X-Git-Url: https://perl5.git.perl.org/perl5.git/blobdiff_plain/38f458ffd56c0eb9f5df18cb6693ca326a4b1374..d2c9cb5392e8c58c3bb1935fc3c098737224567c:/utf8.h?ds=sidebyside diff --git a/utf8.h b/utf8.h index dd4d1e1..fb83507 100644 --- a/utf8.h +++ b/utf8.h @@ -144,11 +144,6 @@ EXTCONST unsigned char PL_utf8skip[]; END_EXTERN_C -#if defined(_MSC_VER) && _MSC_VER < 1400 -/* older MSVC versions have a smallish macro buffer */ -#define PERL_SMALL_MACRO_BUFFER -#endif - /* =for apidoc Am|U8|NATIVE_TO_LATIN1|U8 ch @@ -194,13 +189,8 @@ adding no time nor space requirements to the implementation. =cut */ -#ifdef PERL_SMALL_MACRO_BUFFER -# define NATIVE_TO_LATIN1(ch) ((U8)(ch)) -# define LATIN1_TO_NATIVE(ch) ((U8)(ch)) -#else -# define NATIVE_TO_LATIN1(ch) (__ASSERT_(FITS_IN_8_BITS(ch)) ((U8) ((ch) | 0))) -# define LATIN1_TO_NATIVE(ch) (__ASSERT_(FITS_IN_8_BITS(ch)) ((U8) ((ch) | 0))) -#endif +#define NATIVE_TO_LATIN1(ch) (__ASSERT_(FITS_IN_8_BITS(ch)) ((U8) ((ch) | 0))) +#define LATIN1_TO_NATIVE(ch) (__ASSERT_(FITS_IN_8_BITS(ch)) ((U8) ((ch) | 0))) /* I8 is an intermediate version of UTF-8 used only in UTF-EBCDIC. We thus * consider it to be identical to UTF-8 on ASCII platforms. Strictly speaking @@ -208,13 +198,8 @@ adding no time nor space requirements to the implementation. * because they are 8-bit encodings that serve the same purpose in Perl, and * rarely do we need to distinguish them. The term "NATIVE_UTF8" applies to * whichever one is applicable on the current platform */ -#ifdef PERL_SMALL_MACRO_BUFFER -#define NATIVE_UTF8_TO_I8(ch) ((U8) (ch)) -#define I8_TO_NATIVE_UTF8(ch) ((U8) (ch)) -#else #define NATIVE_UTF8_TO_I8(ch) (__ASSERT_(FITS_IN_8_BITS(ch)) ((U8) ((ch) | 0))) #define I8_TO_NATIVE_UTF8(ch) (__ASSERT_(FITS_IN_8_BITS(ch)) ((U8) ((ch) | 0))) -#endif #define UNI_TO_NATIVE(ch) ((UV) ((ch) | 0)) #define NATIVE_TO_UNI(ch) ((UV) ((ch) | 0)) @@ -272,60 +257,7 @@ Perl's extended UTF-8 means we can have start bytes up through FF, though any beginning with FF yields a code point that is too large for 32-bit ASCII platforms. FF signals to use 13 bytes for the encoded character. This breaks the paradigm that the number of leading bits gives how many total bytes there -are in the character. - -=cut -*/ - -/* Is the representation of the Unicode code point 'cp' the same regardless of - * being encoded in UTF-8 or not? */ -#define OFFUNI_IS_INVARIANT(cp) isASCII(cp) - -/* -=for apidoc Am|bool|UVCHR_IS_INVARIANT|UV cp - -Evaluates to 1 if the representation of code point C is the same whether or -not it is encoded in UTF-8; otherwise evaluates to 0. UTF-8 invariant -characters can be copied as-is when converting to/from UTF-8, saving time. -C is Unicode if above 255; otherwise is platform-native. - -=cut - */ - -#define UVCHR_IS_INVARIANT(cp) OFFUNI_IS_INVARIANT(cp) - -/* Misleadingly named: is the UTF8-encoded byte 'c' part of a variant sequence - * in UTF-8? This is the inverse of UTF8_IS_INVARIANT. The |0 makes sure