3 * This file contains definitions for use with the UTF-8 encoding. It
4 * actually also works with the variant UTF-8 encoding called UTF-EBCDIC, and
5 * hides almost all of the differences between these from the caller. In other
6 * words, someone should #include this file, and if the code is being compiled
7 * on an EBCDIC platform, things should mostly just work.
9 * Copyright (C) 2000, 2001, 2002, 2005, 2006, 2007, 2009,
10 * 2010, 2011 by Larry Wall and others
12 * You may distribute under the terms of either the GNU General Public
13 * License or the Artistic License, as specified in the README file.
17 #ifndef H_UTF8 /* Guard against recursive inclusion */
20 /* Use UTF-8 as the default script encoding?
21 * Turning this on will break scripts having non-UTF-8 binary
22 * data (such as Latin-1) in string literals. */
23 #ifdef USE_UTF8_SCRIPTS
24 # define USE_UTF8_IN_NAMES (!IN_BYTES)
26 # define USE_UTF8_IN_NAMES (PL_hints & HINT_UTF8)
29 #include "regcharclass.h"
30 #include "unicode_constants.h"
32 /* For to_utf8_fold_flags, q.v. */
33 #define FOLD_FLAGS_LOCALE 0x1
34 #define FOLD_FLAGS_FULL 0x2
35 #define FOLD_FLAGS_NOMIX_ASCII 0x4
37 /* For _core_swash_init(), internal core use only */
38 #define _CORE_SWASH_INIT_USER_DEFINED_PROPERTY 0x1
39 #define _CORE_SWASH_INIT_RETURN_IF_UNDEF 0x2
40 #define _CORE_SWASH_INIT_ACCEPT_INVLIST 0x4
43 =head1 Unicode Support
45 =for apidoc is_ascii_string
47 This is a misleadingly-named synonym for L</is_invariant_string>.
48 On ASCII-ish platforms, the name isn't misleading: the ASCII-range characters
49 are exactly the UTF-8 invariants. But EBCDIC machines have more invariants
50 than just the ASCII characters, so C<is_invariant_string> is preferred.
54 #define is_ascii_string(s, len) is_invariant_string(s, len)
56 #define uvchr_to_utf8(a,b) uvchr_to_utf8_flags(a,b,0)
57 #define uvchr_to_utf8_flags(d,uv,flags) \
58 uvoffuni_to_utf8_flags(d,NATIVE_TO_UNI(uv),flags)
59 #define utf8_to_uvchr_buf(s, e, lenp) \
60 utf8n_to_uvchr(s, (U8*)(e) - (U8*)(s), lenp, \
61 ckWARN_d(WARN_UTF8) ? 0 : UTF8_ALLOW_ANY)
63 #define to_uni_fold(c, p, lenp) _to_uni_fold_flags(c, p, lenp, FOLD_FLAGS_FULL)
64 #define to_utf8_fold(c, p, lenp) _to_utf8_fold_flags(c, p, lenp, FOLD_FLAGS_FULL)
65 #define to_utf8_lower(a,b,c) _to_utf8_lower_flags(a,b,c,0)
66 #define to_utf8_upper(a,b,c) _to_utf8_upper_flags(a,b,c,0)
67 #define to_utf8_title(a,b,c) _to_utf8_title_flags(a,b,c,0)
69 /* Source backward compatibility. */
70 #define is_utf8_string_loc(s, len, ep) is_utf8_string_loclen(s, len, ep, 0)
72 #define foldEQ_utf8(s1, pe1, l1, u1, s2, pe2, l2, u2) \
73 foldEQ_utf8_flags(s1, pe1, l1, u1, s2, pe2, l2, u2, 0)
74 #define FOLDEQ_UTF8_NOMIX_ASCII (1 << 0)
75 #define FOLDEQ_LOCALE (1 << 1)
76 #define FOLDEQ_S1_ALREADY_FOLDED (1 << 2)
77 #define FOLDEQ_S2_ALREADY_FOLDED (1 << 3)
78 #define FOLDEQ_S1_FOLDS_SANE (1 << 4)
79 #define FOLDEQ_S2_FOLDS_SANE (1 << 5)
81 #define ibcmp_utf8(s1, pe1, l1, u1, s2, pe2, l2, u2) \
82 cBOOL(! foldEQ_utf8(s1, pe1, l1, u1, s2, pe2, l2, u2))
85 /* The equivalent of these macros but implementing UTF-EBCDIC
86 are in the following header file:
89 #include "utfebcdic.h"
95 EXTCONST unsigned char PL_utf8skip[] = {
96 /* 0x00 */ 1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1, /* ascii */
97 /* 0x10 */ 1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1, /* ascii */
98 /* 0x20 */ 1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1, /* ascii */
99 /* 0x30 */ 1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1, /* ascii */
100 /* 0x40 */ 1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1, /* ascii */
101 /* 0x50 */ 1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1, /* ascii */
102 /* 0x60 */ 1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1, /* ascii */
