Commit | Line | Data |
---|---|---|
a0ed51b3 LW |
1 | /* utf8.h |
2 | * | |
f3cb6f94 KW |
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. | |
8 | * | |
2eee27d7 SS |
9 | * Copyright (C) 2000, 2001, 2002, 2005, 2006, 2007, 2009, |
10 | * 2010, 2011 by Larry Wall and others | |
a0ed51b3 LW |
11 | * |
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. | |
14 | * | |
15 | */ | |
16 | ||
6a5bc5ac KW |
17 | #ifndef PERL_UTF8_H_ /* Guard against recursive inclusion */ |
18 | #define PERL_UTF8_H_ 1 | |
57f0e7e2 | 19 | |
39e02b42 | 20 | /* Use UTF-8 as the default script encoding? |
1e54db1a | 21 | * Turning this on will break scripts having non-UTF-8 binary |
39e02b42 JH |
22 | * data (such as Latin-1) in string literals. */ |
23 | #ifdef USE_UTF8_SCRIPTS | |
24 | # define USE_UTF8_IN_NAMES (!IN_BYTES) | |
25 | #else | |
26 | # define USE_UTF8_IN_NAMES (PL_hints & HINT_UTF8) | |
27 | #endif | |
28 | ||
3cd96634 KW |
29 | #include "regcharclass.h" |
30 | #include "unicode_constants.h" | |
31 | ||
051a06d4 | 32 | /* For to_utf8_fold_flags, q.v. */ |
e4f4ef45 KW |
33 | #define FOLD_FLAGS_LOCALE 0x1 |
34 | #define FOLD_FLAGS_FULL 0x2 | |
35 | #define FOLD_FLAGS_NOMIX_ASCII 0x4 | |
051a06d4 | 36 | |
7bbfa158 | 37 | /* |
7bbfa158 KW |
38 | =for apidoc is_ascii_string |
39 | ||
8871a094 | 40 | This is a misleadingly-named synonym for L</is_utf8_invariant_string>. |
7bbfa158 KW |
41 | On ASCII-ish platforms, the name isn't misleading: the ASCII-range characters |
42 | are exactly the UTF-8 invariants. But EBCDIC machines have more invariants | |
8871a094 KW |
43 | than just the ASCII characters, so C<is_utf8_invariant_string> is preferred. |
44 | ||
45 | =for apidoc is_invariant_string | |
46 | ||
47 | This is a somewhat misleadingly-named synonym for L</is_utf8_invariant_string>. | |
48 | C<is_utf8_invariant_string> is preferred, as it indicates under what conditions | |
49 | the string is invariant. | |
7bbfa158 KW |
50 | |
51 | =cut | |
52 | */ | |
8871a094 KW |
53 | #define is_ascii_string(s, len) is_utf8_invariant_string(s, len) |
54 | #define is_invariant_string(s, len) is_utf8_invariant_string(s, len) | |
7bbfa158 | 55 | |
33f38593 KW |
56 | #define uvoffuni_to_utf8_flags(d,uv,flags) \ |
57 | uvoffuni_to_utf8_flags_msgs(d, uv, flags, 0) | |
de69f3af KW |
58 | #define uvchr_to_utf8(a,b) uvchr_to_utf8_flags(a,b,0) |
59 | #define uvchr_to_utf8_flags(d,uv,flags) \ | |
33f38593 KW |
60 | uvchr_to_utf8_flags_msgs(d,uv,flags, 0) |
61 | #define uvchr_to_utf8_flags_msgs(d,uv,flags,msgs) \ | |
62 | uvoffuni_to_utf8_flags_msgs(d,NATIVE_TO_UNI(uv),flags, msgs) | |
de69f3af | 63 | #define utf8_to_uvchr_buf(s, e, lenp) \ |
9a9a6c98 | 64 | utf8_to_uvchr_buf_helper((const U8 *) (s), (const U8 *) e, lenp) |
f9380377 KW |
65 | #define utf8n_to_uvchr(s, len, lenp, flags) \ |
66 | utf8n_to_uvchr_error(s, len, lenp, flags, 0) | |
37657a5b KW |
67 | #define utf8n_to_uvchr_error(s, len, lenp, flags, errors) \ |
68 | utf8n_to_uvchr_msgs(s, len, lenp, flags, errors, 0) | |
de69f3af | 69 | |
a0270393 | 70 | #define to_uni_fold(c, p, lenp) _to_uni_fold_flags(c, p, lenp, FOLD_FLAGS_FULL) |
a239b1e2 | 71 | |
eda9cac1 | 72 | #define foldEQ_utf8(s1, pe1, l1, u1, s2, pe2, l2, u2) \ |
1604cfb0 | 73 | foldEQ_utf8_flags(s1, pe1, l1, u1, s2, pe2, l2, u2, 0) |
baa60164 | 74 | #define FOLDEQ_UTF8_NOMIX_ASCII (1 << 0) |
cea315b6 | 75 | #define FOLDEQ_LOCALE (1 << 1) |
18f762c3 KW |
76 | #define FOLDEQ_S1_ALREADY_FOLDED (1 << 2) |
77 | #define FOLDEQ_S2_ALREADY_FOLDED (1 << 3) | |
d635b710 KW |
78 | #define FOLDEQ_S1_FOLDS_SANE (1 << 4) |