this - * isn't mistakenly called with a ptr argument */ -#define UTF8_IS_CONTINUED(c) (__ASSERT_(FITS_IN_8_BITS(c)) \ - ((U8)((c) | 0)) & UTF_CONTINUATION_MARK) - -/* Is the byte 'c' the first byte of a multi-byte UTF8-8 encoded sequence? - * This doesn't catch invariants (they are single-byte). It also excludes the - * illegal overlong sequences that begin with C0 and C1. The |0 makes sure - * this isn't mistakenly called with a ptr argument */ -#define UTF8_IS_START(c) (__ASSERT_(FITS_IN_8_BITS(c)) \ - ((U8)((c) | 0)) >= 0xc2) - -/* Is the byte 'c' part of a multi-byte UTF8-8 encoded sequence, and not the - * first byte thereof? The |0 makes sure this isn't mistakenly called with a - * ptr argument */ -#define UTF8_IS_CONTINUATION(c) (__ASSERT_(FITS_IN_8_BITS(c)) \ - (((U8)((c) | 0)) & UTF_IS_CONTINUATION_MASK) == UTF_CONTINUATION_MARK) - -/* Is the UTF8-encoded byte 'c' the first byte of a two byte sequence? Use - * UTF8_IS_NEXT_CHAR_DOWNGRADEABLE() instead if the input isn't known to - * be well-formed. Masking with 0xfe allows the low bit to be 0 or 1; thus - * this matches 0xc[23]. The |0 makes sure this isn't mistakenly called with a - * ptr argument */ -#define UTF8_IS_DOWNGRADEABLE_START(c) (__ASSERT_(FITS_IN_8_BITS(c)) \ - (((U8)((c) | 0)) & 0xfe) == 0xc2) - -/* Is the UTF8-encoded byte 'c' the first byte of a sequence of bytes that - * represent a code point > 255? The |0 makes sure this isn't mistakenly - * called with a ptr argument */ -#define UTF8_IS_ABOVE_LATIN1(c) (__ASSERT_(FITS_IN_8_BITS(c)) \ - ((U8)((c) | 0)) >= 0xc4) +are in the character. */ /* This is the number of low-order bits a continuation byte in a UTF-8 encoded * sequence contributes to the specification of the code point. In the bit @@ -363,6 +295,29 @@ C is Unicode if above 255; otherwise is platform-native. * the underlying reason that B0 works here) */ #define UTF_CONTINUATION_MARK (UTF_IS_CONTINUATION_MASK & 0xB0) +/* Is the byte 'c' part of a multi-byte UTF8-8 encoded sequence, and not the + * first byte thereof? */ +#define UTF8_IS_CONTINUATION(c) (__ASSERT_(FITS_IN_8_BITS(c)) \ + (((NATIVE_UTF8_TO_I8(c) & UTF_IS_CONTINUATION_MASK) \ + == UTF_CONTINUATION_MARK))) + +/* Is the representation of the Unicode code point 'cp' the same regardless of + * being encoded in UTF-8 or not? This is a fundamental property of + * UTF-8,EBCDIC */ +#define OFFUNI_IS_INVARIANT(c) (((WIDEST_UTYPE)(c)) < UTF_CONTINUATION_MARK) + +/* +=for apidoc Am|bool|UVCHR_IS_INVARIANT|UV cp + +Evaluates to 1 if the representation of code point C is the same whether or +not it is encoded in UTF-8; otherwise evaluates to 0. UTF-8 invariant +characters can be copied as-is when converting to/from UTF-8, saving time. +C is Unicode if above 255; otherwise is platform-native. + +=cut + */ +#define UVCHR_IS_INVARIANT(cp) (OFFUNI_IS_INVARIANT(NATIVE_TO_UNI(cp))) + /* Internal macro to be used only in this file to aid in constructing other * publicly accessible macros. * The number of bytes required to express this uv in UTF-8, for just those @@ -421,6 +376,39 @@ encoded as UTF-8. C is a native (ASCII or EBCDIC) code point if less than */ #define UVCHR_SKIP(uv) ( UVCHR_IS_INVARIANT(uv) ? 1 : __BASE_UNI_SKIP(uv)) +#define UTF_MIN_START_BYTE \ + ((UTF_CONTINUATION_MARK >> UTF_ACCUMULATION_SHIFT) | UTF_START_MARK(2)) + +/* Is the byte 'c' the first byte of a multi-byte UTF8-8 encoded sequence? + * This excludes invariants (they are single-byte). It also excludes the + * illegal overlong sequences that begin with C0 and C1 on ASCII platforms, and + * C0-C4 I8 start bytes on EBCDIC ones. On EBCDIC E0 can't start a + * non-overlong sequence, so we define a base macro and for those platforms, + * extend it to also exclude E0 */ +#define UTF8_IS_START_base(c) (__ASSERT_(FITS_IN_8_BITS(c)) \ + (NATIVE_UTF8_TO_I8(c) >= UTF_MIN_START_BYTE)) +#ifdef EBCDIC +# define UTF8_IS_START(c) \ + (UTF8_IS_START_base(c) && (c) != I8_TO_NATIVE_UTF8(0xE0)) +#else +# define UTF8_IS_START(c) UTF8_IS_START_base(c) +#endif + +#define UTF_MIN_ABOVE_LATIN1_BYTE \ + ((0x100 >> UTF_ACCUMULATION_SHIFT) | UTF_START_MARK(2)) + +/* Is the UTF8-encoded byte 'c' the first byte of a sequence of bytes that + * represent a code point > 255? */ +#define UTF8_IS_ABOVE_LATIN1(c) (__ASSERT_(FITS_IN_8_BITS(c)) \ + (NATIVE_UTF8_TO_I8(c) >= UTF_MIN_ABOVE_LATIN1_BYTE)) + +/* Is the UTF8-encoded byte 'c' the first byte of a two byte sequence? Use + * UTF8_IS_NEXT_CHAR_DOWNGRADEABLE() instead if the input isn't known to + * be well-formed. */ +#define UTF8_IS_DOWNGRADEABLE_START(c) (__ASSERT_(FITS_IN_8_BITS(c)) \ + inRANGE(NATIVE_UTF8_TO_I8(c), \ + UTF_MIN_START_BYTE, UTF_MIN_ABOVE_LATIN1_BYTE - 1)) + /* The largest code point representable by two UTF-8 bytes on this platform. * As explained in the comments for __COMMON_UNI_SKIP, 32 start bytes with * UTF_ACCUMULATION_SHIFT bits of information each */ @@ -431,18 +419,26 @@ encoded as UTF-8. C is a native (ASCII or EBCDIC) code point if less than * continuation byte */ #define MAX_PORTABLE_UTF8_TWO_BYTE (32 * (1U << 5) - 1) -/* The maximum number of UTF-8 bytes a single Unicode character can - * uppercase/lowercase/fold into. Unicode guarantees that the maximum - * expansion is UTF8_MAX_FOLD_CHAR_EXPAND characters, but any above-Unicode - * code point will fold to itself, so we only have to look at the expansion of - * the maximum Unicode code point. But this number may be less than the space - * occupied by a very large code point under Perl's extended UTF-8. We have to - * make it large enough to fit any single character. (It turns out that ASCII - * and EBCDIC differ in which is larger) */ +/* + +=for apidoc AmnU|STRLEN|UTF8_MAXBYTES_CASE + +The maximum number of UTF-8 bytes a single Unicode character can +uppercase/lowercase/titlecase/fold into. + +=cut + + * Unicode guarantees that the maximum expansion is UTF8_MAX_FOLD_CHAR_EXPAND + * characters, but any above-Unicode code