103 /* 0x70 */ 1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1, /* ascii */
104 /* 0x80 */ 1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1, /* bogus: continuation byte */
105 /* 0x90 */ 1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1, /* bogus: continuation byte */
106 /* 0xA0 */ 1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1, /* bogus: continuation byte */
107 /* 0xB0 */ 1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1, /* bogus: continuation byte */
108 /* 0xC0 */ 2,2, /* overlong */
109 /* 0xC2 */ 2,2,2,2,2,2,2,2,2,2,2,2,2,2, /* U+0080 to U+03FF */
110 /* 0xD0 */ 2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2, /* U+0400 to U+07FF */
111 /* 0xE0 */ 3,3,3,3,3,3,3,3,3,3,3,3,3,3,3,3, /* U+0800 to U+FFFF */
112 /* 0xF0 */ 4,4,4,4,4,4,4,4,5,5,5,5,6,6, /* above BMP to 2**31 - 1 */
113 /* 0xFE */ 7,13, /* Perl extended (never was official UTF-8). Up to 72bit
114 allowed (64-bit + reserved). */
117 EXTCONST unsigned char PL_utf8skip[];
122 /* Native character to/from iso-8859-1. Are the identity functions on ASCII
124 #define NATIVE_TO_LATIN1(ch) (ch)
125 #define LATIN1_TO_NATIVE(ch) (ch)
127 /* I8 is an intermediate version of UTF-8 used only in UTF-EBCDIC. We thus
128 * consider it to be identical to UTF-8 on ASCII platforms. Strictly speaking
129 * UTF-8 and UTF-EBCDIC are two different things, but we often conflate them
130 * because they are 8-bit encodings that serve the same purpose in Perl, and
131 * rarely do we need to distinguish them. The term "NATIVE_UTF8" applies to
132 * whichever one is applicable on the current platform */
133 #define NATIVE_UTF8_TO_I8(ch) (ch)
134 #define I8_TO_NATIVE_UTF8(ch) (ch)
136 /* Transforms in wide UV chars */
137 #define UNI_TO_NATIVE(ch) (ch)
138 #define NATIVE_TO_UNI(ch) (ch)
142 The following table is from Unicode 3.2.
144 Code Points 1st Byte 2nd Byte 3rd Byte 4th Byte
146 U+0000..U+007F 00..7F
147 U+0080..U+07FF * C2..DF 80..BF
148 U+0800..U+0FFF E0 * A0..BF 80..BF
149 U+1000..U+CFFF E1..EC 80..BF 80..BF
150 U+D000..U+D7FF ED 80..9F 80..BF
151 U+D800..U+DFFF ED A0..BF 80..BF (surrogates)
152 U+E000..U+FFFF EE..EF 80..BF 80..BF
153 U+10000..U+3FFFF F0 * 90..BF 80..BF 80..BF
154 U+40000..U+FFFFF F1..F3 80..BF 80..BF 80..BF
155 U+100000..U+10FFFF F4 80..8F 80..BF 80..BF
156 Below are non-Unicode code points
157 U+110000..U+13FFFF F4 90..BF 80..BF 80..BF
158 U+110000..U+1FFFFF F5..F7 80..BF 80..BF 80..BF
159 U+200000..: F8.. * 88..BF 80..BF 80..BF 80..BF
161 Note the gaps before several of the byte entries above marked by '*'. These are
162 caused by legal UTF-8 avoiding non-shortest encodings: it is technically
163 possible to UTF-8-encode a single code point in different ways, but that is
164 explicitly forbidden, and the shortest possible encoding should always be used
165 (and that is what Perl does). The non-shortest ones are called 'overlongs'.
170 Another way to look at it, as bits:
172 Code Points 1st Byte 2nd Byte 3rd Byte 4th Byte
175 0000 0bbb bbaa aaaa 110b bbbb 10aa aaaa
176 cccc bbbb bbaa aaaa 1110 cccc 10bb bbbb 10aa aaaa
177 00 000d ddcc cccc bbbb bbaa aaaa 1111 0ddd 10cc cccc 10bb bbbb 10aa aaaa
179 As you can see, the continuation bytes all begin with C<10>, and the
180 leading bits of the start byte tell how many bytes there are in the
183 Perl's extended UTF-8 means we can have start bytes up to FF.
187 /* Is the representation of the Unicode code point 'c' the same regardless of
188 * being encoded in UTF-8 or not? */
189 #define UNI_IS_INVARIANT(c) (((UV)c) < 0x80)
191 /* Is the UTF8-encoded byte 'c' part of a variant sequence in UTF-8? This is
192 * the inverse of UTF8_IS_INVARIANT */
193 #define UTF8_IS_CONTINUED(c) (((U8)c) & 0x80)
195 /* Is the byte 'c' the first byte of a multi-byte UTF8-8 encoded sequence?