79 | #define FOLDEQ_S2_FOLDS_SANE (1 << 5) | |
a33c29bc | 80 | |
fcd03d92 KW |
81 | /* This will be described more fully below, but it turns out that the |
82 | * fundamental difference between UTF-8 and UTF-EBCDIC is that the former has | |
83 | * the upper 2 bits of a continuation byte be '10', and the latter has the | |
84 | * upper 3 bits be '101', leaving 6 and 5 significant bits respectively. | |
85 | * | |
86 | * It is helpful to know the EBCDIC value on ASCII platforms, mainly to avoid | |
87 | * some #ifdef's */ | |
88 | #define UTF_EBCDIC_CONTINUATION_BYTE_INFO_BITS 5 | |
89 | ||
1d72bdf6 NIS |
90 | #ifdef EBCDIC |
91 | /* The equivalent of these macros but implementing UTF-EBCDIC | |
92 | are in the following header file: | |
93 | */ | |
94 | ||
95 | #include "utfebcdic.h" | |
fd7cb289 | 96 | |
d06134e5 | 97 | #else /* ! EBCDIC */ |
73c4f7a1 GS |
98 | START_EXTERN_C |
99 | ||
f4d83e55 KW |
100 | /* |
101 | ||
102 | =for apidoc AmnU|STRLEN|UTF8_MAXBYTES | |
103 | ||
104 | The maximum width of a single UTF-8 encoded character, in bytes. | |
105 | ||
106 | NOTE: Strictly speaking Perl's UTF-8 should not be called UTF-8 since UTF-8 | |
107 | is an encoding of Unicode, and Unicode's upper limit, 0x10FFFF, can be | |
108 | expressed with 4 bytes. However, Perl thinks of UTF-8 as a way to encode | |
109 | non-negative integers in a binary format, even those above Unicode. | |
110 | ||
111 | =cut | |
112 | */ | |
111e8ed9 KW |
113 | #define UTF8_MAXBYTES 13 |
114 | ||
a0ed51b3 | 115 | #ifdef DOINIT |
6f06b55f | 116 | EXTCONST unsigned char PL_utf8skip[] = { |
b2635aa8 KW |
117 | /* 0x00 */ 1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1, /* ascii */ |
118 | /* 0x10 */ 1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1, /* ascii */ | |
119 | /* 0x20 */ 1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1, /* ascii */ | |
120 | /* 0x30 */ 1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1, /* ascii */ | |
121 | /* 0x40 */ 1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1, /* ascii */ | |
122 | /* 0x50 */ 1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1, /* ascii */ | |
123 | /* 0x60 */ 1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1, /* ascii */ | |
124 | /* 0x70 */ 1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1, /* ascii */ | |
125 | /* 0x80 */ 1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1, /* bogus: continuation byte */ | |
126 | /* 0x90 */ 1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1, /* bogus: continuation byte */ | |
127 | /* 0xA0 */ 1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1, /* bogus: continuation byte */ | |
128 | /* 0xB0 */ 1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1, /* bogus: continuation byte */ | |
129 | /* 0xC0 */ 2,2, /* overlong */ | |
1ff3baa2 | 130 | /* 0xC2 */ 2,2,2,2,2,2,2,2,2,2,2,2,2,2, /* U+0080 to U+03FF */ |
b2635aa8 KW |
131 | /* 0xD0 */ 2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2, /* U+0400 to U+07FF */ |
132 | /* 0xE0 */ 3,3,3,3,3,3,3,3,3,3,3,3,3,3,3,3, /* U+0800 to U+FFFF */ | |
133 | /* 0xF0 */ 4,4,4,4,4,4,4,4,5,5,5,5,6,6, /* above BMP to 2**31 - 1 */ | |
6937f885 KW |
134 | /* Perl extended (never was official UTF-8). Up to 36 bit */ |
135 | /* 0xFE */ 7, | |
136 | /* More extended, Up to 72 bits (64-bit + reserved) */ | |
111e8ed9 | 137 | /* 0xFF */ UTF8_MAXBYTES |
a0ed51b3 LW |
138 | }; |
139 | #else | |
6f06b55f | 140 | EXTCONST unsigned char PL_utf8skip[]; |
a0ed51b3 LW |
141 | #endif |
142 | ||
73c4f7a1 | 143 | END_EXTERN_C |
7e2040f0 | 144 | |
ef2a4c8f KW |
145 | /* |
146 | ||
