X-Git-Url: https://perl5.git.perl.org/perl5.git/blobdiff_plain/a9f50d33e5a32910a09035ea5927706c74be0d64..2376ab1f7929a7a9f33c6ed28ac0f7c3361894c2:/utf8.h diff --git a/utf8.h b/utf8.h index 7dcea6f..2357fb0 100644 --- a/utf8.h +++ b/utf8.h @@ -43,33 +43,25 @@ #define uvchr_to_utf8_flags(d,uv,flags) \ uvoffuni_to_utf8_flags(d,NATIVE_TO_UNI(uv),flags) #define utf8_to_uvchr_buf(s, e, lenp) \ - utf8n_to_uvchr(s, (e) - (s), lenp, \ + utf8n_to_uvchr(s, (U8*)(e) - (U8*)(s), lenp, \ ckWARN_d(WARN_UTF8) ? 0 : UTF8_ALLOW_ANY) #define to_uni_fold(c, p, lenp) _to_uni_fold_flags(c, p, lenp, FOLD_FLAGS_FULL) -#define to_utf8_fold(c, p, lenp) _to_utf8_fold_flags(c, p, lenp, \ - FOLD_FLAGS_FULL, NULL) -#define to_utf8_lower(a,b,c) _to_utf8_lower_flags(a,b,c,0, NULL) -#define to_utf8_upper(a,b,c) _to_utf8_upper_flags(a,b,c,0, NULL) -#define to_utf8_title(a,b,c) _to_utf8_title_flags(a,b,c,0, NULL) +#define to_utf8_fold(c, p, lenp) _to_utf8_fold_flags(c, p, lenp, FOLD_FLAGS_FULL) +#define to_utf8_lower(a,b,c) _to_utf8_lower_flags(a,b,c,0) +#define to_utf8_upper(a,b,c) _to_utf8_upper_flags(a,b,c,0) +#define to_utf8_title(a,b,c) _to_utf8_title_flags(a,b,c,0) /* Source backward compatibility. */ #define is_utf8_string_loc(s, len, ep) is_utf8_string_loclen(s, len, ep, 0) #define foldEQ_utf8(s1, pe1, l1, u1, s2, pe2, l2, u2) \ foldEQ_utf8_flags(s1, pe1, l1, u1, s2, pe2, l2, u2, 0) -#define FOLDEQ_UTF8_NOMIX_ASCII (1 << 0) -#define FOLDEQ_UTF8_LOCALE (1 << 1) +#define FOLDEQ_UTF8_NOMIX_ASCII (1 << 0) +#define FOLDEQ_LOCALE (1 << 1) #define FOLDEQ_S1_ALREADY_FOLDED (1 << 2) #define FOLDEQ_S2_ALREADY_FOLDED (1 << 3) -/* -=for apidoc ibcmp_utf8 - -This is a synonym for (! foldEQ_utf8()) - -=cut -*/ #define ibcmp_utf8(s1, pe1, l1, u1, s2, pe2, l2, u2) \ cBOOL(! foldEQ_utf8(s1, pe1, l1, u1, s2, pe2, l2, u2)) @@ -233,7 +225,9 @@ Perl's extended UTF-8 means we can have start bytes up to FF. (((UV) UTF_CONTINUATION_MASK) << ((sizeof(UV) * CHARBITS) \ - UTF_ACCUMULATION_SHIFT)) -#ifdef HAS_QUAD +#if UVSIZE >= 8 +# define UTF8_QUAD_MAX UINT64_C(0x1000000000) + /* Input is a true Unicode (not-native) code point */ #define OFFUNISKIP(uv) ( (uv) < 0x80 ? 1 : \ (uv) < 0x800 ? 