196 * This doesn't catch invariants (they are single-byte). It also excludes the
197 * illegal overlong sequences that begin with C0 and C1. */
198 #define UTF8_IS_START(c) (((U8)c) >= 0xc2)
200 /* Is the byte 'c' part of a multi-byte UTF8-8 encoded sequence, and not the
201 * first byte thereof? */
202 #define UTF8_IS_CONTINUATION(c) ((((U8)c) & 0xC0) == 0x80)
204 /* Is the UTF8-encoded byte 'c' the first byte of a two byte sequence? Use
205 * UTF8_IS_NEXT_CHAR_DOWNGRADEABLE() instead if the input isn't known to
206 * be well-formed. Masking with 0xfe allows the low bit to be 0 or 1; thus
207 * this matches 0xc[23]. */
208 #define UTF8_IS_DOWNGRADEABLE_START(c) (((U8)(c) & 0xfe) == 0xc2)
210 /* Is the UTF8-encoded byte 'c' the first byte of a sequence of bytes that
211 * represent a code point > 255? */
212 #define UTF8_IS_ABOVE_LATIN1(c) ((U8)(c) >= 0xc4)
214 /* This defines the 1-bits that are to be in the first byte of a multi-byte
215 * UTF-8 encoded character that give the number of bytes that comprise the
216 * character. 'len' is the number of bytes in the multi-byte sequence. */
217 #define UTF_START_MARK(len) (((len) > 7) ? 0xFF : (0xFF & (0xFE << (7-(len)))))
219 /* Masks out the initial one bits in a start byte, leaving the real data ones.
220 * Doesn't work on an invariant byte. 'len' is the number of bytes in the
221 * multi-byte sequence that comprises the character. */
222 #define UTF_START_MASK(len) (((len) >= 7) ? 0x00 : (0x1F >> ((len)-2)))
224 /* This defines the bits that are to be in the continuation bytes of a multi-byte
225 * UTF-8 encoded character that indicate it is a continuation byte. */
226 #define UTF_CONTINUATION_MARK 0x80
228 /* This is the number of low-order bits a continuation byte in a UTF-8 encoded
229 * sequence contributes to the specification of the code point. In the bit
230 * maps above, you see that the first 2 bits are a constant '10', leaving 6 of
231 * real information */
232 #define UTF_ACCUMULATION_SHIFT 6
234 /* 2**UTF_ACCUMULATION_SHIFT - 1 */
235 #define UTF_CONTINUATION_MASK ((U8)0x3f)
237 /* If a value is anded with this, and the result is non-zero, then using the
238 * original value in UTF8_ACCUMULATE will overflow, shifting bits off the left
240 #define UTF_ACCUMULATION_OVERFLOW_MASK \
241 (((UV) UTF_CONTINUATION_MASK) << ((sizeof(UV) * CHARBITS) \
242 - UTF_ACCUMULATION_SHIFT))
245 # define UTF8_QUAD_MAX UINT64_C(0x1000000000)
247 /* Input is a true Unicode (not-native) code point */
248 #define OFFUNISKIP(uv) ( (uv) < 0x80 ? 1 : \
250 (uv) < 0x10000 ? 3 : \
251 (uv) < 0x200000 ? 4 : \
252 (uv) < 0x4000000 ? 5 : \
253 (uv) < 0x80000000 ? 6 : \
254 (uv) < UTF8_QUAD_MAX ? 7 : 13 )
256 /* No, I'm not even going to *TRY* putting #ifdef inside a #define */
257 #define OFFUNISKIP(uv) ( (uv) < 0x80 ? 1 : \
259 (uv) < 0x10000 ? 3 : \
260 (uv) < 0x200000 ? 4 : \
261 (uv) < 0x4000000 ? 5 : \
262 (uv) < 0x80000000 ? 6 : 7 )
265 /* How wide can a single UTF-8 encoded character become in bytes. */
266 /* NOTE: Strictly speaking Perl's UTF-8 should not be called UTF-8 since UTF-8
267 * is an encoding of Unicode, and Unicode's upper limit, 0x10FFFF, can be
268 * expressed with 4 bytes. However, Perl thinks of UTF-8 as a way to encode
269 * non-negative integers in a binary format, even those above Unicode */
270 #define UTF8_MAXBYTES 13
272 /* The maximum number of UTF-8 bytes a single Unicode character can
273 * uppercase/lowercase/fold into. Unicode guarantees that the maximum
274 * expansion is 3 characters. On ASCIIish platforms, the highest Unicode
275 * character occupies 4 bytes, therefore this number would be 12, but this is