147 | =for apidoc Am|U8|NATIVE_TO_LATIN1|U8 ch | |
148 | ||
149 | Returns the Latin-1 (including ASCII and control characters) equivalent of the | |
150 | input native code point given by C<ch>. Thus, C<NATIVE_TO_LATIN1(193)> on | |
151 | EBCDIC platforms returns 65. These each represent the character C<"A"> on | |
152 | their respective platforms. On ASCII platforms no conversion is needed, so | |
153 | this macro expands to just its input, adding no time nor space requirements to | |
154 | the implementation. | |
155 | ||
156 | For conversion of code points potentially larger than will fit in a character, | |
157 | use L</NATIVE_TO_UNI>. | |
158 | ||
159 | =for apidoc Am|U8|LATIN1_TO_NATIVE|U8 ch | |
160 | ||
161 | Returns the native equivalent of the input Latin-1 code point (including ASCII | |
162 | and control characters) given by C<ch>. Thus, C<LATIN1_TO_NATIVE(66)> on | |
163 | EBCDIC platforms returns 194. These each represent the character C<"B"> on | |
164 | their respective platforms. On ASCII platforms no conversion is needed, so | |
165 | this macro expands to just its input, adding no time nor space requirements to | |
166 | the implementation. | |
167 | ||
168 | For conversion of code points potentially larger than will fit in a character, | |
169 | use L</UNI_TO_NATIVE>. | |
170 | ||
171 | =for apidoc Am|UV|NATIVE_TO_UNI|UV ch | |
172 | ||
173 | Returns the Unicode equivalent of the input native code point given by C<ch>. | |
174 | Thus, C<NATIVE_TO_UNI(195)> on EBCDIC platforms returns 67. These each | |
175 | represent the character C<"C"> on their respective platforms. On ASCII | |
176 | platforms no conversion is needed, so this macro expands to just its input, | |
177 | adding no time nor space requirements to the implementation. | |
178 | ||
179 | =for apidoc Am|UV|UNI_TO_NATIVE|UV ch | |
180 | ||
181 | Returns the native equivalent of the input Unicode code point given by C<ch>. | |
182 | Thus, C<UNI_TO_NATIVE(68)> on EBCDIC platforms returns 196. These each | |
183 | represent the character C<"D"> on their respective platforms. On ASCII | |
184 | platforms no conversion is needed, so this macro expands to just its input, | |
185 | adding no time nor space requirements to the implementation. | |
186 | ||
187 | =cut | |
188 | */ | |
189 | ||
e7ae132e KW |
190 | #define NATIVE_TO_LATIN1(ch) (__ASSERT_(FITS_IN_8_BITS(ch)) ((U8) ((ch) | 0))) |
191 | #define LATIN1_TO_NATIVE(ch) (__ASSERT_(FITS_IN_8_BITS(ch)) ((U8) ((ch) | 0))) | |
59a449d5 KW |
192 | |
193 | /* I8 is an intermediate version of UTF-8 used only in UTF-EBCDIC. We thus | |
194 | * consider it to be identical to UTF-8 on ASCII platforms. Strictly speaking | |
195 | * UTF-8 and UTF-EBCDIC are two different things, but we often conflate them | |
196 | * because they are 8-bit encodings that serve the same purpose in Perl, and | |
197 | * rarely do we need to distinguish them. The term "NATIVE_UTF8" applies to | |
198 | * whichever one is applicable on the current platform */ | |
cb15eeb2 KW |
199 | #define NATIVE_UTF8_TO_I8(ch) (__ASSERT_(FITS_IN_8_BITS(ch)) ((U8) ((ch) | 0))) |
200 | #define I8_TO_NATIVE_UTF8(ch) (__ASSERT_(FITS_IN_8_BITS(ch)) ((U8) ((ch) | 0))) | |
59a449d5 | 201 | |
cb15eeb2 KW |
202 | #define UNI_TO_NATIVE(ch) ((UV) ((ch) | 0)) |
203 | #define NATIVE_TO_UNI(ch) ((UV) ((ch) | 0)) | |
d7578b48 | 204 | |
877d9f0d | 205 | /* |
9041c2e3 | 206 | |
a14e0a36 KW |
207 | The following table is from Unicode 3.2, plus the Perl extensions for above |
208 | U+10FFFF | |
877d9f0d | 209 | |
a14e0a36 | 210 | Code Points 1st Byte 2nd Byte 3rd 4th 5th 6th 7th 8th-13th |
