This is a live mirror of the Perl 5 development currently hosted at https://github.com/perl/perl5
utf8.c, .h: Clarify pod and comment
[perl5.git] / utf8.h
CommitLineData
a0ed51b3
LW
1/* utf8.h
2 *
83706693 3 * Copyright (C) 2000, 2001, 2002, 2005, 2006, 2007, 2009 by Larry Wall and others
a0ed51b3
LW
4 *
5 * You may distribute under the terms of either the GNU General Public
6 * License or the Artistic License, as specified in the README file.
7 *
8 */
9
39e02b42 10/* Use UTF-8 as the default script encoding?
1e54db1a 11 * Turning this on will break scripts having non-UTF-8 binary
39e02b42
JH
12 * data (such as Latin-1) in string literals. */
13#ifdef USE_UTF8_SCRIPTS
14# define USE_UTF8_IN_NAMES (!IN_BYTES)
15#else
16# define USE_UTF8_IN_NAMES (PL_hints & HINT_UTF8)
17#endif
18
fd7cb289
RGS
19/* Source backward compatibility. */
20#define uvuni_to_utf8(d, uv) uvuni_to_utf8_flags(d, uv, 0)
21#define is_utf8_string_loc(s, len, ep) is_utf8_string_loclen(s, len, ep, 0)
22
e6226b18
KW
23/*
24=for apidoc ibcmp_utf8
25
26This is a synonym for (! foldEQ_utf8())
27
28=cut
29*/
30#define ibcmp_utf8(s1, pe1, l1, u1, s2, pe2, l2, u2) \
31 cBOOL(! foldEQ_utf8(s1, pe1, l1, u1, s2, pe2, l2, u2))
32
1d72bdf6
NIS
33#ifdef EBCDIC
34/* The equivalent of these macros but implementing UTF-EBCDIC
35 are in the following header file:
36 */
37
38#include "utfebcdic.h"
fd7cb289 39
d06134e5 40#else /* ! EBCDIC */
73c4f7a1
GS
41START_EXTERN_C
42
a0ed51b3 43#ifdef DOINIT
6f06b55f 44EXTCONST unsigned char PL_utf8skip[] = {
a0ed51b3
LW
451,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1, /* ascii */
461,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1, /* ascii */
471,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1, /* ascii */
481,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1, /* ascii */
491,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1, /* bogus */
501,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1, /* bogus */
512,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2, /* scripts */
3c77ea2b
GS
523,3,3,3,3,3,3,3,3,3,3,3,3,3,3,3,4,4,4,4,4,4,4,4,5,5,5,5,6,6, /* cjk etc. */
537,13, /* Perl extended (not UTF-8). Up to 72bit allowed (64-bit + reserved). */
a0ed51b3
LW
54};
55#else
6f06b55f 56EXTCONST unsigned char PL_utf8skip[];
a0ed51b3
LW
57#endif
58
73c4f7a1 59END_EXTERN_C
7e2040f0 60
1d72bdf6
NIS
61/* Native character to iso-8859-1 */
62#define NATIVE_TO_ASCII(ch) (ch)
63#define ASCII_TO_NATIVE(ch) (ch)
64/* Transform after encoding */
65#define NATIVE_TO_UTF(ch) (ch)
66#define UTF_TO_NATIVE(ch) (ch)
67/* Transforms in wide UV chars */
68#define UNI_TO_NATIVE(ch) (ch)
69#define NATIVE_TO_UNI(ch) (ch)
70/* Transforms in invariant space */
71#define NATIVE_TO_NEED(enc,ch) (ch)
72#define ASCII_TO_NEED(enc,ch) (ch)
d7578b48 73
d06134e5 74/* As there are no translations, avoid the function wrapper */
1754c1a1
NC
75#define utf8n_to_uvchr utf8n_to_uvuni
76#define uvchr_to_utf8 uvuni_to_utf8
2b9d42f0 77
877d9f0d 78/*
9041c2e3 79
8c007b5a 80 The following table is from Unicode 3.2.