276 * smaller than the maximum width a single above-Unicode character can occupy,
277 * so use that instead */
278 #if UTF8_MAXBYTES < 12
279 #error UTF8_MAXBYTES must be at least 12
282 /* ^? is defined to be DEL on ASCII systems. See the definition of toCTRL()
284 #define QUESTION_MARK_CTRL DEL_NATIVE
286 #define MAX_UTF8_TWO_BYTE 0x7FF
288 #define UTF8_MAXBYTES_CASE UTF8_MAXBYTES
290 #endif /* EBCDIC vs ASCII */
292 /* Rest of these are attributes of Unicode and perl's internals rather than the
293 * encoding, or happen to be the same in both ASCII and EBCDIC (at least at
294 * this level; the macros that some of these call may have different
295 * definitions in the two encodings */
297 /* In domain restricted to ASCII, these may make more sense to the reader than
298 * the ones with Latin1 in the name */
299 #define NATIVE_TO_ASCII(ch) NATIVE_TO_LATIN1(ch)
300 #define ASCII_TO_NATIVE(ch) LATIN1_TO_NATIVE(ch)
302 /* More or less misleadingly-named defines, retained for back compat */
303 #define NATIVE_TO_UTF(ch) NATIVE_UTF8_TO_I8(ch)
304 #define NATIVE_TO_I8(ch) NATIVE_UTF8_TO_I8(ch)
305 #define UTF_TO_NATIVE(ch) I8_TO_NATIVE_UTF8(ch)
306 #define I8_TO_NATIVE(ch) I8_TO_NATIVE_UTF8(ch)
307 #define NATIVE8_TO_UNI(ch) NATIVE_TO_LATIN1(ch)
309 /* Adds a UTF8 continuation byte 'new' of information to a running total code
310 * point 'old' of all the continuation bytes so far. This is designed to be
311 * used in a loop to convert from UTF-8 to the code point represented. Note
312 * that this is asymmetric on EBCDIC platforms, in that the 'new' parameter is
313 * the UTF-EBCDIC byte, whereas the 'old' parameter is a Unicode (not EBCDIC)
314 * code point in process of being generated */
315 #define UTF8_ACCUMULATE(old, new) (((old) << UTF_ACCUMULATION_SHIFT) \
316 | ((NATIVE_UTF8_TO_I8((U8)new)) \
317 & UTF_CONTINUATION_MASK))
319 /* This works in the face of malformed UTF-8. */
320 #define UTF8_IS_NEXT_CHAR_DOWNGRADEABLE(s, e) (UTF8_IS_DOWNGRADEABLE_START(*s) \
321 && ( (e) - (s) > 1) \
322 && UTF8_IS_CONTINUATION(*((s)+1)))
324 /* Number of bytes a code point occupies in UTF-8. */
325 #define NATIVE_SKIP(uv) OFFUNISKIP(NATIVE_TO_UNI(uv))
327 /* Most code which says UNISKIP is really thinking in terms of native code
328 * points (0-255) plus all those beyond. This is an imprecise term, but having
329 * it means existing code continues to work. For precision, use NATIVE_SKIP
331 #define UNISKIP(uv) NATIVE_SKIP(uv)
333 /* Convert a two (not one) byte utf8 character to a native code point value.
334 * Needs just one iteration of accumulate. Should not be used unless it is
335 * known that the two bytes are legal: 1) two-byte start, and 2) continuation.
336 * Note that the result can be larger than 255 if the input character is not
338 #define TWO_BYTE_UTF8_TO_NATIVE(HI, LO) \
339 UNI_TO_NATIVE(UTF8_ACCUMULATE((NATIVE_UTF8_TO_I8(HI) & UTF_START_MASK(2)), \
342 /* Should never be used, and be deprecated */
343 #define TWO_BYTE_UTF8_TO_UNI(HI, LO) NATIVE_TO_UNI(TWO_BYTE_UTF8_TO_NATIVE(HI, LO))
345 /* How many bytes in the UTF-8 encoded character whose first (perhaps only)
346 * byte is pointed to by 's' */
347 #define UTF8SKIP(s) PL_utf8skip[*(const U8*)(s)]
349 /* Is the byte 'c' the same character when encoded in UTF-8 as when not. This
350 * works on both UTF-8 encoded strings and non-encoded, as it returns TRUE in
351 * each for the exact same set of bit patterns. (And it works on any byte in a
352 * UTF-8 encoded string) */
353 #define UTF8_IS_INVARIANT(c) UNI_IS_INVARIANT(NATIVE_UTF8_TO_I8(c))
355 /* Like the above, but its name implies a non-UTF8 input, and is implemented
356 * differently (for no particular reason) */