877d9f0d | 211 | |
375122d7 | 212 | U+0000..U+007F 00..7F |
e1b711da | 213 | U+0080..U+07FF * C2..DF 80..BF |
a14e0a36 KW |
214 | U+0800..U+0FFF E0 * A0..BF 80..BF |
215 | U+1000..U+CFFF E1..EC 80..BF 80..BF | |
216 | U+D000..U+D7FF ED 80..9F 80..BF | |
217 | U+D800..U+DFFF ED A0..BF 80..BF (surrogates) | |
218 | U+E000..U+FFFF EE..EF 80..BF 80..BF | |
219 | U+10000..U+3FFFF F0 * 90..BF 80..BF 80..BF | |
220 | U+40000..U+FFFFF F1..F3 80..BF 80..BF 80..BF | |
221 | U+100000..U+10FFFF F4 80..8F 80..BF 80..BF | |
222 | Below are above-Unicode code points | |
223 | U+110000..U+13FFFF F4 90..BF 80..BF 80..BF | |
224 | U+110000..U+1FFFFF F5..F7 80..BF 80..BF 80..BF | |
225 | U+200000..U+FFFFFF F8 * 88..BF 80..BF 80..BF 80..BF | |
226 | U+1000000..U+3FFFFFF F9..FB 80..BF 80..BF 80..BF 80..BF | |
227 | U+4000000..U+3FFFFFFF FC * 84..BF 80..BF 80..BF 80..BF 80..BF | |
228 | U+40000000..U+7FFFFFFF FD 80..BF 80..BF 80..BF 80..BF 80..BF | |
229 | U+80000000..U+FFFFFFFFF FE * 82..BF 80..BF 80..BF 80..BF 80..BF 80..BF | |
230 | U+1000000000.. FF 80..BF 80..BF 80..BF 80..BF 80..BF * 81..BF 80..BF | |
877d9f0d | 231 | |
e1b711da | 232 | Note the gaps before several of the byte entries above marked by '*'. These are |
37e2e78e KW |
233 | caused by legal UTF-8 avoiding non-shortest encodings: it is technically |
234 | possible to UTF-8-encode a single code point in different ways, but that is | |
235 | explicitly forbidden, and the shortest possible encoding should always be used | |
15824458 | 236 | (and that is what Perl does). The non-shortest ones are called 'overlongs'. |
8c007b5a | 237 | |
877d9f0d JH |
238 | */ |
239 | ||
8c007b5a JH |
240 | /* |
241 | Another way to look at it, as bits: | |
242 | ||
b2635aa8 | 243 | Code Points 1st Byte 2nd Byte 3rd Byte 4th Byte |
8c007b5a | 244 | |
b2635aa8 KW |
245 | 0aaa aaaa 0aaa aaaa |
246 | 0000 0bbb bbaa aaaa 110b bbbb 10aa aaaa | |
247 | cccc bbbb bbaa aaaa 1110 cccc 10bb bbbb 10aa aaaa | |
248 | 00 000d ddcc cccc bbbb bbaa aaaa 1111 0ddd 10cc cccc 10bb bbbb 10aa aaaa | |
8c007b5a JH |
249 | |
250 | As you can see, the continuation bytes all begin with C<10>, and the | |
e1b711da | 251 | leading bits of the start byte tell how many bytes there are in the |
8c007b5a JH |
252 | encoded character. |
253 | ||
df863e43 KW |
254 | Perl's extended UTF-8 means we can have start bytes up through FF, though any |
255 | beginning with FF yields a code point that is too large for 32-bit ASCII | |
256 | platforms. FF signals to use 13 bytes for the encoded character. This breaks | |
257 | the paradigm that the number of leading bits gives how many total bytes there | |
ab2e28c2 | 258 | are in the character. */ |
38953e5a | 259 | |
15824458 KW |
260 | /* This is the number of low-order bits a continuation byte in a UTF-8 encoded |
261 | * sequence contributes to the specification of the code point. In the bit | |
262 | * maps above, you see that the first 2 bits are a constant '10', leaving 6 of | |
263 | * real information */ | |
36da1e17 | 264 | # define UTF_CONTINUATION_BYTE_INFO_BITS 6 |
b2635aa8 | 265 | |
fed423a5 KW |
266 | /* ^? is defined to be DEL on ASCII systems. See the definition of toCTRL() |
267 | * for more */ | |
268 | #define QUESTION_MARK_CTRL DEL_NATIVE | |
269 | ||
270 | /* Surrogates, non-character code points and above-Unicode code points are | |
271 | * problematic in some contexts. This allows code that needs to check for | |
a3815e44 | 272 | * those to quickly exclude the vast majority of code points it will |
fed423a5 | 273 | * encounter */ |