2 : \ @@ -269,6 +263,10 @@ Perl's extended UTF-8 means we can have start bytes up to FF. #error UTF8_MAXBYTES must be at least 12 #endif +/* ^? is defined to be DEL on ASCII systems. See the definition of toCTRL() + * for more */ +#define QUESTION_MARK_CTRL DEL_NATIVE + #define MAX_UTF8_TWO_BYTE 0x7FF #define UTF8_MAXBYTES_CASE UTF8_MAXBYTES @@ -338,33 +336,57 @@ Perl's extended UTF-8 means we can have start bytes up to FF. * UTF-8 encoded string) */ #define UTF8_IS_INVARIANT(c) UNI_IS_INVARIANT(NATIVE_UTF8_TO_I8(c)) -#define NATIVE_IS_INVARIANT(c) UNI_IS_INVARIANT(NATIVE_TO_LATIN1(c)) +/* Like the above, but its name implies a non-UTF8 input, and is implemented + * differently (for no particular reason) */ +#define NATIVE_BYTE_IS_INVARIANT(c) UNI_IS_INVARIANT(NATIVE_TO_LATIN1(c)) + +/* Like the above, but accepts any UV as input */ +#define UVCHR_IS_INVARIANT(uv) UNI_IS_INVARIANT(NATIVE_TO_UNI(uv)) #define MAX_PORTABLE_UTF8_TWO_BYTE 0x3FF /* constrained by EBCDIC */ -/* The macros in the next sets are used to generate the two utf8 or utfebcdic - * bytes from an ordinal that is known to fit into two bytes; it must be less - * than 0x3FF to work across both encodings. */ -/* Nocast allows these to be used in the case label of a switch statement; - * however this doesn't won't work for ebcdic, and should be avoided. Use - * regen/unicode_constants instead */ -#define UTF8_TWO_BYTE_HI_nocast(c) I8_TO_NATIVE_UTF8((NATIVE_TO_UNI(c) \ - >> UTF_ACCUMULATION_SHIFT) | UTF_START_MARK(2)) -#define UTF8_TWO_BYTE_LO_nocast(c) I8_TO_NATIVE_UTF8((NATIVE_TO_UNI(c) \ - & UTF_CONTINUATION_MASK) \ - | UTF_CONTINUATION_MARK) - -#define UTF8_TWO_BYTE_HI(c) ((U8) (UTF8_TWO_BYTE_HI_nocast(c))) -#define UTF8_TWO_BYTE_LO(c) ((U8) (UTF8_TWO_BYTE_LO_nocast(c))) - -/* This name is used when the source is a single byte (input not checked). - * These expand identically to the TWO_BYTE versions on ASCII platforms, but - * use to/from LATIN1 instead of UNI, which on EBCDIC eliminates tests */ -#define UTF8_EIGHT_BIT_HI(c) I8_TO_NATIVE_UTF8((NATIVE_TO_LATIN1(c) \ - >> UTF_ACCUMULATION_SHIFT) | UTF_START_MARK(2)) -#define UTF8_EIGHT_BIT_LO(c) I8_TO_NATIVE_UTF8((NATIVE_TO_LATIN1(c) \ - & UTF_CONTINUATION_MASK) \ - | UTF_CONTINUATION_MARK) +/* The macros in the next 4 sets are used to generate the two utf8 or utfebcdic + * bytes from an ordinal that is known to fit into exactly two (not one) bytes; + * it must be less than 0x3FF to work across both encodings. */ + +/* These two are helper macros for the other three sets, and should not be used + * directly anywhere else. 'translate_function' is either NATIVE_TO_LATIN1 + * (which works for code points up to 0xFF) or NATIVE_TO_UNI which works for any + * code point */ +#define __BASE_TWO_BYTE_HI(c, translate_function) \ + I8_TO_NATIVE_UTF8((translate_function(c) >> UTF_ACCUMULATION_SHIFT) \ + | UTF_START_MARK(2)) +#define __BASE_TWO_BYTE_LO(c, translate_function) \ + I8_TO_NATIVE_UTF8((translate_function(c) & UTF_CONTINUATION_MASK) \ + | UTF_CONTINUATION_MARK) + +/* The next two macros should not be used. They were designed to be usable as + * the case label of a switch statement, but this doesn't work for EBCDIC. Use + * regen/unicode_constants.pl instead */ +#define UTF8_TWO_BYTE_HI_nocast(c) __BASE_TWO_BYTE_HI(c, NATIVE_TO_UNI) +#define UTF8_TWO_BYTE_LO_nocast(c) __BASE_TWO_BYTE_LO(c, NATIVE_TO_UNI) + +/* The next two macros are used when the source should be a single byte + * character; checked for under DEBUGGING */ +#define UTF8_EIGHT_BIT_HI(c) (__ASSERT_(FITS_IN_8_BITS(c)) \ + ((U8) __BASE_TWO_BYTE_HI(c, NATIVE_TO_LATIN1))) +#define UTF8_EIGHT_BIT_LO(c) (__ASSERT_(FITS_IN_8_BITS(c)) \ + ((U8) __BASE_TWO_BYTE_LO(c, NATIVE_TO_LATIN1))) + +/* These final two macros in the series are used when the source can be any + * code point whose UTF-8 is known to occupy 2 bytes; they are less efficient + * than the EIGHT_BIT versions on EBCDIC platforms. We use the logical '~' + * operator instead of "<=" to avoid getting compiler warnings. + * MAX_PORTABLE_UTF8_TWO_BYTE should be exactly all one bits in the lower few + * places, so the ~ works */ +#define UTF8_TWO_BYTE_HI(c) \ + (__ASSERT_((sizeof(c) == 1) \ + || !(((WIDEST_UTYPE)(c)) & ~MAX_PORTABLE_UTF8_TWO_BYTE)) \ + ((U8) __BASE_TWO_BYTE_HI(c, NATIVE_TO_LATIN1))) +#define UTF8_TWO_BYTE_LO(c) \ + (__ASSERT_((sizeof(c) == 1) \ + || !(((WIDEST_UTYPE)(c)) & ~MAX_PORTABLE_UTF8_TWO_BYTE)) \ + ((U8) __BASE_TWO_BYTE_LO(c, NATIVE_TO_LATIN1))) /* This is illegal in any well-formed UTF-8 in both EBCDIC and ASCII * as it is only in overlongs. */ @@ -393,8 +415,11 @@ Perl's extended UTF-8 means we can have start bytes up to FF. #define IN_BYTES (CopHINTS_get(PL_curcop) & HINT_BYTES) #define DO_UTF8(sv) (SvUTF8(sv) && !IN_BYTES) #define IN_UNI_8_BIT \ - (CopHINTS_get(PL_curcop) & (HINT_UNI_8_BIT|HINT_LOCALE_NOT_CHARS) \ - && ! IN_LOCALE_RUNTIME && ! IN_BYTES) + (((CopHINTS_get(PL_curcop) & (HINT_UNI_8_BIT)) \ + || (CopHINTS_get(PL_curcop) & HINT_LOCALE_PARTIAL \ + /* -1 below is for :not_characters */ \ + && _is_in_locale_category(FALSE, -1))) \ + && ! IN_BYTES) #define UTF8_ALLOW_EMPTY 0x0001 /* Allow a zero length string */ @@ -421,7 +446,9 @@ Perl's extended UTF-8 means we can have start bytes up to FF. #define UTF8_WARN_SUPER 0x0400 /* points above the legal max */ /* Code points which never were part of the original UTF-8 standard, the first - * byte of which is a FE or FF on ASCII platforms. */ + * byte of which is a FE or FF on ASCII platforms. If the first byte is FF, it + * will overflow a 32-bit word. If the first byte is FE, it will overflow a + * signed 32-bit word. */ #define UTF8_DISALLOW_FE_FF 0x0800 #define UTF8_WARN_FE_FF 0x1000 @@ -516,10 +543,6 @@ Perl's extended UTF-8 means we can have start bytes up to FF. #define UNICODE_IS_SUPER(c) ((c) > PERL_UNICODE_MAX) #define UNICODE_IS_FE_FF(c) ((c) > 0x7FFFFFFF) -#ifdef HAS_QUAD -# define UTF8_QUAD_MAX UINT64_C(0x1000000000) -#endif - #define LATIN_SMALL_LETTER_SHARP_S LATIN_SMALL_LETTER_SHARP_S_NATIVE #define LATIN_SMALL_LETTER_Y_WITH_DIAERESIS \ LATIN_SMALL_LETTER_Y_WITH_DIAERESIS_NATIVE @@ -532,14 +555,15 @@ Perl's extended UTF-8 means we can have start bytes up to FF. #define UNICODE_GREEK_SMALL_LETTER_FINAL_SIGMA 0x03C2 #define UNICODE_GREEK_SMALL_LETTER_SIGMA 0x03C3 #define GREEK_SMALL_LETTER_MU 0x03BC -#define GREEK_CAPITAL_LETTER_MU 0x039C /* Upper and title case of MICRON */ -#define LATIN_CAPITAL_LETTER_Y_WITH_DIAERESIS 0x0178 /* Also is title case */ -#define LATIN_CAPITAL_LETTER_SHARP_S 0x1E9E -#define LATIN_SMALL_LETTER_LONG_S 0x017F +#define GREEK_CAPITAL_LETTER_MU 0x039C /* Upper and title case + of MICRON */ +#define LATIN_CAPITAL_LETTER_Y_WITH_DIAERESIS 0x0178 /* Also is title case */ +#define LATIN_CAPITAL_LETTER_SHARP_S 0x1E9E +#define LATIN_SMALL_LETTER_LONG_S 0x017F #define LATIN_SMALL_LIGATURE_LONG_S_T 0xFB05 #define LATIN_SMALL_LIGATURE_ST 0xFB06 -#define KELVIN_SIGN 0x212A -#define ANGSTROM_SIGN 0x212B +#define KELVIN_SIGN 0x212A +#define ANGSTROM_SIGN 0x212B #define UNI_DISPLAY_ISPRINT 0x0001 #define UNI_DISPLAY_BACKSLASH 0x0002 @@ -553,64 +577,83 @@ Perl's extended UTF-8 means we can have start bytes up to FF. ((end) > (input) + 1) && \ toFOLD((input)[0]) == 's' && \ toFOLD((input)[1]) == 's') + #define SHARP_S_SKIP 2 /* If you want to exclude surrogates, and beyond legal Unicode, see the blame * log for earlier versions which gave details for these */ +/* A helper macro for isUTF8_CHAR, so use that one, and not this one. This is + * retained solely for backwards compatibility and may be deprecated and + * removed in a future Perl version. + * + * regen/regcharclass.pl generates is_UTF8_CHAR_utf8() macros for up to these + * number of bytes. So this has to be coordinated with that file */ +#ifdef EBCDIC +# define IS_UTF8_CHAR_FAST(n) ((n) <= 3) +#else +# define IS_UTF8_CHAR_FAST(n) ((n) <= 4) +#endif + #ifndef EBCDIC -/* This was generated by regen/regcharclass.pl, and then moved here. The lines - * that generated it were then commented out. This was done solely because it - * takes on the order of 10 minutes to generate, and is never going to change. - * The EBCDIC equivalent hasn't been commented out in regcharclass.pl, so it - * should generate and run the correct stuff */ +/* A helper macro for isUTF8_CHAR, so use that one instead of this. This was + * generated by regen/regcharclass.pl, and then moved here. The lines that + * generated it were then commented out. This was done solely because it takes + * on the order of 10 minutes to generate, and is never going to change, unless + * the generated code is improved. + * + * The EBCDIC versions have been cut to not cover all of legal Unicode, so + * don't take too long to generate, and there is a separate one for each code + * page, so they are in regcharclass.h instead of here */ /* - UTF8_CHAR: Matches utf8 from 1 to 4 bytes + UTF8_CHAR: Matches legal UTF-8 encoded characters from 2 through 4 bytes - 0x0 - 0x1FFFFF + 0x80 - 0x1FFFFF */ /*** GENERATED CODE ***/ -#define is_UTF8_CHAR_utf8_safe(s,e) \ -( ((e)-(s) > 3) ? \ - ( ( ( ((U8*)s)[0] & 0x80 ) == 0x00 ) ? 1 \ - : ( 0xC2 <= ((U8*)s)[0] && ((U8*)s)[0] <= 0xDF ) ? \ - ( ( ( ((U8*)s)[1] & 0xC0 ) == 0x80 ) ? 2 : 0 ) \ - : ( 0xE0 == ((U8*)s)[0] ) ? \ - ( ( ( ( ((U8*)s)[1] & 0xE0 ) == 0xA0 ) && ( ( ((U8*)s)[2] & 0xC0 ) == 0x80 ) ) ? 3 : 0 )\ - : ( 0xE1 <= ((U8*)s)[0] && ((U8*)s)[0] <= 0xEF ) ? \ - ( ( ( ( ((U8*)s)[1] & 0xC0 ) == 0x80 ) && ( ( ((U8*)s)[2] & 0xC0 ) == 0x80 ) ) ? 3 : 0 )\ - : ( 0xF0 == ((U8*)s)[0] ) ? \ - ( ( ( ( 0x90 <= ((U8*)s)[1] && ((U8*)s)[1] <= 0xBF ) && ( ( ((U8*)s)[2] & 0xC0 ) == 0x80 ) ) && ( ( ((U8*)s)[3] & 0xC0 ) == 0x80 ) ) ? 4 : 0 )\ - : ( ( ( ( 0xF1 <= ((U8*)s)[0] && ((U8*)s)[0] <= 0xF7 ) && ( ( ((U8*)s)[1] & 0xC0 ) == 0x80 ) ) && ( ( ((U8*)s)[2] & 0xC0 ) == 0x80 ) ) && ( ( ((U8*)s)[3] & 0xC0 ) == 0x80 ) ) ? 4 : 0 )\ -: ((e)-(s) > 2) ? \ - ( ( ( ((U8*)s)[0] & 0x80 ) == 0x00 ) ? 1 \ - : ( 0xC2 <= ((U8*)s)[0] && ((U8*)s)[0] <= 0xDF ) ? \ - ( ( ( ((U8*)s)[1] & 0xC0 ) == 0x80 ) ? 2 : 0 ) \ - : ( 0xE0 == ((U8*)s)[0] ) ? \ - ( ( ( ( ((U8*)s)[1] & 0xE0 ) == 0xA0 ) && ( ( ((U8*)s)[2] & 0xC0 ) == 0x80 ) ) ? 3 : 0 )\ - : ( ( ( 0xE1 <= ((U8*)s)[0] && ((U8*)s)[0] <= 0xEF ) && ( ( ((U8*)s)[1] & 0xC0 ) == 0x80 ) ) && ( ( ((U8*)s)[2] & 0xC0 ) == 0x80 ) ) ? 3 : 0 )\ -: ((e)-(s) > 1) ? \ - ( ( ( ((U8*)s)[0] & 0x80 ) == 0x00 ) ? 1 \ - : ( ( 0xC2 <= ((U8*)s)[0] && ((U8*)s)[0] <= 0xDF ) && ( ( ((U8*)s)[1] & 0xC0 ) == 0x80 ) ) ? 2 : 0 )\ -: ((e)-(s) > 0) ? \ - ( ( ((U8*)s)[0] & 0x80 ) == 0x00 ) \ -: 0 ) +#define is_UTF8_CHAR_utf8_no_length_checks(s) \ +( ( 0xC2 <= ((U8*)s)[0] && ((U8*)s)[0] <= 0xDF ) ? \ + ( ( ( ((U8*)s)[1] & 0xC0 ) == 0x80 ) ? 2 : 0 ) \ +: ( 0xE0 == ((U8*)s)[0] ) ? \ + ( ( ( ( ((U8*)s)[1] & 0xE0 ) == 0xA0 ) && ( ( ((U8*)s)[2] & 0xC0 ) == 0x80 ) ) ? 3 : 0 )\ +: ( 0xE1 <= ((U8*)s)[0] && ((U8*)s)[0] <= 0xEF ) ? \ + ( ( ( ( ((U8*)s)[1] & 0xC0 ) == 0x80 ) && ( ( ((U8*)s)[2] & 0xC0 ) == 0x80 ) ) ? 3 : 0 )\ +: ( 0xF0 == ((U8*)s)[0] ) ? \ + ( ( ( ( 0x90 <= ((U8*)s)[1] && ((U8*)s)[1] <= 0xBF ) && ( ( ((U8*)s)[2] & 0xC0 ) == 0x80 ) ) && ( ( ((U8*)s)[3] & 0xC0 ) == 0x80 ) ) ? 4 : 0 )\ +: ( ( ( ( 0xF1 <= ((U8*)s)[0] && ((U8*)s)[0] <= 0xF7 ) && ( ( ((U8*)s)[1] & 0xC0 ) == 0x80 ) ) && ( ( ((U8*)s)[2] & 0xC0 ) == 0x80 ) ) && ( ( ((U8*)s)[3] & 0xC0 ) == 0x80 ) ) ? 4 : 0 ) #endif -/* IS_UTF8_CHAR(p) is strictly speaking wrong (not UTF-8) because it - * (1) allows UTF-8 encoded UTF-16 surrogates - * (2) it allows code points past U+10FFFF. - * The Perl_is_utf8_char() full "slow" code will handle the Perl - * "extended UTF-8". */ -#define IS_UTF8_CHAR(p, n) (is_UTF8_CHAR_utf8_safe(p, (p) + (n)) == n) +/* +=head1 Unicode Support -/* regen/regcharclass.pl generates is_UTF8_CHAR_utf8_safe() macros for up to - * these number of bytes. So this has to be coordinated with it */ -#ifdef EBCDIC -# define IS_UTF8_CHAR_FAST(n) ((n) <= 5) -#else -# define IS_UTF8_CHAR_FAST(n) ((n) <= 4) -#endif +=for apidoc Am|STRLEN|isUTF8_CHAR|const U8 *s|const U8 *e + +Returns the number of bytes beginning at C which form a legal UTF-8 (or +UTF-EBCDIC) encoded character, looking no further than C bytes into +C. Returns 0 if the sequence starting at C through C is not +well-formed UTF-8 + +Note that an INVARIANT character (i.e. ASCII on non-EBCDIC +machines) is a valid UTF-8 character. + +=cut +*/ + +#define isUTF8_CHAR(s, e) (UNLIKELY((e) <= (s)) \ + ? 0 \ + : (UTF8_IS_INVARIANT(*s)) \ + ? 1 \ + : UNLIKELY(((e) - (s)) < UTF8SKIP(s)) \ + ? 0 \ + : LIKELY(IS_UTF8_CHAR_FAST(UTF8SKIP(s))) \ + ? is_UTF8_CHAR_utf8_no_length_checks(s) \ + : _is_utf8_char_slow(s, e)) + +#define is_utf8_char_buf(buf, buf_end) isUTF8_CHAR(buf, buf_end) + +/* Do not use; should be deprecated. Use isUTF8_CHAR() instead; this is + * retained solely for backwards compatibility */ +#define IS_UTF8_CHAR(p, n) (isUTF8_CHAR(p, (p) + (n)) == n) #endif /* H_UTF8 */