357 #define NATIVE_BYTE_IS_INVARIANT(c) UNI_IS_INVARIANT(NATIVE_TO_LATIN1(c))
359 /* Like the above, but accepts any UV as input */
360 #define UVCHR_IS_INVARIANT(uv) UNI_IS_INVARIANT(NATIVE_TO_UNI(uv))
362 #define MAX_PORTABLE_UTF8_TWO_BYTE 0x3FF /* constrained by EBCDIC */
364 /* The macros in the next 4 sets are used to generate the two utf8 or utfebcdic
365 * bytes from an ordinal that is known to fit into exactly two (not one) bytes;
366 * it must be less than 0x3FF to work across both encodings. */
368 /* These two are helper macros for the other three sets, and should not be used
369 * directly anywhere else. 'translate_function' is either NATIVE_TO_LATIN1
370 * (which works for code points up to 0xFF) or NATIVE_TO_UNI which works for any
372 #define __BASE_TWO_BYTE_HI(c, translate_function) \
373 I8_TO_NATIVE_UTF8((translate_function(c) >> UTF_ACCUMULATION_SHIFT) \
375 #define __BASE_TWO_BYTE_LO(c, translate_function) \
376 I8_TO_NATIVE_UTF8((translate_function(c) & UTF_CONTINUATION_MASK) \
377 | UTF_CONTINUATION_MARK)
379 /* The next two macros should not be used. They were designed to be usable as
380 * the case label of a switch statement, but this doesn't work for EBCDIC. Use
381 * regen/unicode_constants.pl instead */
382 #define UTF8_TWO_BYTE_HI_nocast(c) __BASE_TWO_BYTE_HI(c, NATIVE_TO_UNI)
383 #define UTF8_TWO_BYTE_LO_nocast(c) __BASE_TWO_BYTE_LO(c, NATIVE_TO_UNI)
385 /* The next two macros are used when the source should be a single byte
386 * character; checked for under DEBUGGING */
387 #define UTF8_EIGHT_BIT_HI(c) (__ASSERT_(FITS_IN_8_BITS(c)) \
388 ((U8) __BASE_TWO_BYTE_HI(c, NATIVE_TO_LATIN1)))
389 #define UTF8_EIGHT_BIT_LO(c) (__ASSERT_(FITS_IN_8_BITS(c)) \
390 ((U8) __BASE_TWO_BYTE_LO(c, NATIVE_TO_LATIN1)))
392 /* These final two macros in the series are used when the source can be any
393 * code point whose UTF-8 is known to occupy 2 bytes; they are less efficient
394 * than the EIGHT_BIT versions on EBCDIC platforms. We use the logical '~'
395 * operator instead of "<=" to avoid getting compiler warnings.
396 * MAX_UTF8_TWO_BYTE should be exactly all one bits in the lower few
397 * places, so the ~ works */
398 #define UTF8_TWO_BYTE_HI(c) \
399 (__ASSERT_((sizeof(c) == 1) \
400 || !(((WIDEST_UTYPE)(c)) & ~MAX_UTF8_TWO_BYTE)) \
401 ((U8) __BASE_TWO_BYTE_HI(c, NATIVE_TO_UNI)))
402 #define UTF8_TWO_BYTE_LO(c) \
403 (__ASSERT_((sizeof(c) == 1) \
404 || !(((WIDEST_UTYPE)(c)) & ~MAX_UTF8_TWO_BYTE)) \
405 ((U8) __BASE_TWO_BYTE_LO(c, NATIVE_TO_UNI)))
407 /* This is illegal in any well-formed UTF-8 in both EBCDIC and ASCII
408 * as it is only in overlongs. */
409 #define ILLEGAL_UTF8_BYTE I8_TO_NATIVE_UTF8(0xC1)
412 * 'UTF' is whether or not p is encoded in UTF8. The names 'foo_lazy_if' stem
413 * from an earlier version of these macros in which they didn't call the
414 * foo_utf8() macros (i.e. were 'lazy') unless they decided that *p is the
415 * beginning of a utf8 character. Now that foo_utf8() determines that itself,
416 * no need to do it again here
418 #define isIDFIRST_lazy_if(p,UTF) ((IN_BYTES || !UTF ) \
420 : isIDFIRST_utf8((const U8*)p))
421 #define isWORDCHAR_lazy_if(p,UTF) ((IN_BYTES || (!UTF )) \
423 : isWORDCHAR_utf8((const U8*)p))
424 #define isALNUM_lazy_if(p,UTF) isWORDCHAR_lazy_if(p,UTF)
426 #define UTF8_MAXLEN UTF8_MAXBYTES
428 /* A Unicode character can fold to up to 3 characters */
429 #define UTF8_MAX_FOLD_CHAR_EXPAND 3
431 #define IN_BYTES (CopHINTS_get(PL_curcop) & HINT_BYTES)
432 #define DO_UTF8(sv) (SvUTF8(sv) && !IN_BYTES)
433 #define IN_UNI_8_BIT \
434 (((CopHINTS_get(PL_curcop) & (HINT_UNI_8_BIT)) \
