV** const listsvp = hv_fetchs(hv, "LIST", FALSE);
-
- SV** const typesvp = hv_fetchs(hv, "TYPE", FALSE);
- SV** const bitssvp = hv_fetchs(hv, "BITS", FALSE);
- SV** const nonesvp = hv_fetchs(hv, "NONE", FALSE);
- /*SV** const extssvp = hv_fetchs(hv, "EXTRAS", FALSE);*/
- const U8* const typestr = (U8*)SvPV_nolen(*typesvp);
- const STRLEN bits = SvUV(*bitssvp);
- const STRLEN octets = bits >> 3; /* if bits == 1, then octets == 0 */
- const UV none = SvUV(*nonesvp);
- SV **specials_p = hv_fetchs(hv, "SPECIALS", 0);
-
- HV* ret = newHV();
-
- PERL_ARGS_ASSERT__SWASH_INVERSION_HASH;
-
- /* Must have at least 8 bits to get the mappings */
- if (bits != 8 && bits != 16 && bits != 32) {
- Perl_croak(aTHX_ "panic: swash_inversion_hash doesn't expect bits %"
- UVuf, (UV)bits);
- }
-
- if (specials_p) { /* It might be "special" (sometimes, but not always, a
- mapping to more than one character */
-
- /* Construct an inverse mapping hash for the specials */
- HV * const specials_hv = MUTABLE_HV(SvRV(*specials_p));
- HV * specials_inverse = newHV();
- char *char_from; /* the lhs of the map */
- I32 from_len; /* its byte length */
- char *char_to; /* the rhs of the map */
- I32 to_len; /* its byte length */
- SV *sv_to; /* and in a sv */
- AV* from_list; /* list of things that map to each 'to' */
-
- hv_iterinit(specials_hv);
-
- /* The keys are the characters (in UTF-8) that map to the corresponding
- * UTF-8 string value. Iterate through the list creating the inverse
- * list. */
- while ((sv_to = hv_iternextsv(specials_hv, &char_from, &from_len))) {
- SV** listp;
- if (! SvPOK(sv_to)) {
- Perl_croak(aTHX_ "panic: value returned from hv_iternextsv() "
- "unexpectedly is not a string, flags=%lu",
- (unsigned long)SvFLAGS(sv_to));
- }
- /*DEBUG_U(PerlIO_printf(Perl_debug_log, "Found mapping from %" UVXf ", First char of to is %" UVXf "\n", valid_utf8_to_uvchr((U8*) char_from, 0), valid_utf8_to_uvchr((U8*) SvPVX(sv_to), 0)));*/
-
- /* Each key in the inverse list is a mapped-to value, and the key's
- * hash value is a list of the strings (each in UTF-8) that map to
- * it. Those strings are all one character long */
- if ((listp = hv_fetch(specials_inverse,
- SvPVX(sv_to),
- SvCUR(sv_to), 0)))
- {
- from_list = (AV*) *listp;
- }
- else { /* No entry yet for it: create one */
- from_list = newAV();
- if (! hv_store(specials_inverse,
- SvPVX(sv_to),
- SvCUR(sv_to),
- (SV*) from_list, 0))
- {
- Perl_croak(aTHX_ "panic: hv_store() unexpectedly failed");
- }
- }
-
- /* Here have the list associated with this 'to' (perhaps newly
- * created and empty). Just add to it. Note that we ASSUME that
- * the input is guaranteed to not have duplications, so we don't
- * check for that. Duplications just slow down execution time. */
- av_push(from_list, newSVpvn_utf8(char_from, from_len, TRUE));
- }
-
- /* Here, 'specials_inverse' contains the inverse mapping. Go through
- * it looking for cases like the FB05/FB06 examples above. There would
- * be an entry in the hash like
- * 'st' => [ FB05, FB06 ]
- * In this example we will create two lists that get stored in the
- * returned hash, 'ret':
- * FB05 => [ FB05, FB06 ]
- * FB06 => [ FB05, FB06 ]
- *
- * Note that there is nothing to do if the array only has one element.