877d9f0d
JH
81
82 Code Points 1st Byte 2nd Byte 3rd Byte 4th Byte
83
375122d7 84 U+0000..U+007F 00..7F
e1b711da 85 U+0080..U+07FF * C2..DF 80..BF
37e2e78e 86 U+0800..U+0FFF E0 * A0..BF 80..BF
375122d7 87 U+1000..U+CFFF E1..EC 80..BF 80..BF
e1b711da 88 U+D000..U+D7FF ED 80..9F 80..BF
37e2e78e 89 U+D800..U+DFFF +++++++ utf16 surrogates, not legal utf8 +++++++
375122d7 90 U+E000..U+FFFF EE..EF 80..BF 80..BF
37e2e78e 91 U+10000..U+3FFFF F0 * 90..BF 80..BF 80..BF
877d9f0d
JH
92 U+40000..U+FFFFF F1..F3 80..BF 80..BF 80..BF
93 U+100000..U+10FFFF F4 80..8F 80..BF 80..BF
94
e1b711da 95Note the gaps before several of the byte entries above marked by '*'. These are
37e2e78e
KW
96caused by legal UTF-8 avoiding non-shortest encodings: it is technically
97possible to UTF-8-encode a single code point in different ways, but that is
98explicitly forbidden, and the shortest possible encoding should always be used
99(and that is what Perl does).
8c007b5a 100
877d9f0d
JH
101 */
102
8c007b5a
JH
103/*
104 Another way to look at it, as bits:
105
106 Code Points 1st Byte 2nd Byte 3rd Byte 4th Byte
107
108 0aaaaaaa 0aaaaaaa
109 00000bbbbbaaaaaa 110bbbbb 10aaaaaa
110 ccccbbbbbbaaaaaa 1110cccc 10bbbbbb 10aaaaaa
111 00000dddccccccbbbbbbaaaaaa 11110ddd 10cccccc 10bbbbbb 10aaaaaa
112
113As you can see, the continuation bytes all begin with C<10>, and the
e1b711da 114leading bits of the start byte tell how many bytes there are in the
8c007b5a
JH
115encoded character.
116
65ab9279
TC
117Perl's extended UTF-8 means we can have start bytes up to FF.
118
8c007b5a
JH
119*/
120
121
c4d5f83a 122#define UNI_IS_INVARIANT(c) (((UV)c) < 0x80)
d06134e5
KW
123/* Note that C0 and C1 are invalid in legal UTF8, so the lower bound of the
124 * below might ought to be C2 */
65ab9279 125#define UTF8_IS_START(c) (((U8)c) >= 0xc0)
c512ce4f
JH
126#define UTF8_IS_CONTINUATION(c) (((U8)c) >= 0x80 && (((U8)c) <= 0xbf))
127#define UTF8_IS_CONTINUED(c) (((U8)c) & 0x80)
db42d148 128#define UTF8_IS_DOWNGRADEABLE_START(c) (((U8)c & 0xfc) == 0xc0)
8850bf83 129
22901f30
RGS
130#define UTF_START_MARK(len) (((len) > 7) ? 0xFF : (0xFE << (7-(len))))
131#define UTF_START_MASK(len) (((len) >= 7) ? 0x00 : (0x1F >> ((len)-2)))
1d72bdf6
NIS
132
133#define UTF_CONTINUATION_MARK 0x80
134#define UTF_ACCUMULATION_SHIFT 6
135#define UTF_CONTINUATION_MASK ((U8)0x3f)
c512ce4f 136
1d68d6cd 137#ifdef HAS_QUAD
5bbb0b5a 138#define UNISKIP(uv) ( (uv) < 0x80 ? 1 : \
1d68d6cd
SC
139 (uv) < 0x800 ? 2 : \
140 (uv) < 0x10000 ? 3 : \
141 (uv) < 0x200000 ? 4 : \
142 (uv) < 0x4000000 ? 5 : \
143 (uv) < 0x80000000 ? 6 : \
9041c2e3 144 (uv) < UTF8_QUAD_MAX ? 7 : 13 )
1d68d6cd
SC
145#else
146/* No, I'm not even going to *TRY* putting #ifdef inside a #define */
5bbb0b5a 147#define UNISKIP(uv) ( (uv) < 0x80 ? 1 : \
1d68d6cd
SC
148 (uv) < 0x800 ? 2 : \
149 (uv) < 0x10000 ? 3 : \
150 (uv) < 0x200000 ? 4 : \
151 (uv) < 0x4000000 ? 5 : \
152 (uv) < 0x80000000 ? 6 : 7 )
153#endif
154
d06134e5
KW
155#endif /* EBCDIC vs ASCII */
156
157/* Rest of these are attributes of Unicode and perl's internals rather than the
158 * encoding, or happen to be the same in both ASCII and EBCDIC (at least at
159 * this level; the macros that some of these call may have different
160 * definitions in the two encodings */
161
162#define NATIVE8_TO_UNI(ch) NATIVE_TO_ASCII(ch) /* a clearer synonym */
163
164#define UTF8_ACCUMULATE(old, new) (((old) << UTF_ACCUMULATION_SHIFT) | (((U8)new) & UTF_CONTINUATION_MASK))
165
2950f2a7
KW
166/* Convert a two (not one) byte utf8 character to a unicode code point value.