435 || (CopHINTS_get(PL_curcop) & HINT_LOCALE_PARTIAL \
436 /* -1 below is for :not_characters */ \
437 && _is_in_locale_category(FALSE, -1))) \
441 #define UTF8_ALLOW_EMPTY 0x0001 /* Allow a zero length string */
443 /* Allow first byte to be a continuation byte */
444 #define UTF8_ALLOW_CONTINUATION 0x0002
446 /* Allow second... bytes to be non-continuation bytes */
447 #define UTF8_ALLOW_NON_CONTINUATION 0x0004
449 /* expecting more bytes than were available in the string */
450 #define UTF8_ALLOW_SHORT 0x0008
452 /* Overlong sequence; i.e., the code point can be specified in fewer bytes. */
453 #define UTF8_ALLOW_LONG 0x0010
455 #define UTF8_DISALLOW_SURROGATE 0x0020 /* Unicode surrogates */
456 #define UTF8_WARN_SURROGATE 0x0040
458 #define UTF8_DISALLOW_NONCHAR 0x0080 /* Unicode non-character */
459 #define UTF8_WARN_NONCHAR 0x0100 /* code points */
461 #define UTF8_DISALLOW_SUPER 0x0200 /* Super-set of Unicode: code */
462 #define UTF8_WARN_SUPER 0x0400 /* points above the legal max */
464 /* Code points which never were part of the original UTF-8 standard, the first
465 * byte of which is a FE or FF on ASCII platforms. If the first byte is FF, it
466 * will overflow a 32-bit word. If the first byte is FE, it will overflow a
467 * signed 32-bit word. */
468 #define UTF8_DISALLOW_FE_FF 0x0800
469 #define UTF8_WARN_FE_FF 0x1000
471 #define UTF8_CHECK_ONLY 0x2000
473 /* For backwards source compatibility. They do nothing, as the default now
474 * includes what they used to mean. The first one's meaning was to allow the
475 * just the single non-character 0xFFFF */
476 #define UTF8_ALLOW_FFFF 0
477 #define UTF8_ALLOW_SURROGATE 0
479 #define UTF8_DISALLOW_ILLEGAL_INTERCHANGE (UTF8_DISALLOW_SUPER|UTF8_DISALLOW_NONCHAR|UTF8_DISALLOW_SURROGATE|UTF8_DISALLOW_FE_FF)
480 #define UTF8_WARN_ILLEGAL_INTERCHANGE \
481 (UTF8_WARN_SUPER|UTF8_WARN_NONCHAR|UTF8_WARN_SURROGATE|UTF8_WARN_FE_FF)
482 #define UTF8_ALLOW_ANY \
483 (~(UTF8_DISALLOW_ILLEGAL_INTERCHANGE|UTF8_WARN_ILLEGAL_INTERCHANGE))
484 #define UTF8_ALLOW_ANYUV \
486 & ~(UTF8_DISALLOW_ILLEGAL_INTERCHANGE|UTF8_WARN_ILLEGAL_INTERCHANGE))
487 #define UTF8_ALLOW_DEFAULT (ckWARN(WARN_UTF8) ? 0 : \
490 /* Surrogates, non-character code points and above-Unicode code points are
491 * problematic in some contexts. This allows code that needs to check for
492 * those to to quickly exclude the vast majority of code points it will
494 #define UTF8_FIRST_PROBLEMATIC_CODE_POINT_FIRST_BYTE \
495 FIRST_SURROGATE_UTF8_FIRST_BYTE
497 #define UTF8_IS_SURROGATE(s) cBOOL(is_SURROGATE_utf8(s))
498 #define UTF8_IS_REPLACEMENT(s, send) cBOOL(is_REPLACEMENT_utf8_safe(s,send))
501 * U+10FFFF: \xF4\x8F\xBF\xBF \xF9\xA1\xBF\xBF\xBF max legal Unicode
502 * U+110000: \xF4\x90\x80\x80 \xF9\xA2\xA0\xA0\xA0
503 * U+110001: \xF4\x90\x80\x81 \xF9\xA2\xA0\xA0\xA1
505 * BE AWARE that this test doesn't rule out malformed code points, in
506 * particular overlongs */
507 #ifdef EBCDIC /* Both versions assume well-formed UTF8 */
508 # define UTF8_IS_SUPER(s) (NATIVE_UTF8_TO_I8(* (U8*) (s)) >= 0xF9 \
509 && (NATIVE_UTF8_TO_I8(* (U8*) (s)) > 0xF9 \
510 || (NATIVE_UTF8_TO_I8(* ((U8*) (s) + 1)) >= 0xA2)))
512 # define UTF8_IS_SUPER(s) (*(U8*) (s) >= 0xF4 \
513 && (*(U8*) (s) > 0xF4 || (*((U8*) (s) + 1) >= 0x90)))
516 /* These are now machine generated, and the 'given' clause is no longer
518 #define UTF8_IS_NONCHAR_GIVEN_THAT_NON_SUPER_AND_GE_PROBLEMATIC(s) \
519 cBOOL(is_NONCHAR_utf8(s))
520 #define UTF8_IS_NONCHAR_(s) \
521 UTF8_IS_NONCHAR_GIVEN_THAT_NON_SUPER_AND_GE_PROBLEMATIC(s)
523 #define UNICODE_SURROGATE_FIRST 0xD800
524 #define UNICODE_SURROGATE_LAST 0xDFFF
525 #define UNICODE_REPLACEMENT 0xFFFD