- * (In the normal 1-1 case handled below, we don't have to worry about
- * two lists, as everything gets tied to the single list that is
- * generated for the single character 'to'. But here, we are omitting
- * that list, ('st' in the example), so must have multiple lists.) */
- while ((from_list = (AV *) hv_iternextsv(specials_inverse,
- &char_to, &to_len)))
- {
- if (av_tindex_skip_len_mg(from_list) > 0) {
- SSize_t i;
-
- /* We iterate over all combinations of i,j to place each code
- * point on each list */
- for (i = 0; i <= av_tindex_skip_len_mg(from_list); i++) {
- SSize_t j;
- AV* i_list = newAV();
- SV** entryp = av_fetch(from_list, i, FALSE);
- if (entryp == NULL) {
- Perl_croak(aTHX_ "panic: av_fetch() unexpectedly"
- " failed");
- }
- if (hv_fetch(ret, SvPVX(*entryp), SvCUR(*entryp), FALSE)) {
- Perl_croak(aTHX_ "panic: unexpected entry for %s",
- SvPVX(*entryp));
- }
- if (! hv_store(ret, SvPVX(*entryp), SvCUR(*entryp),
- (SV*) i_list, FALSE))
- {
- Perl_croak(aTHX_ "panic: hv_store() unexpectedly failed");
- }
-
- /* For DEBUG_U: UV u = valid_utf8_to_uvchr((U8*) SvPVX(*entryp), 0);*/
- for (j = 0; j <= av_tindex_skip_len_mg(from_list); j++) {
- entryp = av_fetch(from_list, j, FALSE);
- if (entryp == NULL) {
- Perl_croak(aTHX_ "panic: av_fetch() unexpectedly failed");
- }
-
- /* When i==j this adds itself to the list */
- av_push(i_list, newSVuv(utf8_to_uvchr_buf(
- (U8*) SvPVX(*entryp),
- (U8*) SvPVX(*entryp) + SvCUR(*entryp),
- 0)));
- /*DEBUG_U(PerlIO_printf(Perl_debug_log, "%s: %d: Adding %" UVXf " to list for %" UVXf "\n", __FILE__, __LINE__, valid_utf8_to_uvchr((U8*) SvPVX(*entryp), 0), u));*/
- }
- }
- }
- }
- SvREFCNT_dec(specials_inverse); /* done with it */
- } /* End of specials */
-
- /* read $swash->{LIST} */
-
-#if UNICODE_MAJOR_VERSION == 3 \
- && UNICODE_DOT_VERSION == 0 \
- && UNICODE_DOT_DOT_VERSION == 1
-
- /* For this version only U+130 and U+131 are equivalent under qr//i. Add a
- * rule so that things work under /iaa and /il */
-
- SV * mod_listsv = sv_mortalcopy(*listsvp);
- sv_catpv(mod_listsv, "130\t130\t131\n");
- l = (U8*)SvPV(mod_listsv, lcur);
-
-#else
-
- l = (U8*)SvPV(*listsvp, lcur);
-
-#endif
-
- lend = l + lcur;
-
- /* Go through each input line */
- while (l < lend) {
- UV min, max, val;
- UV inverse;
- l = swash_scan_list_line(l, lend, &min, &max, &val,
- cBOOL(octets), typestr);
- if (l > lend) {
- break;
- }
-
- /* Each element in the range is to be inverted */
- for (inverse = min; inverse <= max; inverse++) {
- AV* list;
- SV** listp;
- IV i;
- bool found_key = FALSE;
- bool found_inverse = FALSE;
-
- /* The key is the inverse mapping */
- char key[UTF8_MAXBYTES+1];
- char* key_end = (char *) uvchr_to_utf8((U8*) key, val);
- STRLEN key_len = key_end - key;
-
- /* Get the list for the map */
- if ((listp = hv_fetch(ret, key, key_len, FALSE))) {
- list = (AV*) *listp;
- }
- else { /* No entry yet for it: create one */
