static const char unees[] =
"Malformed UTF-8 character (unexpected end of string)";
static const char cp_above_legal_max[] =
- "It is deprecated to use code point 0x%"UVXf"; the permissible max is 0x%"UVXf"";
+ "Use of code point 0x%"UVXf" is deprecated; the permissible max is 0x%"UVXf"";
-#define MAX_NON_DEPRECATED_CP (IV_MAX)
+#define MAX_NON_DEPRECATED_CP ((UV) (IV_MAX))
/*
=head1 Unicode Support
=cut
*/
+#define HANDLE_UNICODE_SURROGATE(uv, flags) \
+ STMT_START { \
+ if (flags & UNICODE_WARN_SURROGATE) { \
+ Perl_ck_warner_d(aTHX_ packWARN(WARN_SURROGATE), \
+ "UTF-16 surrogate U+%04"UVXf, uv); \
+ } \
+ if (flags & UNICODE_DISALLOW_SURROGATE) { \
+ return NULL; \
+ } \
+ } STMT_END;
+
+#define HANDLE_UNICODE_NONCHAR(uv, flags) \
+ STMT_START { \
+ if (flags & UNICODE_WARN_NONCHAR) { \
+ Perl_ck_warner_d(aTHX_ packWARN(WARN_NONCHAR), \
+ "Unicode non-character U+%04"UVXf" is not " \
+ "recommended for open interchange", uv); \
+ } \
+ if (flags & UNICODE_DISALLOW_NONCHAR) { \
+ return NULL; \
+ } \
+ } STMT_END;
+
+/* Use shorter names internally in this file */
+#define SHIFT UTF_ACCUMULATION_SHIFT
+#undef MARK
+#define MARK UTF_CONTINUATION_MARK
+#define MASK UTF_CONTINUATION_MASK
+
U8 *
Perl_uvoffuni_to_utf8_flags(pTHX_ U8 *d, UV uv, UV flags)
{
PERL_ARGS_ASSERT_UVOFFUNI_TO_UTF8_FLAGS;
if (OFFUNI_IS_INVARIANT(uv)) {
- *d++ = (U8) LATIN1_TO_NATIVE(uv);
+ *d++ = LATIN1_TO_NATIVE(uv);
return d;
}
- /* The first problematic code point is the first surrogate */
- if (uv >= UNICODE_SURROGATE_FIRST) {
- if (UNICODE_IS_SURROGATE(uv)) {
- if (flags & UNICODE_WARN_SURROGATE) {
- Perl_ck_warner_d(aTHX_ packWARN(WARN_SURROGATE),
- "UTF-16 surrogate U+%04"UVXf, uv);
- }
- if (flags & UNICODE_DISALLOW_SURROGATE) {
- return NULL;
- }
- }
- else if (UNICODE_IS_SUPER(uv)) {
- if ( UNLIKELY(uv > MAX_NON_DEPRECATED_CP)
- && ckWARN_d(WARN_DEPRECATED))
+ if (uv <= MAX_UTF8_TWO_BYTE) {
+ *d++ = I8_TO_NATIVE_UTF8(( uv >> SHIFT) | UTF_START_MARK(2));
+ *d++ = I8_TO_NATIVE_UTF8(( uv & MASK) | MARK);
+ return d;
+ }
+
+ /* Not 2-byte; test for and handle 3-byte result. In the test immediately
+ * below, the 16 is for start bytes E0-EF (which are all the possible ones
+ * for 3 byte characters). The 2 is for 2 continuation bytes; these each
+ * contribute SHIFT bits. This yields 0x4000 on EBCDIC platforms, 0x1_0000
+ * on ASCII; so 3 bytes covers the range 0x400-0x3FFF on EBCDIC;
+ * 0x800-0xFFFF on ASCII */
+ if (uv < (16 * (1U << (2 * SHIFT)))) {
+ *d++ = I8_TO_NATIVE_UTF8(( uv >> ((3 - 1) * SHIFT)) | UTF_START_MARK(3));
+ *d++ = I8_TO_NATIVE_UTF8(((uv >> ((2 - 1) * SHIFT)) & MASK) | MARK);
+ *d++ = I8_TO_NATIVE_UTF8(( uv /* (1 - 1) */ & MASK) | MARK);
+
+#ifndef EBCDIC /* These problematic code points are 4 bytes on EBCDIC, so
+ aren't tested here */
+ /* The most likely code points in this range are below the surrogates.
+ * Do an extra test to quickly exclude those. */
+ if (UNLIKELY(uv >= UNICODE_SURROGATE_FIRST)) {
+ if (UNLIKELY( UNICODE_IS_32_CONTIGUOUS_NONCHARS(uv)
+ || UNICODE_IS_END_PLANE_NONCHAR_GIVEN_NOT_SUPER(uv)))
{
- Perl_warner(aTHX_ packWARN(WARN_DEPRECATED),
- cp_above_legal_max, uv, MAX_NON_DEPRECATED_CP);
+ HANDLE_UNICODE_NONCHAR(uv, flags);
}
- if ( (flags & UNICODE_WARN_SUPER)
- || (UNICODE_IS_ABOVE_31_BIT(uv) && (flags & UNICODE_WARN_ABOVE_31_BIT)))
- {
- Perl_ck_warner_d(aTHX_ packWARN(WARN_NON_UNICODE),
+ else if (UNLIKELY(UNICODE_IS_SURROGATE(uv))) {
+ HANDLE_UNICODE_SURROGATE(uv, flags);
+ }
+ }
+#endif
+ return d;
+ }
- /* Choose the more dire applicable warning */
- (UNICODE_IS_ABOVE_31_BIT(uv))
- ? "Code point 0x%"UVXf" is not Unicode, and not portable"
- : "Code point 0x%"UVXf" is not Unicode, may not be portable",
- uv);
- }
- if (flags & UNICODE_DISALLOW_SUPER
- || (UNICODE_IS_ABOVE_31_BIT(uv) && (flags & UNICODE_DISALLOW_ABOVE_31_BIT)))
- {
- return NULL;
- }
- }
- else if (UNICODE_IS_NONCHAR(uv)) {
- if (flags & UNICODE_WARN_NONCHAR) {
- Perl_ck_warner_d(aTHX_ packWARN(WARN_NONCHAR),
- "Unicode non-character U+%04"UVXf" is not recommended for open interchange",
- uv);
- }
- if (flags & UNICODE_DISALLOW_NONCHAR) {
- return NULL;
- }
- }
+ /* Not 3-byte; that means the code point is at least 0x1_0000 on ASCII
+ * platforms, and 0x4000 on EBCDIC. There are problematic cases that can
+ * happen starting with 4-byte characters on ASCII platforms. We unify the
+ * code for these with EBCDIC, even though some of them require 5-bytes on
+ * those, because khw believes the code saving is worth the very slight
+ * performance hit on these high EBCDIC code points. */
+
+ if (UNLIKELY(UNICODE_IS_SUPER(uv))) {
+ if ( UNLIKELY(uv > MAX_NON_DEPRECATED_CP)
+ && ckWARN_d(WARN_DEPRECATED))
+ {
+ Perl_warner(aTHX_ packWARN(WARN_DEPRECATED),
+ cp_above_legal_max, uv, MAX_NON_DEPRECATED_CP);
+ }
+ if ( (flags & UNICODE_WARN_SUPER)
+ || ( UNICODE_IS_ABOVE_31_BIT(uv)
+ && (flags & UNICODE_WARN_ABOVE_31_BIT)))
+ {
+ Perl_ck_warner_d(aTHX_ packWARN(WARN_NON_UNICODE),
+
+ /* Choose the more dire applicable warning */
+ (UNICODE_IS_ABOVE_31_BIT(uv))
+ ? "Code point 0x%"UVXf" is not Unicode, and not portable"
+ : "Code point 0x%"UVXf" is not Unicode, may not be portable",
+ uv);
+ }
+ if (flags & UNICODE_DISALLOW_SUPER
+ || ( UNICODE_IS_ABOVE_31_BIT(uv)
+ && (flags & UNICODE_DISALLOW_ABOVE_31_BIT)))
+ {
+ return NULL;
+ }
+ }
+ else if (UNLIKELY(UNICODE_IS_END_PLANE_NONCHAR_GIVEN_NOT_SUPER(uv))) {
+ HANDLE_UNICODE_NONCHAR(uv, flags);
+ }
+
+ /* Test for and handle 4-byte result. In the test immediately below, the
+ * 8 is for start bytes F0-F7 (which are all the possible ones for 4 byte
+ * characters). The 3 is for 3 continuation bytes; these each contribute
+ * SHIFT bits. This yields 0x4_0000 on EBCDIC platforms, 0x20_0000 on
+ * ASCII, so 4 bytes covers the range 0x4000-0x3_FFFF on EBCDIC;
+ * 0x1_0000-0x1F_FFFF on ASCII */
+ if (uv < (8 * (1U << (3 * SHIFT)))) {
+ *d++ = I8_TO_NATIVE_UTF8(( uv >> ((4 - 1) * SHIFT)) | UTF_START_MARK(4));
+ *d++ = I8_TO_NATIVE_UTF8(((uv >> ((3 - 1) * SHIFT)) & MASK) | MARK);
+ *d++ = I8_TO_NATIVE_UTF8(((uv >> ((2 - 1) * SHIFT)) & MASK) | MARK);
+ *d++ = I8_TO_NATIVE_UTF8(( uv /* (1 - 1) */ & MASK) | MARK);
+
+#ifdef EBCDIC /* These were handled on ASCII platforms in the code for 3-byte
+ 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);
+ }
+ else if (UNLIKELY(UNICODE_IS_SURROGATE(uv))) {
+ HANDLE_UNICODE_SURROGATE(uv, flags);
+ }
+#endif
+
+ return d;
}
-#if defined(EBCDIC)
+ /* Not 4-byte; that means the code point is at least 0x20_0000 on ASCII
+ * platforms, and 0x4000 on EBCDIC. At this point we switch to a loop
+ * format. The unrolled version above turns out to not save all that much
+ * time, and at these high code points (well above the legal Unicode range
+ * on ASCII platforms, and well above anything in common use in EBCDIC),
+ * khw believes that less code outweighs slight performance gains. */
+
{
STRLEN len = OFFUNISKIP(uv);
U8 *p = d+len-1;
while (p > d) {
- *p-- = (U8) I8_TO_NATIVE_UTF8((uv & UTF_CONTINUATION_MASK) | UTF_CONTINUATION_MARK);
+ *p-- = I8_TO_NATIVE_UTF8((uv & UTF_CONTINUATION_MASK) | UTF_CONTINUATION_MARK);
uv >>= UTF_ACCUMULATION_SHIFT;
}
- *p = (U8) I8_TO_NATIVE_UTF8((uv & UTF_START_MASK(len)) | UTF_START_MARK(len));
+ *p = I8_TO_NATIVE_UTF8((uv & UTF_START_MASK(len)) | UTF_START_MARK(len));
return d+len;
}
-#else /* Non loop style */
- if (uv < 0x800) {
- *d++ = (U8)(( uv >> 6) | 0xc0);
- *d++ = (U8)(( uv & 0x3f) | 0x80);
- return d;
- }
- if (uv < 0x10000) {
- *d++ = (U8)(( uv >> 12) | 0xe0);
- *d++ = (U8)(((uv >> 6) & 0x3f) | 0x80);
- *d++ = (U8)(( uv & 0x3f) | 0x80);
- return d;
- }
- if (uv < 0x200000) {
- *d++ = (U8)(( uv >> 18) | 0xf0);
- *d++ = (U8)(((uv >> 12) & 0x3f) | 0x80);
- *d++ = (U8)(((uv >> 6) & 0x3f) | 0x80);
- *d++ = (U8)(( uv & 0x3f) | 0x80);
- return d;
- }
- if (uv < 0x4000000) {
- *d++ = (U8)(( uv >> 24) | 0xf8);
- *d++ = (U8)(((uv >> 18) & 0x3f) | 0x80);
- *d++ = (U8)(((uv >> 12) & 0x3f) | 0x80);
- *d++ = (U8)(((uv >> 6) & 0x3f) | 0x80);
- *d++ = (U8)(( uv & 0x3f) | 0x80);
- return d;
- }
- if (uv < 0x80000000) {
- *d++ = (U8)(( uv >> 30) | 0xfc);
- *d++ = (U8)(((uv >> 24) & 0x3f) | 0x80);
- *d++ = (U8)(((uv >> 18) & 0x3f) | 0x80);
- *d++ = (U8)(((uv >> 12) & 0x3f) | 0x80);
- *d++ = (U8)(((uv >> 6) & 0x3f) | 0x80);
- *d++ = (U8)(( uv & 0x3f) | 0x80);
- return d;
- }
-#ifdef UTF8_QUAD_MAX
- if (uv < UTF8_QUAD_MAX)
-#endif
- {
- *d++ = 0xfe; /* Can't match U+FEFF! */
- *d++ = (U8)(((uv >> 30) & 0x3f) | 0x80);
- *d++ = (U8)(((uv >> 24) & 0x3f) | 0x80);
- *d++ = (U8)(((uv >> 18) & 0x3f) | 0x80);
- *d++ = (U8)(((uv >> 12) & 0x3f) | 0x80);
- *d++ = (U8)(((uv >> 6) & 0x3f) | 0x80);
- *d++ = (U8)(( uv & 0x3f) | 0x80);
- return d;
- }
-#ifdef UTF8_QUAD_MAX
- {
- *d++ = 0xff; /* Can't match U+FFFE! */
- *d++ = 0x80; /* 6 Reserved bits */
- *d++ = (U8)(((uv >> 60) & 0x0f) | 0x80); /* 2 Reserved bits */
- *d++ = (U8)(((uv >> 54) & 0x3f) | 0x80);
- *d++ = (U8)(((uv >> 48) & 0x3f) | 0x80);
- *d++ = (U8)(((uv >> 42) & 0x3f) | 0x80);
- *d++ = (U8)(((uv >> 36) & 0x3f) | 0x80);
- *d++ = (U8)(((uv >> 30) & 0x3f) | 0x80);
- *d++ = (U8)(((uv >> 24) & 0x3f) | 0x80);
- *d++ = (U8)(((uv >> 18) & 0x3f) | 0x80);
- *d++ = (U8)(((uv >> 12) & 0x3f) | 0x80);
- *d++ = (U8)(((uv >> 6) & 0x3f) | 0x80);
- *d++ = (U8)(( uv & 0x3f) | 0x80);
- return d;
- }
-#endif
-#endif /* Non loop style */
}
+
/*
=for apidoc uvchr_to_utf8
follows:
If C<uv> is a Unicode surrogate code point and C<UNICODE_WARN_SURROGATE> is set,
-the function will raise a warning, provided UTF8 warnings are enabled. If instead
-C<UNICODE_DISALLOW_SURROGATE> is set, the function will fail and return NULL.
-If both flags are set, the function will both warn and return NULL.
+the function will raise a warning, provided UTF8 warnings are enabled. If
+instead C<UNICODE_DISALLOW_SURROGATE> is set, the function will fail and return
+NULL. If both flags are set, the function will both warn and return NULL.
Similarly, the C<UNICODE_WARN_NONCHAR> and C<UNICODE_DISALLOW_NONCHAR> flags
affect how the function handles a Unicode non-character.
* 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 */
-#define CALL_UPPER_CASE(INP, OUTP, LENP) Perl_to_utf8_case(aTHX_ INP, OUTP, LENP, &PL_utf8_toupper, "ToUc", "")
-#define CALL_TITLE_CASE(INP, OUTP, LENP) Perl_to_utf8_case(aTHX_ INP, OUTP, LENP, &PL_utf8_totitle, "ToTc", "")
-#define CALL_LOWER_CASE(INP, OUTP, LENP) Perl_to_utf8_case(aTHX_ INP, OUTP, LENP, &PL_utf8_tolower, "ToLc", "")
+#define CALL_UPPER_CASE(uv, s, d, lenp) _to_utf8_case(uv, s, d, lenp, &PL_utf8_toupper, "ToUc", "")
+#define CALL_TITLE_CASE(uv, s, d, lenp) _to_utf8_case(uv, s, d, lenp, &PL_utf8_totitle, "ToTc", "")
+#define CALL_LOWER_CASE(uv, s, d, lenp) _to_utf8_case(uv, s, d, lenp, &PL_utf8_tolower, "ToLc", "")
-/* 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
+/* 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(INP, OUTP, LENP, SPECIALS) Perl_to_utf8_case(aTHX_ INP, OUTP, LENP, &PL_utf8_tofold, "ToCf", (SPECIALS) ? "" : NULL)
+#define CALL_FOLD_CASE(uv, s, d, lenp, specials) _to_utf8_case(uv, s, d, lenp, &PL_utf8_tofold, "ToCf", (specials) ? "" : NULL)
UV
Perl_to_uni_upper(pTHX_ UV c, U8* p, STRLEN *lenp)
}
uvchr_to_utf8(p, c);
- return CALL_UPPER_CASE(p, p, lenp);
+ return CALL_UPPER_CASE(c, p, p, lenp);
}
UV
}
uvchr_to_utf8(p, c);
- return CALL_TITLE_CASE(p, p, lenp);
+ return CALL_TITLE_CASE(c, p, p, lenp);
}
STATIC U8
}
uvchr_to_utf8(p, c);
- return CALL_LOWER_CASE(p, p, lenp);
+ return CALL_LOWER_CASE(c, p, p, lenp);
}
UV
assert (! (flags & FOLD_FLAGS_LOCALE));
- if (c == MICRO_SIGN) {
+ if (UNLIKELY(c == MICRO_SIGN)) {
converted = GREEK_SMALL_LETTER_MU;
}
#if UNICODE_MAJOR_VERSION > 3 /* no multifolds in early Unicode */ \
|| (UNICODE_MAJOR_VERSION == 3 && ( UNICODE_DOT_VERSION > 0) \
|| UNICODE_DOT_DOT_VERSION > 0)
- else if ((flags & FOLD_FLAGS_FULL) && c == LATIN_SMALL_LETTER_SHARP_S) {
-
+ else if ( (flags & FOLD_FLAGS_FULL)
+ && UNLIKELY(c == LATIN_SMALL_LETTER_SHARP_S))
+ {
/* If can't cross 127/128 boundary, can't return "ss"; instead return
* two U+017F characters, as fc("\df") should eq fc("\x{17f}\x{17f}")
* under those circumstances. */
/* 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(p, p, lenp, flags & FOLD_FLAGS_FULL);
+ return CALL_FOLD_CASE(c, p, p, lenp, flags & FOLD_FLAGS_FULL);
}
else { /* Otherwise, _to_utf8_fold_flags has the intelligence to deal with
the special flags. */
/*
=for apidoc to_utf8_case
+Instead use the appropriate one of L</toUPPER_utf8>,
+L</toTITLE_utf8>,
+L</toLOWER_utf8>,
+or L</toFOLD_utf8>.
+
C<p> contains the pointer to the UTF-8 string encoding
the character that is being converted. This routine assumes that the character
at C<p> is well-formed.
Perl_to_utf8_case(pTHX_ const U8 *p, U8* ustrp, STRLEN *lenp,
SV **swashp, const char *normal, const char *special)
{
+ PERL_ARGS_ASSERT_TO_UTF8_CASE;
+
+ return _to_utf8_case(valid_utf8_to_uvchr(p, NULL), p, ustrp, lenp, swashp, normal, special);
+}
+
+ /* 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)
+{
STRLEN len = 0;
- const UV uv1 = valid_utf8_to_uvchr(p, NULL);
- PERL_ARGS_ASSERT_TO_UTF8_CASE;
+ PERL_ARGS_ASSERT__TO_UTF8_CASE;
+
+ /* For code points that don't change case, we already know that the output
+ * of this function is the unchanged input, so we can skip doing look-ups
+ * for them. Unfortunately the case-changing code points are scattered
+ * around. But there are some long consecutive ranges where there are no
+ * case changing code points. By adding tests, we can eliminate the lookup
+ * for all the ones in such ranges. This is currently done here only for
+ * just a few cases where the scripts are in common use in modern commerce
+ * (and scripts adjacent to those which can be included without additional
+ * tests). */
+
+ if (uv1 >= 0x0590) {
+ /* This keeps from needing further processing the code points most
+ * likely to be used in the following non-cased scripts: Hebrew,
+ * Arabic, Syriac, Thaana, NKo, Samaritan, Mandaic, Devanagari,
+ * Bengali, Gurmukhi, Gujarati, Oriya, Tamil, Telugu, Kannada,
+ * Malayalam, Sinhala, Thai, Lao, Tibetan, Myanmar */
+ if (uv1 < 0x10A0) {
+ goto cases_to_self;
+ }
- /* Note that swash_fetch() doesn't output warnings for these because it
- * assumes we will */
- if (uv1 >= UNICODE_SURROGATE_FIRST) {
- if (uv1 <= UNICODE_SURROGATE_LAST) {
- if (ckWARN_d(WARN_SURROGATE)) {
- const char* desc = (PL_op) ? OP_DESC(PL_op) : normal;
- Perl_warner(aTHX_ packWARN(WARN_SURROGATE),
- "Operation \"%s\" returns its argument for UTF-16 surrogate U+%04"UVXf"", desc, uv1);
- }
- }
- else if (UNICODE_IS_SUPER(uv1)) {
- if ( UNLIKELY(uv1 > MAX_NON_DEPRECATED_CP)
- && ckWARN_d(WARN_DEPRECATED))
- {
- Perl_warner(aTHX_ packWARN(WARN_DEPRECATED),
- cp_above_legal_max, uv1, MAX_NON_DEPRECATED_CP);
+ /* The following largish code point ranges also don't have case
+ * changes, but khw didn't think they warranted extra tests to speed
+ * them up (which would slightly slow down everything else above them):
+ * 1100..139F Hangul Jamo, Ethiopic
+ * 1400..1CFF Unified Canadian Aboriginal Syllabics, Ogham, Runic,
+ * Tagalog, Hanunoo, Buhid, Tagbanwa, Khmer, Mongolian,
+ * Limbu, Tai Le, New Tai Lue, Buginese, Tai Tham,
+ * Combining Diacritical Marks Extended, Balinese,
+ * Sundanese, Batak, Lepcha, Ol Chiki
+ * 2000..206F General Punctuation
+ */
+
+ if (uv1 >= 0x2D30) {
+
+ /* This keeps the from needing further processing the code points
+ * most likely to be used in the following non-cased major scripts:
+ * CJK, Katakana, Hiragana, plus some less-likely scripts.
+ *
+ * (0x2D30 above might have to be changed to 2F00 in the unlikely
+ * event that Unicode eventually allocates the unused block as of
+ * v8.0 2FE0..2FEF to code points that are cased. khw has verified
+ * that the test suite will start having failures to alert you
+ * should that happen) */
+ if (uv1 < 0xA640) {
+ goto cases_to_self;
}
- if (ckWARN_d(WARN_NON_UNICODE)) {
- const char* desc = (PL_op) ? OP_DESC(PL_op) : normal;
- Perl_warner(aTHX_ packWARN(WARN_NON_UNICODE),
- "Operation \"%s\" returns its argument for non-Unicode code point 0x%04"UVXf"", desc, uv1);
- }
- }
+
+ if (uv1 >= 0xAC00) {
+ if (UNLIKELY(UNICODE_IS_SURROGATE(uv1))) {
+ if (ckWARN_d(WARN_SURROGATE)) {
+ const char* desc = (PL_op) ? OP_DESC(PL_op) : normal;
+ Perl_warner(aTHX_ packWARN(WARN_SURROGATE),
+ "Operation \"%s\" returns its argument for UTF-16 surrogate U+%04"UVXf"", desc, uv1);
+ }
+ goto cases_to_self;
+ }
+
+ /* AC00..FAFF Catches Hangul syllables and private use, plus
+ * some others */
+ if (uv1 < 0xFB00) {
+ goto cases_to_self;
+
+ }
+
+ if (UNLIKELY(UNICODE_IS_SUPER(uv1))) {
+ if ( UNLIKELY(uv1 > MAX_NON_DEPRECATED_CP)
+ && ckWARN_d(WARN_DEPRECATED))
+ {
+ Perl_warner(aTHX_ packWARN(WARN_DEPRECATED),
+ cp_above_legal_max, uv1, MAX_NON_DEPRECATED_CP);
+ }
+ if (ckWARN_d(WARN_NON_UNICODE)) {
+ const char* desc = (PL_op) ? OP_DESC(PL_op) : normal;
+ Perl_warner(aTHX_ packWARN(WARN_NON_UNICODE),
+ "Operation \"%s\" returns its argument for non-Unicode code point 0x%04"UVXf"", desc, uv1);
+ }
+ goto cases_to_self;
+ }
+#ifdef HIGHEST_CASE_CHANGING_CP_FOR_USE_ONLY_BY_UTF8_DOT_C
+ if (UNLIKELY(uv1
+ > HIGHEST_CASE_CHANGING_CP_FOR_USE_ONLY_BY_UTF8_DOT_C))
+ {
+
+ /* As of this writing, this means we avoid swash creation
+ * for anything beyond low Plane 1 */
+ goto cases_to_self;
+ }
+#endif
+ }
+ }
/* Note that non-characters are perfectly legal, so no warning should
- * be given */
+ * be given. There are so few of them, that it isn't worth the extra
+ * tests to avoid swash creation */
}
if (!*swashp) /* load on-demand */
/* 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);
}
}
else { /* UTF-8, ord above 255 */
- result = CALL_UPPER_CASE(p, ustrp, lenp);
+ result = CALL_UPPER_CASE(valid_utf8_to_uvchr(p, NULL), p, ustrp, lenp);
if (flags) {
result = check_locale_boundary_crossing(p, result, ustrp, lenp);
}
}
else { /* UTF-8, ord above 255 */
- result = CALL_TITLE_CASE(p, ustrp, lenp);
+ result = CALL_TITLE_CASE(valid_utf8_to_uvchr(p, NULL), p, ustrp, lenp);
if (flags) {
result = check_locale_boundary_crossing(p, result, ustrp, lenp);
}
}
else { /* UTF-8, ord above 255 */
- result = CALL_LOWER_CASE(p, ustrp, lenp);
+ result = CALL_LOWER_CASE(valid_utf8_to_uvchr(p, NULL), p, ustrp, lenp);
if (flags) {
result = check_locale_boundary_crossing(p, result, ustrp, lenp);
}
}
else { /* UTF-8, ord above 255 */
- result = CALL_FOLD_CASE(p, ustrp, lenp, flags & FOLD_FLAGS_FULL);
+ result = CALL_FOLD_CASE(valid_utf8_to_uvchr(p, NULL), p, ustrp, lenp, flags & FOLD_FLAGS_FULL);
if (flags & FOLD_FLAGS_LOCALE) {
* (see lib/unicore/SpecCase.txt) The SWASHGET in lib/utf8_heavy.pl is
* the lower-level routine, and it is similarly broken for returning
* multiple values. --jhi
- * For those, you should use to_utf8_case() instead */
+ * For those, you should use S__to_utf8_case() instead */
/* Now SWASHGET is recasted into S_swatch_get in this file. */
/* Note:
* 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
+ * 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
)) {
/* A side effect of this function will be to warn */
(void) utf8n_to_uvchr(s, e - s, &char_len, UTF8_WARN_SUPER);
- ok = FALSE;
- }
+ ok = FALSE;
+ }
}
else if (UTF8_IS_SURROGATE(s, e)) {
if (ckWARN_d(WARN_SURROGATE)) {
https://perl5.git.perl.org / perl5.git / blobdiff