526 #define UNICODE_BYTE_ORDER_MARK 0xFEFF
528 /* Though our UTF-8 encoding can go beyond this,
529 * let's be conservative and do as Unicode says. */
530 #define PERL_UNICODE_MAX 0x10FFFF
532 #define UNICODE_WARN_SURROGATE 0x0001 /* UTF-16 surrogates */
533 #define UNICODE_WARN_NONCHAR 0x0002 /* Non-char code points */
534 #define UNICODE_WARN_SUPER 0x0004 /* Above 0x10FFFF */
535 #define UNICODE_WARN_FE_FF 0x0008 /* Above 0x10FFFF */
536 #define UNICODE_DISALLOW_SURROGATE 0x0010
537 #define UNICODE_DISALLOW_NONCHAR 0x0020
538 #define UNICODE_DISALLOW_SUPER 0x0040
539 #define UNICODE_DISALLOW_FE_FF 0x0080
540 #define UNICODE_WARN_ILLEGAL_INTERCHANGE \
541 (UNICODE_WARN_SURROGATE|UNICODE_WARN_NONCHAR|UNICODE_WARN_SUPER)
542 #define UNICODE_DISALLOW_ILLEGAL_INTERCHANGE \
543 (UNICODE_DISALLOW_SURROGATE|UNICODE_DISALLOW_NONCHAR|UNICODE_DISALLOW_SUPER)
545 /* For backward source compatibility, as are now the default */
546 #define UNICODE_ALLOW_SURROGATE 0
547 #define UNICODE_ALLOW_SUPER 0
548 #define UNICODE_ALLOW_ANY 0
550 #define UNICODE_IS_SURROGATE(c) ((c) >= UNICODE_SURROGATE_FIRST && \
551 (c) <= UNICODE_SURROGATE_LAST)
552 #define UNICODE_IS_REPLACEMENT(c) ((c) == UNICODE_REPLACEMENT)
553 #define UNICODE_IS_BYTE_ORDER_MARK(c) ((c) == UNICODE_BYTE_ORDER_MARK)
554 #define UNICODE_IS_NONCHAR(c) ((c >= 0xFDD0 && c <= 0xFDEF) \
555 /* The other noncharacters end in FFFE or FFFF, which \
556 * the mask below catches both of, but beyond the last \
557 * official unicode code point, they aren't \
558 * noncharacters, since those aren't Unicode \
559 * characters at all */ \
560 || ((((c & 0xFFFE) == 0xFFFE)) && ! UNICODE_IS_SUPER(c)))
561 #define UNICODE_IS_SUPER(c) ((c) > PERL_UNICODE_MAX)
562 #define UNICODE_IS_FE_FF(c) ((c) > 0x7FFFFFFF)
564 #define LATIN_SMALL_LETTER_SHARP_S LATIN_SMALL_LETTER_SHARP_S_NATIVE
565 #define LATIN_SMALL_LETTER_Y_WITH_DIAERESIS \
566 LATIN_SMALL_LETTER_Y_WITH_DIAERESIS_NATIVE
567 #define MICRO_SIGN MICRO_SIGN_NATIVE
568 #define LATIN_CAPITAL_LETTER_A_WITH_RING_ABOVE \
569 LATIN_CAPITAL_LETTER_A_WITH_RING_ABOVE_NATIVE
570 #define LATIN_SMALL_LETTER_A_WITH_RING_ABOVE \
571 LATIN_SMALL_LETTER_A_WITH_RING_ABOVE_NATIVE
572 #define UNICODE_GREEK_CAPITAL_LETTER_SIGMA 0x03A3
573 #define UNICODE_GREEK_SMALL_LETTER_FINAL_SIGMA 0x03C2
574 #define UNICODE_GREEK_SMALL_LETTER_SIGMA 0x03C3
575 #define GREEK_SMALL_LETTER_MU 0x03BC
576 #define GREEK_CAPITAL_LETTER_MU 0x039C /* Upper and title case
578 #define LATIN_CAPITAL_LETTER_Y_WITH_DIAERESIS 0x0178 /* Also is title case */
579 #define LATIN_CAPITAL_LETTER_SHARP_S 0x1E9E
580 #define LATIN_SMALL_LETTER_LONG_S 0x017F
581 #define LATIN_SMALL_LIGATURE_LONG_S_T 0xFB05
582 #define LATIN_SMALL_LIGATURE_ST 0xFB06
583 #define KELVIN_SIGN 0x212A
584 #define ANGSTROM_SIGN 0x212B
586 #define UNI_DISPLAY_ISPRINT 0x0001
587 #define UNI_DISPLAY_BACKSLASH 0x0002
588 #define UNI_DISPLAY_QQ (UNI_DISPLAY_ISPRINT|UNI_DISPLAY_BACKSLASH)
589 #define UNI_DISPLAY_REGEX (UNI_DISPLAY_ISPRINT|UNI_DISPLAY_BACKSLASH)
591 #define ANYOF_FOLD_SHARP_S(node, input, end) \
592 (ANYOF_BITMAP_TEST(node, LATIN_SMALL_LETTER_SHARP_S) && \
593 (ANYOF_NONBITMAP(node)) && \
594 (ANYOF_FLAGS(node) & ANYOF_LOC_NONBITMAP_FOLD) && \
595 ((end) > (input) + 1) && \
596 isALPHA_FOLD_EQ((input)[0], 's'))
598 #define SHARP_S_SKIP 2
600 /* If you want to exclude surrogates, and beyond legal Unicode, see the blame
601 * log for earlier versions which gave details for these */
603 /* A helper macro for isUTF8_CHAR, so use that one, and not this one. This is
604 * retained solely for backwards compatibility and may be deprecated and
605 * removed in a future Perl version.