point will fold to itself, so we only + * have to look at the expansion of the maximum Unicode code point. But this + * number may be less than the space occupied by a very large code point under + * Perl's extended UTF-8. We have to make it large enough to fit any single + * character. (It turns out that ASCII and EBCDIC differ in which is larger) + * +=cut +*/ #define UTF8_MAXBYTES_CASE \ - (UTF8_MAXBYTES >= (UTF8_MAX_FOLD_CHAR_EXPAND * OFFUNISKIP(0x10FFFF)) \ - ? UTF8_MAXBYTES \ - : (UTF8_MAX_FOLD_CHAR_EXPAND * OFFUNISKIP(0x10FFFF))) + MAX(UTF8_MAXBYTES, UTF8_MAX_FOLD_CHAR_EXPAND * OFFUNISKIP(0x10FFFF)) /* Rest of these are attributes of Unicode and perl's internals rather than the * encoding, or happen to be the same in both ASCII and EBCDIC (at least at @@ -465,7 +461,7 @@ encoded as UTF-8. C is a native (ASCII or EBCDIC) code point if less than * UTF-8 encoded character that mark it as a start byte and give the number of * bytes that comprise the character. 'len' is the number of bytes in the * multi-byte sequence. */ -#define UTF_START_MARK(len) (((len) > 7) ? 0xFF : (0xFF & (0xFE << (7-(len))))) +#define UTF_START_MARK(len) (((len) > 7) ? 0xFF : ((U8) (0xFE << (7-(len))))) /* Masks out the initial one bits in a start byte, leaving the real data ones. * Doesn't work on an invariant byte. 'len' is the number of bytes in the @@ -535,15 +531,62 @@ encoded as UTF-8. C is a native (ASCII or EBCDIC) code point if less than /* =for apidoc Am|STRLEN|UTF8SKIP|char* s -returns the number of bytes in the UTF-8 encoded character whose first (perhaps -only) byte is pointed to by C. +returns the number of bytes a non-malformed UTF-8 encoded character whose first +(perhaps only) byte is pointed to by C. + +If there is a possibility of malformed input, use instead: + +=over + +=item L> if you know the maximum ending pointer in the +buffer pointed to by C; or + +=item L> if you don't know it. + +=back + +It is better to restructure your code so the end pointer is passed down so that +you know what it actually is at the point of this call, but if that isn't +possible, L> can minimize the chance of accessing beyond the end +of the input buffer. =cut */ #define UTF8SKIP(s) PL_utf8skip[*(const U8*)(s)] + +/* +=for apidoc Am|STRLEN|UTF8_SKIP|char* s +This is a synonym for L> + +=cut +*/ + #define UTF8_SKIP(s) UTF8SKIP(s) /* +=for apidoc Am|STRLEN|UTF8_CHK_SKIP|char* s + +This is a safer version of L>, but still not as safe as +L>. This version doesn't blindly assume that the input +string pointed to by C is well-formed, but verifies that there isn't a NUL +terminating character before the expected end of the next character in C. +The length C returns stops just before any such NUL. + +Perl tends to add NULs, as an insurance policy, after the end of strings in +SV's, so it is likely that using this macro will prevent inadvertent reading +beyond the end of the input buffer, even if it is malformed UTF-8. + +This macro is intended to be used by XS modules where the inputs could be +malformed, and it isn't feasible to restructure to use the safer +L>, for example when interfacing with a C library. + +=cut +*/ + +#define UTF8_CHK_SKIP(s) \ + (s[0] == '\0' ? 