167 * Needs just one iteration of accumulate. Should not be used unless it is
168 * known that the two bytes are legal: 1) two-byte start, and 2) continuation.
169 * Note that the result can be larger than 255 if the input character is not
170 * downgradable */
171#define TWO_BYTE_UTF8_TO_UNI(HI, LO) \
172 UTF8_ACCUMULATE((NATIVE_TO_UTF(HI) & UTF_START_MASK(2)), \
173 NATIVE_TO_UTF(LO))
174
d06134e5
KW
175#define UTF8SKIP(s) PL_utf8skip[*(const U8*)(s)]
176
177#define UTF8_IS_INVARIANT(c) UNI_IS_INVARIANT(NATIVE_TO_UTF(c))
178#define NATIVE_IS_INVARIANT(c) UNI_IS_INVARIANT(NATIVE8_TO_UNI(c))
179
180#define MAX_PORTABLE_UTF8_TWO_BYTE 0x3FF /* constrained by EBCDIC */
181
182/* The macros in the next sets are used to generate the two utf8 or utfebcdic
183 * bytes from an ordinal that is known to fit into two bytes; it must be less
184 * than 0x3FF to work across both encodings. */
185/* Nocast allows these to be used in the case label of a switch statement */
0072151e 186#define UTF8_TWO_BYTE_HI_nocast(c) UTF_TO_NATIVE(((c) >> UTF_ACCUMULATION_SHIFT) | (0xFF & UTF_START_MARK(2)))
14c28784 187#define UTF8_TWO_BYTE_LO_nocast(c) UTF_TO_NATIVE(((c) & UTF_CONTINUATION_MASK) | UTF_CONTINUATION_MARK)
d06134e5
KW
188
189#define UTF8_TWO_BYTE_HI(c) ((U8) (UTF8_TWO_BYTE_HI_nocast(c)))
190#define UTF8_TWO_BYTE_LO(c) ((U8) (UTF8_TWO_BYTE_LO_nocast(c)))
191
192/* This name is used when the source is a single byte */
193#define UTF8_EIGHT_BIT_HI(c) UTF8_TWO_BYTE_HI((U8)(c))
194#define UTF8_EIGHT_BIT_LO(c) UTF8_TWO_BYTE_LO((U8)(c))
195
7e2040f0
GS
196/*
197 * Note: we try to be careful never to call the isXXX_utf8() functions
d06134e5
KW
198 * unless we're pretty sure we've seen the beginning of a UTF-8 or UTFEBCDIC
199 * character. Otherwise we risk loading in the heavy-duty swash_init and
200 * swash_fetch routines unnecessarily.