607 * regen/regcharclass.pl generates is_UTF8_CHAR_utf8() macros for up to these
608 * number of bytes. So this has to be coordinated with that file */
610 # define IS_UTF8_CHAR_FAST(n) ((n) <= 3)
612 # define IS_UTF8_CHAR_FAST(n) ((n) <= 4)
616 /* A helper macro for isUTF8_CHAR, so use that one instead of this. This was
617 * generated by regen/regcharclass.pl, and then moved here. The lines that
618 * generated it were then commented out. This was done solely because it takes
619 * on the order of 10 minutes to generate, and is never going to change, unless
620 * the generated code is improved.
622 * The EBCDIC versions have been cut to not cover all of legal Unicode, so
623 * don't take too long to generate, and there is a separate one for each code
624 * page, so they are in regcharclass.h instead of here */
626 UTF8_CHAR: Matches legal UTF-8 encoded characters from 2 through 4 bytes
630 /*** GENERATED CODE ***/
631 #define is_UTF8_CHAR_utf8_no_length_checks(s) \
632 ( ( 0xC2 <= ((U8*)s)[0] && ((U8*)s)[0] <= 0xDF ) ? \
633 ( ( ( ((U8*)s)[1] & 0xC0 ) == 0x80 ) ? 2 : 0 ) \
634 : ( 0xE0 == ((U8*)s)[0] ) ? \
635 ( ( ( ( ((U8*)s)[1] & 0xE0 ) == 0xA0 ) && ( ( ((U8*)s)[2] & 0xC0 ) == 0x80 ) ) ? 3 : 0 )\
636 : ( 0xE1 <= ((U8*)s)[0] && ((U8*)s)[0] <= 0xEF ) ? \
637 ( ( ( ( ((U8*)s)[1] & 0xC0 ) == 0x80 ) && ( ( ((U8*)s)[2] & 0xC0 ) == 0x80 ) ) ? 3 : 0 )\
638 : ( 0xF0 == ((U8*)s)[0] ) ? \
639 ( ( ( ( 0x90 <= ((U8*)s)[1] && ((U8*)s)[1] <= 0xBF ) && ( ( ((U8*)s)[2] & 0xC0 ) == 0x80 ) ) && ( ( ((U8*)s)[3] & 0xC0 ) == 0x80 ) ) ? 4 : 0 )\
640 : ( ( ( ( 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 )
645 =for apidoc Am|STRLEN|isUTF8_CHAR|const U8 *s|const U8 *e
647 Returns the number of bytes beginning at C<s> which form a legal UTF-8 (or
648 UTF-EBCDIC) encoded character, looking no further than C<e - s> bytes into
649 C<s>. Returns 0 if the sequence starting at C<s> through C<e - 1> is not
652 Note that an INVARIANT character (i.e. ASCII on non-EBCDIC
653 machines) is a valid UTF-8 character.
658 #define isUTF8_CHAR(s, e) (UNLIKELY((e) <= (s)) \
660 : (UTF8_IS_INVARIANT(*s)) \
662 : UNLIKELY(((e) - (s)) < UTF8SKIP(s)) \
664 : LIKELY(IS_UTF8_CHAR_FAST(UTF8SKIP(s))) \
665 ? is_UTF8_CHAR_utf8_no_length_checks(s) \
666 : _is_utf8_char_slow(s, e))
668 #define is_utf8_char_buf(buf, buf_end) isUTF8_CHAR(buf, buf_end)
670 /* Do not use; should be deprecated. Use isUTF8_CHAR() instead; this is
671 * retained solely for backwards compatibility */
672 #define IS_UTF8_CHAR(p, n) (isUTF8_CHAR(p, (p) + (n)) == n)
678 * c-indentation-style: bsd
680 * indent-tabs-mode: nil
683 * ex: set ts=8 sts=4 sw=4 et: