xref: /openbsd/gnu/usr.bin/perl/utf8.h (revision 7b36286a)
1 /*    utf8.h
2  *
3  *    Copyright (C) 2000, 2001, 2002, 2005 by Larry Wall and others
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 
10 /* Use UTF-8 as the default script encoding?
11  * Turning this on will break scripts having non-UTF-8 binary
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 
19 #ifdef EBCDIC
20 /* The equivalent of these macros but implementing UTF-EBCDIC
21    are in the following header file:
22  */
23 
24 #include "utfebcdic.h"
25 #else
26 START_EXTERN_C
27 
28 #ifdef DOINIT
29 EXTCONST unsigned char PL_utf8skip[] = {
30 1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1, /* ascii */
31 1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1, /* ascii */
32 1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1, /* ascii */
33 1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1, /* ascii */
34 1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1, /* bogus */
35 1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1, /* bogus */
36 2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2, /* scripts */
37 3,3,3,3,3,3,3,3,3,3,3,3,3,3,3,3,4,4,4,4,4,4,4,4,5,5,5,5,6,6,	 /* cjk etc. */
38 7,13, /* Perl extended (not UTF-8).  Up to 72bit allowed (64-bit + reserved). */
39 };
40 #else
41 EXTCONST unsigned char PL_utf8skip[];
42 #endif
43 
44 END_EXTERN_C
45 #define UTF8SKIP(s) PL_utf8skip[*(const U8*)s]
46 
47 /* Native character to iso-8859-1 */
48 #define NATIVE_TO_ASCII(ch)      (ch)
49 #define ASCII_TO_NATIVE(ch)      (ch)
50 /* Transform after encoding */
51 #define NATIVE_TO_UTF(ch)        (ch)
52 #define UTF_TO_NATIVE(ch)        (ch)
53 /* Transforms in wide UV chars */
54 #define UNI_TO_NATIVE(ch)        (ch)
55 #define NATIVE_TO_UNI(ch)        (ch)
56 /* Transforms in invariant space */
57 #define NATIVE_TO_NEED(enc,ch)   (ch)
58 #define ASCII_TO_NEED(enc,ch)    (ch)
59 
60 /* As there are no translations avoid the function wrapper */
61 #define Perl_utf8n_to_uvchr Perl_utf8n_to_uvuni
62 #define Perl_uvchr_to_utf8  Perl_uvuni_to_utf8
63 
64 /*
65 
66  The following table is from Unicode 3.2.
67 
68  Code Points		1st Byte  2nd Byte  3rd Byte  4th Byte
69 
70    U+0000..U+007F	00..7F
71    U+0080..U+07FF	C2..DF    80..BF
72    U+0800..U+0FFF	E0        A0..BF    80..BF
73    U+1000..U+CFFF       E1..EC    80..BF    80..BF
74    U+D000..U+D7FF       ED        80..9F    80..BF
75    U+D800..U+DFFF       ******* ill-formed *******
76    U+E000..U+FFFF       EE..EF    80..BF    80..BF
77   U+10000..U+3FFFF	F0        90..BF    80..BF    80..BF
78   U+40000..U+FFFFF	F1..F3    80..BF    80..BF    80..BF
79  U+100000..U+10FFFF	F4        80..8F    80..BF    80..BF
80 
81 Note the A0..BF in U+0800..U+0FFF, the 80..9F in U+D000...U+D7FF,
82 the 90..BF in U+10000..U+3FFFF, and the 80...8F in U+100000..U+10FFFF.
83 The "gaps" are caused by legal UTF-8 avoiding non-shortest encodings:
84 it is technically possible to UTF-8-encode a single code point in different
85 ways, but that is explicitly forbidden, and the shortest possible encoding
86 should always be used (and that is what Perl does).