- list = newAV();
- if (! hv_store(ret, key, key_len, (SV*) list, FALSE)) {
- Perl_croak(aTHX_ "panic: hv_store() unexpectedly failed");
- }
- }
-
- /* Look through list to see if this inverse mapping already is
- * listed, or if there is a mapping to itself already */
- for (i = 0; i <= av_tindex_skip_len_mg(list); i++) {
- SV** entryp = av_fetch(list, i, FALSE);
- SV* entry;
- UV uv;
- if (entryp == NULL) {
- Perl_croak(aTHX_ "panic: av_fetch() unexpectedly failed");
- }
- entry = *entryp;
- uv = SvUV(entry);
- /*DEBUG_U(PerlIO_printf(Perl_debug_log, "list for %" UVXf " contains %" UVXf "\n", val, uv));*/
- if (uv == val) {
- found_key = TRUE;
- }
- if (uv == inverse) {
- found_inverse = TRUE;
- }
-
- /* No need to continue searching if found everything we are
- * looking for */
- if (found_key && found_inverse) {
- break;
- }
- }
-
- /* Make sure there is a mapping to itself on the list */
- if (! found_key) {
- av_push(list, newSVuv(val));
- /*DEBUG_U(PerlIO_printf(Perl_debug_log, "%s: %d: Adding %" UVXf " to list for %" UVXf "\n", __FILE__, __LINE__, val, val));*/
- }
-
-
- /* Simply add the value to the list */
- if (! found_inverse) {
- av_push(list, newSVuv(inverse));
- /*DEBUG_U(PerlIO_printf(Perl_debug_log, "%s: %d: Adding %" UVXf " to list for %" UVXf "\n", __FILE__, __LINE__, inverse, val));*/
- }
-
- /* swatch_get() increments the value of val for each element in the
- * range. That makes more compact tables possible. You can
- * express the capitalization, for example, of all consecutive
- * letters with a single line: 0061\t007A\t0041 This maps 0061 to
- * 0041, 0062 to 0042, etc. I (khw) have never understood 'none',
- * and it's not documented; it appears to be used only in
- * implementing tr//; I copied the semantics from swatch_get(), just
- * in case */
- if (!none || val < none) {
- ++val;
- }
- }
- }
-
- return ret;
-}
-
SV*
Perl__swash_to_invlist(pTHX_ SV* const swash)
{
/* Get the 0th element, which is needed to setup the inversion list
* */
while (isSPACE(*l)) l++;
+ after_atou = (char *) lend;
if (!grok_atoUV((const char *)l, &element0, &after_atou)) {
Perl_croak(aTHX_ "panic: Expecting a valid 0th element for"
" inversion list");
" elements than available", elements);
}
while (isSPACE(*l)) l++;
+ after_atou = (char *) lend;
if (!grok_atoUV((const char *)l, other_elements_ptr++,
&after_atou))
{
* that effect. However, if the caller knows what
* it's doing, it can pass this flag to indicate that,
* and the assertion is skipped.
- * FOLDEQ_S2_ALREADY_FOLDED Similarly.
+ * FOLDEQ_S2_ALREADY_FOLDED Similar to FOLDEQ_S1_ALREADY_FOLDED, but applies
+ * to s2, and s2 doesn't have to be UTF-8 encoded.
+ * This introduces an asymmetry to save a few branches
+ * in a loop. Currently, this is not a problem, as
+ * never are both inputs pre-folded. Simply call this
+ * function with the pre-folded one as the second
+ * string.
* FOLDEQ_S2_FOLDS_SANE
*/
I32
PERL_ARGS_ASSERT_FOLDEQ_UTF8_FLAGS;
- assert( ! ((flags & (FOLDEQ_UTF8_NOMIX_ASCII | FOLDEQ_LOCALE))
- && (((flags & FOLDEQ_S1_ALREADY_FOLDED)
- && !(flags & FOLDEQ_S1_FOLDS_SANE))
- || ((flags & FOLDEQ_S2_ALREADY_FOLDED)
- && !(flags & FOLDEQ_S2_FOLDS_SANE)))));
+ assert( ! ( (flags & (FOLDEQ_UTF8_NOMIX_ASCII | FOLDEQ_LOCALE))
+ && (( (flags & FOLDEQ_S1_ALREADY_FOLDED)
+ && !(flags & FOLDEQ_S1_FOLDS_SANE))
+ || ( (flags & FOLDEQ_S2_ALREADY_FOLDED)
+ && !(flags & FOLDEQ_S2_FOLDS_SANE)))));
/* The algorithm is to trial the folds without regard to the flags on
* the first line of the above assert(), and then see if the result
* violates them. This means that the inputs can't be pre-folded to a
flags_for_folder |= FOLD_FLAGS_LOCALE;
}
}
+ if (flags & FOLDEQ_UTF8_NOMIX_ASCII) {
+ flags_for_folder |= FOLD_FLAGS_NOMIX_ASCII;
+ }
if (pe1) {
e1 = *(U8**)pe1;
if (n2 == 0) { /* Same for s2 */
if (flags & FOLDEQ_S2_ALREADY_FOLDED) {
- f2 = (U8 *) p2;
- assert(u2);
- n2 = UTF8SKIP(f2);
+
+ /* Point to the already-folded character. But for non-UTF-8
+ * variants, convert to UTF-8 for the algorithm below */
+ if (UTF8_IS_INVARIANT(*p2)) {
+ f2 = (U8 *) p2;
+ n2 = 1;
+ }
+ else if (u2) {
+ f2 = (U8 *) p2;
+ n2 = UTF8SKIP(f2);
+ }
+ else {
+ foldbuf2[0] = UTF8_EIGHT_BIT_HI(*p2);
+ foldbuf2[1] = UTF8_EIGHT_BIT_LO(*p2);
+ f2 = foldbuf2;
+ n2 = 2;
+ }
}
else {
if (isASCII(*p2) && ! (flags & FOLDEQ_LOCALE)) {
{
PERL_ARGS_ASSERT_UVUNI_TO_UTF8;
- return Perl_uvoffuni_to_utf8_flags(aTHX_ d, uv, 0);
+ return uvoffuni_to_utf8_flags(d, uv, 0);
}
/*
}
/*
+=for apidoc utf8_to_uvchr
+
+Returns the native code point of the first character in the string C<s>
+which is assumed to be in UTF-8 encoding; C<retlen> will be set to the
+length, in bytes, of that character.
+
+Some, but not all, UTF-8 malformations are detected, and in fact, some
+malformed input could cause reading beyond the end of the input buffer, which
+is why this function is deprecated. Use L</utf8_to_uvchr_buf> instead.
+
+If C<s> points to one of the detected malformations, and UTF8 warnings are
+enabled, zero is returned and C<*retlen> is set (if C<retlen> isn't
+C<NULL>) to -1. If those warnings are off, the computed value if well-defined (or
+the Unicode REPLACEMENT CHARACTER, if not) is silently returned, and C<*retlen>
+is set (if C<retlen> isn't NULL) so that (S<C<s> + C<*retlen>>) is the
+next possible position in C<s> that could begin a non-malformed character.
+See L</utf8n_to_uvchr> for details on when the REPLACEMENT CHARACTER is returned.
+
+=cut
+*/
+
+UV
+Perl_utf8_to_uvchr(pTHX_ const U8 *s, STRLEN *retlen)
+{
+ PERL_ARGS_ASSERT_UTF8_TO_UVCHR;
+
+ /* This function is unsafe if malformed UTF-8 input is given it, which is
+ * why the function is deprecated. If the first byte of the input
+ * indicates that there are more bytes remaining in the sequence that forms
+ * the character than there are in the input buffer, it can read past the
+ * end. But we can make it safe if the input string happens to be
+ * NUL-terminated, as many strings in Perl are, by refusing to read past a
+ * NUL. A NUL indicates the start of the next character anyway. If the
+ * input isn't NUL-terminated, the function remains unsafe, as it always
+ * has been.
+ *
+ * An initial NUL has to be handled separately, but all ASCIIs can be
+ * handled the same way, speeding up this common case */
+
+ if (UTF8_IS_INVARIANT(*s)) { /* Assumes 's' contains at least 1 byte */
+ return (UV) *s;
+ }
+
+ return utf8_to_uvchr_buf(s,
+ s + my_strnlen((char *) s, UTF8SKIP(s)),
+ retlen);
+}
+
+/*
* ex: set ts=8 sts=4 sw=4 et:
*/
https://perl5.git.perl.org / perl5.git / blobdiff