7e2040f0 201 */
d06134e5 202#define isIDFIRST_lazy_if(p,c) ((IN_BYTES || (!c || ! UTF8_IS_START(*((const U8*)p)))) \
7e2040f0 203 ? isIDFIRST(*(p)) \
e1ec3a88 204 : isIDFIRST_utf8((const U8*)p))
d06134e5 205#define isALNUM_lazy_if(p,c) ((IN_BYTES || (!c || ! UTF8_IS_START(*((const U8*)p)))) \
7e2040f0 206 ? isALNUM(*(p)) \
e1ec3a88 207 : isALNUM_utf8((const U8*)p))
1d72bdf6 208
7e2040f0
GS
209#define isIDFIRST_lazy(p) isIDFIRST_lazy_if(p,1)
210#define isALNUM_lazy(p) isALNUM_lazy_if(p,1)
3bd709b1 211
89ebb4a3
JH
212#define UTF8_MAXBYTES 13
213/* How wide can a single UTF-8 encoded character become in bytes.
214 * NOTE: Strictly speaking Perl's UTF-8 should not be called UTF-8
215 * since UTF-8 is an encoding of Unicode and given Unicode's current
216 * upper limit only four bytes is possible. Perl thinks of UTF-8
217 * as a way to encode non-negative integers in a binary format. */
218#define UTF8_MAXLEN UTF8_MAXBYTES
219
89ebb4a3
JH
220/* The maximum number of UTF-8 bytes a single Unicode character can
221 * uppercase/lowercase/fold into; this number depends on the Unicode
222 * version. An example of maximal expansion is the U+03B0 which
223 * uppercases to U+03C5 U+0308 U+0301. The Unicode databases that
d06134e5 224 * tell these things are UnicodeData.txt, CaseFolding.txt, and
89ebb4a3
JH
225 * SpecialCasing.txt. */
226#define UTF8_MAXBYTES_CASE 6
3bd709b1 227
a98fe34d 228#define IN_BYTES (CopHINTS_get(PL_curcop) & HINT_BYTES)
0064a8a9 229#define DO_UTF8(sv) (SvUTF8(sv) && !IN_BYTES)
1863b879 230#define IN_UNI_8_BIT ( (CopHINTS_get(PL_curcop) & HINT_UNI_8_BIT) \
61fc5122 231 && ! IN_LOCALE_RUNTIME && ! IN_BYTES)
1d72bdf6
NIS
232
233#define UTF8_ALLOW_EMPTY 0x0001
234#define UTF8_ALLOW_CONTINUATION 0x0002
235#define UTF8_ALLOW_NON_CONTINUATION 0x0004
37e2e78e 236#define UTF8_ALLOW_FE_FF 0x0008 /* Allow FE or FF start bytes, \
ff97e5cf 237 yields above 0x7fffFFFF = 31 bits */
37e2e78e 238#define UTF8_ALLOW_SHORT 0x0010 /* expecting more bytes */
1d72bdf6 239#define UTF8_ALLOW_SURROGATE 0x0020
9f7f3913 240#define UTF8_ALLOW_FFFF 0x0040 /* Allow UNICODE_ILLEGAL */
37e2e78e 241#define UTF8_ALLOW_LONG 0x0080 /* expecting fewer bytes */
1d72bdf6 242#define UTF8_ALLOW_ANYUV (UTF8_ALLOW_EMPTY|UTF8_ALLOW_FE_FF|\
61843245 243 UTF8_ALLOW_SURROGATE|UTF8_ALLOW_FFFF)
c867b360 244#define UTF8_ALLOW_ANY 0x00FF
e83d50c9 245#define UTF8_CHECK_ONLY 0x0200
9f7f3913
ST
246#define UTF8_ALLOW_DEFAULT (ckWARN(WARN_UTF8) ? 0 : \
247 UTF8_ALLOW_ANYUV)
1d72bdf6 248
c867b360
JH
249#define UNICODE_SURROGATE_FIRST 0xD800
250#define UNICODE_SURROGATE_LAST 0xDFFF
251#define UNICODE_REPLACEMENT 0xFFFD
252#define UNICODE_BYTE_ORDER_MARK 0xFEFF
253#define UNICODE_ILLEGAL 0xFFFF
1d72bdf6 254
b851fbc1 255/* Though our UTF-8 encoding can go beyond this,
b3ab6785 256 * let's be conservative and do as Unicode 5.1 says. */
b851fbc1
JH
257#define PERL_UNICODE_MAX 0x10FFFF
258
259#define UNICODE_ALLOW_SURROGATE 0x0001 /* Allow UTF-16 surrogates (EVIL) */
260#define UNICODE_ALLOW_FDD0 0x0002 /* Allow the U+FDD0...U+FDEF */
9f7f3913
ST
261#define UNICODE_ALLOW_FFFF 0x0004 /* Allow U+FFF[EF], U+1FFF[EF], ... */
262#define UNICODE_ALLOW_SUPER 0x0008 /* Allow past 0x10FFFF */
c867b360 263#define UNICODE_ALLOW_ANY 0x000F
b851fbc1 264
1d72bdf6
NIS
265#define UNICODE_IS_SURROGATE(c) ((c) >= UNICODE_SURROGATE_FIRST && \
266 (c) <= UNICODE_SURROGATE_LAST)
a10ec373 267#define UNICODE_IS_REPLACEMENT(c) ((c) == UNICODE_REPLACEMENT)
872c91ae 268#define UNICODE_IS_BYTE_ORDER_MARK(c) ((c) == UNICODE_BYTE_ORDER_MARK)
1d72bdf6
NIS
269#define UNICODE_IS_ILLEGAL(c) ((c) == UNICODE_ILLEGAL)
270
f067b878
NA
271#ifdef HAS_QUAD
272# define UTF8_QUAD_MAX UINT64_C(0x1000000000)
273#endif
3bd709b1 274
09091399
JH
275#define UNICODE_GREEK_CAPITAL_LETTER_SIGMA 0x03A3
276#define UNICODE_GREEK_SMALL_LETTER_FINAL_SIGMA 0x03C2
277#define UNICODE_GREEK_SMALL_LETTER_SIGMA 0x03C3
9dcbe121 278#define GREEK_SMALL_LETTER_MU 0x03BC
09091399 279
9e55ce06 280#define UNI_DISPLAY_ISPRINT 0x0001
c728cb41
JH
281#define UNI_DISPLAY_BACKSLASH 0x0002
282#define UNI_DISPLAY_QQ (UNI_DISPLAY_ISPRINT|UNI_DISPLAY_BACKSLASH)
283#define UNI_DISPLAY_REGEX (UNI_DISPLAY_ISPRINT|UNI_DISPLAY_BACKSLASH)
9e55ce06 284
5cd46e1f
KW
285#ifndef EBCDIC
286# define LATIN_SMALL_LETTER_SHARP_S 0x00DF
287# define LATIN_SMALL_LETTER_Y_WITH_DIAERESIS 0x00FF
288# define MICRO_SIGN 0x00B5
289#endif
290
291#define ANYOF_FOLD_SHARP_S(node, input, end) \
292 (ANYOF_BITMAP_TEST(node, LATIN_SMALL_LETTER_SHARP_S) && \
3ff7ceb3 293 (ANYOF_FLAGS(node) & ANYOF_NONBITMAP) && \
07b6858f
JH
294 (ANYOF_FLAGS(node) & ANYOF_FOLD) && \
295 ((end) > (input) + 1) && \
296 toLOWER((input)[0]) == 's' && \
297 toLOWER((input)[1]) == 's')
ebc501f0 298#define SHARP_S_SKIP 2
3b0fc154 299
77263263
ST
300#ifdef EBCDIC
301/* IS_UTF8_CHAR() is not ported to EBCDIC */
302#else
3b0fc154
JH
303#define IS_UTF8_CHAR_1(p) \
304 ((p)[0] <= 0x7F)
305#define IS_UTF8_CHAR_2(p) \
306 ((p)[0] >= 0xC2 && (p)[0] <= 0xDF && \
307 (p)[1] >= 0x80 && (p)[1] <= 0xBF)
308#define IS_UTF8_CHAR_3a(p) \
309 ((p)[0] == 0xE0 && \
310 (p)[1] >= 0xA0 && (p)[1] <= 0xBF && \
311 (p)[2] >= 0x80 && (p)[2] <= 0xBF)
312#define IS_UTF8_CHAR_3b(p) \
313 ((p)[0] >= 0xE1 && (p)[0] <= 0xEC && \
314 (p)[1] >= 0x80 && (p)[1] <= 0xBF && \
315 (p)[2] >= 0x80 && (p)[2] <= 0xBF)
316#define IS_UTF8_CHAR_3c(p) \
317 ((p)[0] == 0xED && \
318 (p)[1] >= 0x80 && (p)[1] <= 0xBF && \
319 (p)[2] >= 0x80 && (p)[2] <= 0xBF)
320/* In IS_UTF8_CHAR_3c(p) one could use
321 * (p)[1] >= 0x80 && (p)[1] <= 0x9F
322 * if one wanted to exclude surrogates. */
323#define IS_UTF8_CHAR_3d(p) \
324 ((p)[0] >= 0xEE && (p)[0] <= 0xEF && \
325 (p)[1] >= 0x80 && (p)[1] <= 0xBF && \
326 (p)[2] >= 0x80 && (p)[2] <= 0xBF)
327#define IS_UTF8_CHAR_4a(p) \
328 ((p)[0] == 0xF0 && \
329 (p)[1] >= 0x90 && (p)[1] <= 0xBF && \
330 (p)[2] >= 0x80 && (p)[2] <= 0xBF && \
331 (p)[3] >= 0x80 && (p)[3] <= 0xBF)
332#define IS_UTF8_CHAR_4b(p) \
333 ((p)[0] >= 0xF1 && (p)[0] <= 0xF3 && \
334 (p)[1] >= 0x80 && (p)[1] <= 0xBF && \
335 (p)[2] >= 0x80 && (p)[2] <= 0xBF && \
336 (p)[3] >= 0x80 && (p)[3] <= 0xBF)
337/* In IS_UTF8_CHAR_4c(p) one could use
338 * (p)[0] == 0xF4
339 * if one wanted to stop at the Unicode limit U+10FFFF.
340 * The 0xF7 allows us to go to 0x1fffff (0x200000 would
341 * require five bytes). Not doing any further code points
342 * since that is not needed (and that would not be strict
343 * UTF-8, anyway). The "slow path" in Perl_is_utf8_char()
344 * will take care of the "extended UTF-8". */
345#define IS_UTF8_CHAR_4c(p) \
346 ((p)[0] == 0xF4 && (p)[0] <= 0xF7 && \
347 (p)[1] >= 0x80 && (p)[1] <= 0xBF && \
348 (p)[2] >= 0x80 && (p)[2] <= 0xBF && \
349 (p)[3] >= 0x80 && (p)[3] <= 0xBF)
350
351#define IS_UTF8_CHAR_3(p) \
352 (IS_UTF8_CHAR_3a(p) || \
353 IS_UTF8_CHAR_3b(p) || \
354 IS_UTF8_CHAR_3c(p) || \
355 IS_UTF8_CHAR_3d(p))
356#define IS_UTF8_CHAR_4(p) \
357 (IS_UTF8_CHAR_4a(p) || \
358 IS_UTF8_CHAR_4b(p) || \
359 IS_UTF8_CHAR_4c(p))
360
361/* IS_UTF8_CHAR(p) is strictly speaking wrong (not UTF-8) because it
362 * (1) allows UTF-8 encoded UTF-16 surrogates
363 * (2) it allows code points past U+10FFFF.
364 * The Perl_is_utf8_char() full "slow" code will handle the Perl
365 * "extended UTF-8". */
366#define IS_UTF8_CHAR(p, n) \
367 ((n) == 1 ? IS_UTF8_CHAR_1(p) : \
368 (n) == 2 ? IS_UTF8_CHAR_2(p) : \
369 (n) == 3 ? IS_UTF8_CHAR_3(p) : \
370 (n) == 4 ? IS_UTF8_CHAR_4(p) : 0)
371
768c67ee
JH
372#define IS_UTF8_CHAR_FAST(n) ((n) <= 4)
373
77263263 374#endif /* IS_UTF8_CHAR() for UTF-8 */
e9a8c099
MHM
375
376/*
377 * Local variables:
378 * c-indentation-style: bsd
379 * c-basic-offset: 4
380 * indent-tabs-mode: t
381 * End:
382 *
383 * ex: set ts=8 sts=4 sw=4 noet:
384 */