87 
88  */
89 
90 /*
91  Another way to look at it, as bits:
92 
93  Code Points                    1st Byte   2nd Byte  3rd Byte  4th Byte
94 
95                     0aaaaaaa     0aaaaaaa
96             00000bbbbbaaaaaa     110bbbbb  10aaaaaa
97             ccccbbbbbbaaaaaa     1110cccc  10bbbbbb  10aaaaaa
98   00000dddccccccbbbbbbaaaaaa     11110ddd  10cccccc  10bbbbbb  10aaaaaa
99 
100 As you can see, the continuation bytes all begin with C<10>, and the
101 leading bits of the start byte tell how many bytes the are in the
102 encoded character.
103 
104 */
105 
106 
107 #define UNI_IS_INVARIANT(c)		(((UV)c) <  0x80)
108 #define UTF8_IS_INVARIANT(c)		UNI_IS_INVARIANT(NATIVE_TO_UTF(c))
109 #define NATIVE_IS_INVARIANT(c)		UNI_IS_INVARIANT(NATIVE_TO_ASCII(c))
110 #define UTF8_IS_START(c)		(((U8)c) >= 0xc0 && (((U8)c) <= 0xfd))
111 #define UTF8_IS_CONTINUATION(c)		(((U8)c) >= 0x80 && (((U8)c) <= 0xbf))
112 #define UTF8_IS_CONTINUED(c) 		(((U8)c) &  0x80)
113 #define UTF8_IS_DOWNGRADEABLE_START(c)	(((U8)c & 0xfc) == 0xc0)
114 
115 #define UTF_START_MARK(len) ((len >  7) ? 0xFF : (0xFE << (7-len)))
116 #define UTF_START_MASK(len) ((len >= 7) ? 0x00 : (0x1F >> (len-2)))
117 
118 #define UTF_CONTINUATION_MARK		0x80
119 #define UTF_ACCUMULATION_SHIFT		6
120 #define UTF_CONTINUATION_MASK		((U8)0x3f)
121 #define UTF8_ACCUMULATE(old, new)	(((old) << UTF_ACCUMULATION_SHIFT) | (((U8)new) & UTF_CONTINUATION_MASK))
122 
123 #define UTF8_EIGHT_BIT_HI(c)	((((U8)(c))>>UTF_ACCUMULATION_SHIFT)|UTF_START_MARK(2))
124 #define UTF8_EIGHT_BIT_LO(c)	(((((U8)(c)))&UTF_CONTINUATION_MASK)|UTF_CONTINUATION_MARK)
125 
126 #ifdef HAS_QUAD
127 #define UNISKIP(uv) ( (uv) < 0x80           ? 1 : \
128 		      (uv) < 0x800          ? 2 : \
129 		      (uv) < 0x10000        ? 3 : \
130 		      (uv) < 0x200000       ? 4 : \
131 		      (uv) < 0x4000000      ? 5 : \
132 		      (uv) < 0x80000000     ? 6 : \
133                       (uv) < UTF8_QUAD_MAX ? 7 : 13 )
134 #else
135 /* No, I'm not even going to *TRY* putting #ifdef inside a #define */
136 #define UNISKIP(uv) ( (uv) < 0x80           ? 1 : \
137 		      (uv) < 0x800          ? 2 : \
138 		      (uv) < 0x10000        ? 3 : \
139 		      (uv) < 0x200000       ? 4 : \
140 		      (uv) < 0x4000000      ? 5 : \
141 		      (uv) < 0x80000000     ? 6 : 7 )
142 #endif
143 
144 /*
145  * Note: we try to be careful never to call the isXXX_utf8() functions
146  * unless we're pretty sure we've seen the beginning of a UTF-8 character
147  * (that is, the two high bits are set).  Otherwise we risk loading in the
148  * heavy-duty SWASHINIT and SWASHGET routines unnecessarily.