1 : MIN(UTF8SKIP(s), \ + my_strnlen((char *) (s), UTF8SKIP(s)))) +/* =for apidoc Am|STRLEN|UTF8_SAFE_SKIP|char* s|char* e returns 0 if S= e>>; otherwise returns the number of bytes in the @@ -588,6 +631,11 @@ with a ptr argument. * above show, doesn't matter as to its implementation */ #define NATIVE_BYTE_IS_INVARIANT(c) UVCHR_IS_INVARIANT(c) +/* Misleadingly named: is the UTF8-encoded byte 'c' part of a variant sequence + * in UTF-8? This is the inverse of UTF8_IS_INVARIANT. */ +#define UTF8_IS_CONTINUED(c) (__ASSERT_(FITS_IN_8_BITS(c)) \ + (! UTF8_IS_INVARIANT(c))) + /* 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. */ @@ -644,30 +692,15 @@ with a ptr argument. * beginning of a utf8 character. Now that foo_utf8() determines that itself, * no need to do it again here */ -#define isIDFIRST_lazy_if(p,UTF) \ - _is_utf8_FOO(_CC_IDFIRST, (const U8 *) p, "isIDFIRST_lazy_if", \ - "isIDFIRST_lazy_if_safe", \ - cBOOL(UTF && ! IN_BYTES), 0, __FILE__,__LINE__) - #define isIDFIRST_lazy_if_safe(p, e, UTF) \ ((IN_BYTES || !UTF) \ ? isIDFIRST(*(p)) \ : isIDFIRST_utf8_safe(p, e)) - -#define isWORDCHAR_lazy_if(p,UTF) \ - _is_utf8_FOO(_CC_IDFIRST, (const U8 *) p, "isWORDCHAR_lazy_if", \ - "isWORDCHAR_lazy_if_safe", \ - cBOOL(UTF && ! IN_BYTES), 0, __FILE__,__LINE__) - #define isWORDCHAR_lazy_if_safe(p, e, UTF) \ ((IN_BYTES || !UTF) \ ? isWORDCHAR(*(p)) \ : isWORDCHAR_utf8_safe((U8 *) p, (U8 *) e)) - -#define isALNUM_lazy_if(p,UTF) \ - _is_utf8_FOO(_CC_IDFIRST, (const U8 *) p, "isALNUM_lazy_if", \ - "isWORDCHAR_lazy_if_safe", \ - cBOOL(UTF && ! IN_BYTES), 0, __FILE__,__LINE__) +#define isALNUM_lazy_if_safe(p, e, UTF) isWORDCHAR_lazy_if_safe(p, e, UTF) #define UTF8_MAXLEN UTF8_MAXBYTES @@ -898,6 +931,10 @@ Evaluates to 0xFFFD, the code point of the Unicode REPLACEMENT CHARACTER #define UNICODE_DISALLOW_NONCHAR 0x0020 #define UNICODE_DISALLOW_SUPER 0x0040 #define UNICODE_DISALLOW_PERL_EXTENDED 0x0080 + +#ifdef PERL_CORE +# define UNICODE_ALLOW_ABOVE_IV_MAX 0x0100 +#endif #define UNICODE_DISALLOW_ABOVE_31_BIT UNICODE_DISALLOW_PERL_EXTENDED #define UNICODE_GOT_SURROGATE UNICODE_DISALLOW_SURROGATE @@ -972,7 +1009,13 @@ Evaluates to 0xFFFD, the code point of the Unicode REPLACEMENT CHARACTER #define UNI_DISPLAY_ISPRINT 0x0001 #define UNI_DISPLAY_BACKSLASH 0x0002 -#define UNI_DISPLAY_QQ (UNI_DISPLAY_ISPRINT|UNI_DISPLAY_BACKSLASH) +#define UNI_DISPLAY_BACKSPACE 0x0004 /* Allow \b when also + UNI_DISPLAY_BACKSLASH */ +#define UNI_DISPLAY_QQ (UNI_DISPLAY_ISPRINT \ + |UNI_DISPLAY_BACKSLASH \ + |UNI_DISPLAY_BACKSPACE) + +/* Character classes could also allow \b, but not patterns in general */ #define UNI_DISPLAY_REGEX (UNI_DISPLAY_ISPRINT|UNI_DISPLAY_BACKSLASH) #define ANYOF_FOLD_SHARP_S(node, input, end) \