149  */
150 #define isIDFIRST_lazy_if(p,c) ((IN_BYTES || (!c || (*((const U8*)p) < 0xc0))) \
151 				? isIDFIRST(*(p)) \
152 				: isIDFIRST_utf8((U8*)p))
153 #define isALNUM_lazy_if(p,c)   ((IN_BYTES || (!c || (*((const U8*)p) < 0xc0))) \
154 				? isALNUM(*(p)) \
155 				: isALNUM_utf8((U8*)p))
156 
157 
158 #endif /* EBCDIC vs ASCII */
159 
160 /* Rest of these are attributes of Unicode and perl's internals rather than the encoding */
161 
162 #define isIDFIRST_lazy(p)	isIDFIRST_lazy_if(p,1)
163 #define isALNUM_lazy(p)		isALNUM_lazy_if(p,1)
164 
165 #define UTF8_MAXBYTES 13
166 /* How wide can a single UTF-8 encoded character become in bytes.
167  * NOTE: Strictly speaking Perl's UTF-8 should not be called UTF-8
168  * since UTF-8 is an encoding of Unicode and given Unicode's current
169  * upper limit only four bytes is possible.  Perl thinks of UTF-8
170  * as a way to encode non-negative integers in a binary format. */
171 #define UTF8_MAXLEN UTF8_MAXBYTES
172 
173 #define UTF8_MAXLEN_UCLC 3		/* Obsolete, do not use. */
174 #define UTF8_MAXLEN_UCLC_MULT 39	/* Obsolete, do not use. */
175 #define UTF8_MAXLEN_FOLD 3		/* Obsolete, do not use. */
176 #define UTF8_MAXLEN_FOLD_MULT 39	/* Obsolete, do not use. */
177 
178 /* The maximum number of UTF-8 bytes a single Unicode character can
179  * uppercase/lowercase/fold into; this number depends on the Unicode
180  * version.  An example of maximal expansion is the U+03B0 which
181  * uppercases to U+03C5 U+0308 U+0301.  The Unicode databases that
182  * tell these things are UnicodeDatabase.txt, CaseFolding.txt, and
183  * SpecialCasing.txt. */
184 #define UTF8_MAXBYTES_CASE	6
185 
186 #define IN_BYTES (PL_curcop->op_private & HINT_BYTES)
187 #define DO_UTF8(sv) (SvUTF8(sv) && !IN_BYTES)
188 
189 #define UTF8_ALLOW_EMPTY		0x0001
190 #define UTF8_ALLOW_CONTINUATION		0x0002
191 #define UTF8_ALLOW_NON_CONTINUATION	0x0004
192 #define UTF8_ALLOW_FE_FF		0x0008
193 #define UTF8_ALLOW_SHORT		0x0010
194 #define UTF8_ALLOW_SURROGATE		0x0020
195 #define UTF8_ALLOW_FFFF			0x0040 /* Allows also FFFE. */
196 #define UTF8_ALLOW_LONG			0x0080
197 #define UTF8_ALLOW_ANYUV		(UTF8_ALLOW_EMPTY|UTF8_ALLOW_FE_FF|\
198 					 UTF8_ALLOW_SURROGATE|UTF8_ALLOW_FFFF)
199 #define UTF8_ALLOW_ANY			0x00FF
200 #define UTF8_CHECK_ONLY			0x0200
201 
202 #define UNICODE_SURROGATE_FIRST		0xD800
203 #define UNICODE_SURROGATE_LAST		0xDFFF
204 #define UNICODE_REPLACEMENT		0xFFFD
205 #define UNICODE_BYTE_ORDER_MARK		0xFEFF
206 #define UNICODE_ILLEGAL			0xFFFF
207 
208 /* Though our UTF-8 encoding can go beyond this,
209  * let's be conservative and do as Unicode 3.2 says. */
210 #define PERL_UNICODE_MAX	0x10FFFF
211 
212 #define UNICODE_ALLOW_SURROGATE 0x0001	/* Allow UTF-16 surrogates (EVIL) */
213 #define UNICODE_ALLOW_FDD0	0x0002	/* Allow the U+FDD0...U+FDEF */
214 #define UNICODE_ALLOW_FFFF	0x0004	/* Allow 0xFFF[EF], 0x1FFF[EF], ... */
215 #define UNICODE_ALLOW_SUPER	0x0008	/* Allow past 10xFFFF */
216 #define UNICODE_ALLOW_ANY	0x000F
217 
218 #define UNICODE_IS_SURROGATE(c)		((c) >= UNICODE_SURROGATE_FIRST && \
219 					 (c) <= UNICODE_SURROGATE_LAST)
220 #define UNICODE_IS_REPLACEMENT(c)	((c) == UNICODE_REPLACEMENT)
221 #define UNICODE_IS_BYTE_ORDER_MARK(c)	((c) == UNICODE_BYTE_ORDER_MARK)
222 #define UNICODE_IS_ILLEGAL(c)		((c) == UNICODE_ILLEGAL)
223 
224 #ifdef HAS_QUAD
225 #    define UTF8_QUAD_MAX	UINT64_C(0x1000000000)
226 #endif
227 
228 #define UTF8_IS_ASCII(c) UTF8_IS_INVARIANT(c)
229 
230 #define UNICODE_LATIN_SMALL_LETTER_SHARP_S	0x00DF
231 #define UNICODE_GREEK_CAPITAL_LETTER_SIGMA	0x03A3
232 #define UNICODE_GREEK_SMALL_LETTER_FINAL_SIGMA	0x03C2
233 #define UNICODE_GREEK_SMALL_LETTER_SIGMA	0x03C3
234 
235 #define EBCDIC_LATIN_SMALL_LETTER_SHARP_S	0x0059
236 
237 #define UNI_DISPLAY_ISPRINT	0x0001
238 #define UNI_DISPLAY_BACKSLASH	0x0002
239 #define UNI_DISPLAY_QQ		(UNI_DISPLAY_ISPRINT|UNI_DISPLAY_BACKSLASH)
240 #define UNI_DISPLAY_REGEX	(UNI_DISPLAY_ISPRINT|UNI_DISPLAY_BACKSLASH)
241 
242 #ifdef EBCDIC
243 #   define ANYOF_FOLD_SHARP_S(node, input, end)	\
244 	(ANYOF_BITMAP_TEST(node, EBCDIC_LATIN_SMALL_LETTER_SHARP_S) && \
245 	 (ANYOF_FLAGS(node) & ANYOF_UNICODE) && \
246 	 (ANYOF_FLAGS(node) & ANYOF_FOLD) && \
247 	 ((end) > (input) + 1) && \
248 	 toLOWER((input)[0]) == 's' && \
249 	 toLOWER((input)[1]) == 's')
250 #else
251 #   define ANYOF_FOLD_SHARP_S(node, input, end)	\
252 	(ANYOF_BITMAP_TEST(node, UNICODE_LATIN_SMALL_LETTER_SHARP_S) && \
253 	 (ANYOF_FLAGS(node) & ANYOF_UNICODE) && \
254 	 (ANYOF_FLAGS(node) & ANYOF_FOLD) && \
255 	 ((end) > (input) + 1) && \
256 	 toLOWER((input)[0]) == 's' && \
257 	 toLOWER((input)[1]) == 's')
258 #endif
259 #define SHARP_S_SKIP 2
260 
261 #ifdef EBCDIC
262 /* IS_UTF8_CHAR() is not ported to EBCDIC */
263 #else
264 #define IS_UTF8_CHAR_1(p)	\
265 	((p)[0] <= 0x7F)
266 #define IS_UTF8_CHAR_2(p)	\
267 	((p)[0] >= 0xC2 && (p)[0] <= 0xDF && \
268 	 (p)[1] >= 0x80 && (p)[1] <= 0xBF)
269 #define IS_UTF8_CHAR_3a(p)	\
270 	((p)[0] == 0xE0 && \
271 	 (p)[1] >= 0xA0 && (p)[1] <= 0xBF && \
272 	 (p)[2] >= 0x80 && (p)[2] <= 0xBF)
273 #define IS_UTF8_CHAR_3b(p)	\
274 	((p)[0] >= 0xE1 && (p)[0] <= 0xEC && \
275 	 (p)[1] >= 0x80 && (p)[1] <= 0xBF && \
276 	 (p)[2] >= 0x80 && (p)[2] <= 0xBF)
277 #define IS_UTF8_CHAR_3c(p)	\
278 	((p)[0] == 0xED && \
279 	 (p)[1] >= 0x80 && (p)[1] <= 0xBF && \
280 	 (p)[2] >= 0x80 && (p)[2] <= 0xBF)
281 /* In IS_UTF8_CHAR_3c(p) one could use
282  * (p)[1] >= 0x80 && (p)[1] <= 0x9F
283  * if one wanted to exclude surrogates. */
284 #define IS_UTF8_CHAR_3d(p)	\
285 	((p)[0] >= 0xEE && (p)[0] <= 0xEF && \
286 	 (p)[1] >= 0x80 && (p)[1] <= 0xBF && \
287 	 (p)[2] >= 0x80 && (p)[2] <= 0xBF)
288 #define IS_UTF8_CHAR_4a(p)	\
289 	((p)[0] == 0xF0 && \
290 	 (p)[1] >= 0x90 && (p)[1] <= 0xBF && \
291 	 (p)[2] >= 0x80 && (p)[2] <= 0xBF && \
292 	 (p)[3] >= 0x80 && (p)[3] <= 0xBF)
293 #define IS_UTF8_CHAR_4b(p)	\
294 	((p)[0] >= 0xF1 && (p)[0] <= 0xF3 && \
295 	 (p)[1] >= 0x80 && (p)[1] <= 0xBF && \
296 	 (p)[2] >= 0x80 && (p)[2] <= 0xBF && \
297 	 (p)[3] >= 0x80 && (p)[3] <= 0xBF)
298 /* In IS_UTF8_CHAR_4c(p) one could use
299  * (p)[0] == 0xF4
300  * if one wanted to stop at the Unicode limit U+10FFFF.
301  * The 0xF7 allows us to go to 0x1fffff (0x200000 would
302  * require five bytes).  Not doing any further code points
303  * since that is not needed (and that would not be strict
304  * UTF-8, anyway).  The "slow path" in Perl_is_utf8_char()
305  * will take care of the "extended UTF-8". */
306 #define IS_UTF8_CHAR_4c(p)	\
307 	((p)[0] == 0xF4 && (p)[0] <= 0xF7 && \
308 	 (p)[1] >= 0x80 && (p)[1] <= 0xBF && \
309 	 (p)[2] >= 0x80 && (p)[2] <= 0xBF && \
310 	 (p)[3] >= 0x80 && (p)[3] <= 0xBF)
311 
312 #define IS_UTF8_CHAR_3(p)	\
313 	(IS_UTF8_CHAR_3a(p) || \
314 	 IS_UTF8_CHAR_3b(p) || \
315 	 IS_UTF8_CHAR_3c(p) || \
316 	 IS_UTF8_CHAR_3d(p))
317 #define IS_UTF8_CHAR_4(p)	\
318 	(IS_UTF8_CHAR_4a(p) || \
319 	 IS_UTF8_CHAR_4b(p) || \
320 	 IS_UTF8_CHAR_4c(p))
321 
322 /* IS_UTF8_CHAR(p) is strictly speaking wrong (not UTF-8) because it
323  * (1) allows UTF-8 encoded UTF-16 surrogates
324  * (2) it allows code points past U+10FFFF.
325  * The Perl_is_utf8_char() full "slow" code will handle the Perl
326  * "extended UTF-8". */
327 #define IS_UTF8_CHAR(p, n)	\
328 	((n) == 1 ? IS_UTF8_CHAR_1(p) : \
329  	 (n) == 2 ? IS_UTF8_CHAR_2(p) : \
330 	 (n) == 3 ? IS_UTF8_CHAR_3(p) : \
331 	 (n) == 4 ? IS_UTF8_CHAR_4(p) : 0)
332 
333 #define IS_UTF8_CHAR_FAST(n) ((n) <= 4)
334 
335 #endif /* IS_UTF8_CHAR() for UTF-8 */
336