a6951642 KW |
274 | #define isUTF8_POSSIBLY_PROBLEMATIC(c) (__ASSERT_(FITS_IN_8_BITS(c)) \ |
275 | (U8) c >= 0xED) | |
fed423a5 KW |
276 | |
277 | #endif /* EBCDIC vs ASCII */ | |
278 | ||
28ca3ab5 KW |
279 | /* It turns out that in a number of cases, that handling ASCII vs EBCDIC is a |
280 | * matter of being off-by-one. So this is a convenience macro, used to avoid | |
281 | * some #ifdefs. */ | |
282 | #define ONE_IF_EBCDIC_ZERO_IF_NOT \ | |
283 | (UTF_CONTINUATION_BYTE_INFO_BITS == UTF_EBCDIC_CONTINUATION_BYTE_INFO_BITS) | |
284 | ||
36da1e17 KW |
285 | /* Since the significant bits in a continuation byte are stored in the |
286 | * least-significant positions, we often find ourselves shifting by that | |
287 | * amount. This is a clearer name in such situations */ | |
288 | #define UTF_ACCUMULATION_SHIFT UTF_CONTINUATION_BYTE_INFO_BITS | |
289 | ||
290 | /* 2**info_bits - 1. This masks out all but the bits that carry | |
d7bcd45a KW |
291 | * real information in a continuation byte. This turns out to be 0x3F in |
292 | * UTF-8, 0x1F in UTF-EBCDIC. */ | |
36da1e17 KW |
293 | #define UTF_CONTINUATION_MASK \ |
294 | ((U8) nBIT_MASK(UTF_CONTINUATION_BYTE_INFO_BITS)) | |
fed423a5 | 295 | |
9f3cfb7a KW |
296 | /* For use in UTF8_IS_CONTINUATION(). This turns out to be 0xC0 in UTF-8, |
297 | * E0 in UTF-EBCDIC */ | |
298 | #define UTF_IS_CONTINUATION_MASK ((U8) (0xFF << UTF_ACCUMULATION_SHIFT)) | |
299 | ||
38f458ff KW |
300 | /* This defines the bits that are to be in the continuation bytes of a |
301 | * multi-byte UTF-8 encoded character that mark it is a continuation byte. | |
302 | * This turns out to be 0x80 in UTF-8, 0xA0 in UTF-EBCDIC. (khw doesn't know | |
fcd03d92 KW |
303 | * the underlying reason that B0 works here, except it just happens to work. |
304 | * One could solve for two linear equations and come up with it.) */ | |
38f458ff KW |
305 | #define UTF_CONTINUATION_MARK (UTF_IS_CONTINUATION_MASK & 0xB0) |
306 | ||
f4225fa0 KW |
307 | /* Is the byte 'c' part of a multi-byte UTF8-8 encoded sequence, and not the |
308 | * first byte thereof? */ | |
309 | #define UTF8_IS_CONTINUATION(c) (__ASSERT_(FITS_IN_8_BITS(c)) \ | |
310 | (((NATIVE_UTF8_TO_I8(c) & UTF_IS_CONTINUATION_MASK) \ | |
311 | == UTF_CONTINUATION_MARK))) | |
312 | ||
2dc97505 KW |
313 | /* Is the representation of the Unicode code point 'cp' the same regardless of |
314 | * being encoded in UTF-8 or not? This is a fundamental property of | |
315 | * UTF-8,EBCDIC */ | |
316 | #define OFFUNI_IS_INVARIANT(c) (((WIDEST_UTYPE)(c)) < UTF_CONTINUATION_MARK) | |
317 | ||
ab2e28c2 KW |
318 | /* |
319 | =for apidoc Am|bool|UVCHR_IS_INVARIANT|UV cp | |
320 | ||
321 | Evaluates to 1 if the representation of code point C<cp> is the same whether or | |
322 | not it is encoded in UTF-8; otherwise evaluates to 0. UTF-8 invariant | |
323 | characters can be copied as-is when converting to/from UTF-8, saving time. | |
324 | C<cp> is Unicode if above 255; otherwise is platform-native. | |
325 | ||
326 | =cut | |
327 | */ | |
328 | #define UVCHR_IS_INVARIANT(cp) (OFFUNI_IS_INVARIANT(NATIVE_TO_UNI(cp))) | |
329 | ||
c26e6896 KW |
330 | /* This defines the 1-bits that are to be in the first byte of a multi-byte |
331 | * UTF-8 encoded character that mark it as a start byte and give the number of | |
332 | * bytes that comprise the character. 'len' is that number. | |
333 | * | |
334 | * To illustrate: len = 2 => ((U8) ~ 0b0011_1111) or 1100_0000 | |
335 | * 7 => ((U8) ~ 0b0000_0001) or 1111_1110 | |
336 | * > 7 => 0xFF | |
337 | * | |
338 | * This is not to be used on a single-byte character. As in many places in | |
339 | * perl, U8 must be 8 bits | |
340 | */ | |
341 | #define UTF_START_MARK(len) ((U8) ~(0xFF >> (len))) | |
342 | ||
0176a72c KW |
343 | /* Masks out the initial one bits in a start byte, leaving the following 0 bit |
344 | * and the real data bits. 'len' is the number of bytes in the multi-byte | |
345 | * sequence that comprises the character. | |
346 | * | |
347 | * To illustrate: len = 2 => 0b0011_1111 works on start byte 110xxxxx | |
348 | * 6 => 0b0000_0011 works on start byte 1111110x | |
349 | * >= 7 => There are no data bits in the start byte | |
350 | * Note that on ASCII platforms, this can be passed a len=1 byte; and all the | |
351 | * real data bits will be returned: | |
352 | len = 1 => 0b0111_1111 | |
353 | * This isn't true on EBCDIC platforms, where some len=1 bytes are of the form | |
354 | * 0b101x_xxxx, so this can't be used there on single-byte characters. */ | |
355 | #define UTF_START_MASK(len) (0xFF >> (len)) | |
356 | ||
7028aeba KW |
357 | /* Internal macro to be used only in this file to aid in constructing other |
358 | * publicly accessible macros. | |
359 | * The number of bytes required to express this uv in UTF-8, for just those | |
360 | * uv's requiring 2 through 6 bytes, as these are common to all platforms and | |
361 | * word sizes. The number of bytes needed is given by the number of leading 1 | |
362 | * bits in the start byte. There are 32 start bytes that have 2 initial 1 bits | |
363 | * (C0-DF); there are 16 that have 3 initial 1 bits (E0-EF); 8 that have 4 | |
364 | * initial 1 bits (F0-F8); 4 that have 5 initial 1 bits (F9-FB), and 2 that | |
365 | * have 6 initial 1 bits (FC-FD). The largest number a string of n bytes can | |
366 | * represent is (the number of possible start bytes for 'n') | |
367 | * * (the number of possiblities for each start byte | |
368 | * The latter in turn is | |
369 | * 2 ** ( (how many continuation bytes there are) | |
370 | * * (the number of bits of information each | |
371 | * continuation byte holds)) | |
372 | * | |
373 | * If we were on a platform where we could use a fast find first set bit | |
374 | * instruction (or count leading zeros instruction) this could be replaced by | |
375 | * using that to find the log2 of the uv, and divide that by the number of bits | |
376 | * of information in each continuation byte, adjusting for large cases and how | |
377 | * much information is in a start byte for that length */ | |
72164d3a | 378 | #define __COMMON_UNI_SKIP(uv) \ |
7028aeba KW |
379 | (UV) (uv) < (32 * (1U << ( UTF_ACCUMULATION_SHIFT))) ? 2 : \ |
380 | (UV) (uv) < (16 * (1U << (2 * UTF_ACCUMULATION_SHIFT))) ? 3 : \ | |
381 | (UV) (uv) < ( 8 * (1U << (3 * UTF_ACCUMULATION_SHIFT))) ? 4 : \ | |
382 | (UV) (uv) < ( 4 * (1U << (4 * UTF_ACCUMULATION_SHIFT))) ? 5 : \ | |
383 | (UV) (uv) < ( 2 * (1U << (5 * UTF_ACCUMULATION_SHIFT))) ? 6 : | |
72164d3a KW |
384 | |
385 | /* Internal macro to be used only in this file. | |
386 | * This adds to __COMMON_UNI_SKIP the details at this platform's upper range. | |
fed423a5 | 387 | * For any-sized EBCDIC platforms, or 64-bit ASCII ones, we need one more test |
72164d3a KW |
388 | * to see if just 7 bytes is needed, or if the maximum is needed. For 32-bit |
389 | * ASCII platforms, everything is representable by 7 bytes */ | |
fed423a5 | 390 | #if defined(UV_IS_QUAD) || defined(EBCDIC) |
72164d3a | 391 | # define __BASE_UNI_SKIP(uv) (__COMMON_UNI_SKIP(uv) \ |
8974941d KW |
392 | LIKELY((UV) (uv) < ((UV) 1U << (6 * UTF_ACCUMULATION_SHIFT))) \ |
393 | ? 7 \ | |
394 | : UTF8_MAXBYTES) | |
1d68d6cd | 395 | #else |
72164d3a | 396 | # define __BASE_UNI_SKIP(uv) (__COMMON_UNI_SKIP(uv) 7) |
1d68d6cd SC |
397 | #endif |
398 | ||
fed423a5 KW |
399 | /* The next two macros use the base macro defined above, and add in the tests |
400 | * at the low-end of the range, for just 1 byte, yielding complete macros, | |
401 | * publicly accessible. */ | |
402 | ||
403 | /* Input is a true Unicode (not-native) code point */ | |
404 | #define OFFUNISKIP(uv) (OFFUNI_IS_INVARIANT(uv) ? 1 : __BASE_UNI_SKIP(uv)) | |
2084b489 | 405 | |
5352a763 KW |
406 | /* |
407 | ||
408 | =for apidoc Am|STRLEN|UVCHR_SKIP|UV cp | |
409 | returns the number of bytes required to represent the code point C<cp> when | |
410 | encoded as UTF-8. C<cp> is a native (ASCII or EBCDIC) code point if less than | |
411 | 255; a Unicode code point otherwise. | |
412 | ||
413 | =cut | |
414 | */ | |
fdb6583d | 415 | #define UVCHR_SKIP(uv) ( UVCHR_IS_INVARIANT(uv) ? 1 : __BASE_UNI_SKIP(uv)) |
5352a763 | 416 | |
4bab39bc KW |
417 | #define UTF_MIN_START_BYTE \ |
418 | ((UTF_CONTINUATION_MARK >> UTF_ACCUMULATION_SHIFT) | UTF_START_MARK(2)) | |
419 | ||
420 | /* Is the byte 'c' the first byte of a multi-byte UTF8-8 encoded sequence? | |
59645eb1 | 421 | * This excludes invariants (they are single-byte). It also excludes the |
4bab39bc | 422 | * illegal overlong sequences that begin with C0 and C1 on ASCII platforms, and |
59645eb1 KW |
423 | * C0-C4 I8 start bytes on EBCDIC ones. On EBCDIC E0 can't start a |
424 | * non-overlong sequence, so we define a base macro and for those platforms, | |
425 | * extend it to also exclude E0 */ | |
426 | #define UTF8_IS_START_base(c) (__ASSERT_(FITS_IN_8_BITS(c)) \ | |
4bab39bc | 427 | (NATIVE_UTF8_TO_I8(c) >= UTF_MIN_START_BYTE)) |
59645eb1 KW |
428 | #ifdef EBCDIC |
429 | # define UTF8_IS_START(c) \ | |
430 | (UTF8_IS_START_base(c) && (c) != I8_TO_NATIVE_UTF8(0xE0)) | |
431 | #else | |
432 | # define UTF8_IS_START(c) UTF8_IS_START_base(c) | |
433 | #endif | |
4bab39bc | 434 | |
1df63428 KW |
435 | #define UTF_MIN_ABOVE_LATIN1_BYTE \ |
436 | ((0x100 >> UTF_ACCUMULATION_SHIFT) | UTF_START_MARK(2)) | |
437 | ||
438 | /* Is the UTF8-encoded byte 'c' the first byte of a sequence of bytes that | |
439 | * represent a code point > 255? */ | |
440 | #define UTF8_IS_ABOVE_LATIN1(c) (__ASSERT_(FITS_IN_8_BITS(c)) \ | |
441 | (NATIVE_UTF8_TO_I8(c) >= UTF_MIN_ABOVE_LATIN1_BYTE)) | |
442 | ||
7c88d61e KW |
443 | /* Is the UTF8-encoded byte 'c' the first byte of a two byte sequence? Use |
444 | * UTF8_IS_NEXT_CHAR_DOWNGRADEABLE() instead if the input isn't known to | |
445 | * be well-formed. */ | |
446 | #define UTF8_IS_DOWNGRADEABLE_START(c) (__ASSERT_(FITS_IN_8_BITS(c)) \ | |
03dc0b1b | 447 | inRANGE_helper_(U8, NATIVE_UTF8_TO_I8(c), \ |
7c88d61e KW |
448 | UTF_MIN_START_BYTE, UTF_MIN_ABOVE_LATIN1_BYTE - 1)) |
449 | ||
b651802e KW |
450 | /* The largest code point representable by two UTF-8 bytes on this platform. |
451 | * As explained in the comments for __COMMON_UNI_SKIP, 32 start bytes with | |
fed423a5 | 452 | * UTF_ACCUMULATION_SHIFT bits of information each */ |
aa206fb7 KW |
453 | #define MAX_UTF8_TWO_BYTE (32 * (1U << UTF_ACCUMULATION_SHIFT) - 1) |
454 | ||
b651802e | 455 | /* The largest code point representable by two UTF-8 bytes on any platform that |
fcd03d92 KW |
456 | * Perl runs on. */ |
457 | #define MAX_PORTABLE_UTF8_TWO_BYTE \ | |
458 | nBIT_UMAX(5 + MIN( UTF_CONTINUATION_BYTE_INFO_BITS, \ | |
459 | UTF_EBCDIC_CONTINUATION_BYTE_INFO_BITS)) | |
aa206fb7 | 460 | |
f2c50040 KW |
461 | /* |
462 | ||
463 | =for apidoc AmnU|STRLEN|UTF8_MAXBYTES_CASE | |
464 | ||
465 | The maximum number of UTF-8 bytes a single Unicode character can | |
466 | uppercase/lowercase/titlecase/fold into. | |
467 | ||
468 | =cut | |
469 | ||
470 | * Unicode guarantees that the maximum expansion is UTF8_MAX_FOLD_CHAR_EXPAND | |
471 | * characters, but any above-Unicode code point will fold to itself, so we only | |
472 | * have to look at the expansion of the maximum Unicode code point. But this | |
473 | * number may be less than the space occupied by a very large code point under | |
474 | * Perl's extended UTF-8. We have to make it large enough to fit any single | |
475 | * character. (It turns out that ASCII and EBCDIC differ in which is larger) | |
476 | * | |
477 | =cut | |
478 | */ | |
c03c0950 | 479 | #define UTF8_MAXBYTES_CASE \ |
ae9a9edb | 480 | MAX(UTF8_MAXBYTES, UTF8_MAX_FOLD_CHAR_EXPAND * OFFUNISKIP(0x10FFFF)) |
c03c0950 | 481 | |
d06134e5 KW |
482 | /* Rest of these are attributes of Unicode and perl's internals rather than the |
483 | * encoding, or happen to be the same in both ASCII and EBCDIC (at least at | |
484 | * this level; the macros that some of these call may have different | |
485 | * definitions in the two encodings */ | |
486 | ||
59a449d5 KW |
487 | /* In domain restricted to ASCII, these may make more sense to the reader than |
488 | * the ones with Latin1 in the name */ | |
489 | #define NATIVE_TO_ASCII(ch) NATIVE_TO_LATIN1(ch) | |
490 | #define ASCII_TO_NATIVE(ch) LATIN1_TO_NATIVE(ch) | |
491 | ||
492 | /* More or less misleadingly-named defines, retained for back compat */ | |
493 | #define NATIVE_TO_UTF(ch) NATIVE_UTF8_TO_I8(ch) | |
494 | #define NATIVE_TO_I8(ch) NATIVE_UTF8_TO_I8(ch) | |
495 | #define UTF_TO_NATIVE(ch) I8_TO_NATIVE_UTF8(ch) | |
496 | #define I8_TO_NATIVE(ch) I8_TO_NATIVE_UTF8(ch) | |
497 | #define NATIVE8_TO_UNI(ch) NATIVE_TO_LATIN1(ch) | |
d06134e5 | 498 | |
537124e4 KW |
499 | /* Adds a UTF8 continuation byte 'new' of information to a running total code |
500 | * point 'old' of all the continuation bytes so far. This is designed to be | |
155d2738 KW |
501 | * used in a loop to convert from UTF-8 to the code point represented. Note |
502 | * that this is asymmetric on EBCDIC platforms, in that the 'new' parameter is | |
503 | * the UTF-EBCDIC byte, whereas the 'old' parameter is a Unicode (not EBCDIC) | |
504 | * code point in process of being generated */ | |
a6951642 KW |
505 | #define UTF8_ACCUMULATE(old, new) (__ASSERT_(FITS_IN_8_BITS(new)) \ |
506 | ((old) << UTF_ACCUMULATION_SHIFT) \ | |
009097b1 | 507 | | ((NATIVE_UTF8_TO_I8(new)) \ |
155d2738 | 508 | & UTF_CONTINUATION_MASK)) |
d06134e5 | 509 | |
4ab10950 | 510 | /* This works in the face of malformed UTF-8. */ |
4e1ed312 | 511 | #define UTF8_IS_NEXT_CHAR_DOWNGRADEABLE(s, e) \ |
0b47b392 KW |
512 | ( ( (e) - (s) > 1) \ |
513 | && UTF8_IS_DOWNGRADEABLE_START(*(s)) \ | |
4e1ed312 | 514 | && UTF8_IS_CONTINUATION(*((s)+1))) |
4ab10950 | 515 | |
5aaebcb3 | 516 | /* Number of bytes a code point occupies in UTF-8. */ |
5352a763 | 517 | #define NATIVE_SKIP(uv) UVCHR_SKIP(uv) |
bd18bd40 | 518 | |
5aaebcb3 KW |
519 | /* Most code which says UNISKIP is really thinking in terms of native code |
520 | * points (0-255) plus all those beyond. This is an imprecise term, but having | |
2accb712 | 521 | * it means existing code continues to work. For precision, use UVCHR_SKIP, |
5352a763 KW |
522 | * NATIVE_SKIP, or OFFUNISKIP */ |
523 | #define UNISKIP(uv) UVCHR_SKIP(uv) | |
5aaebcb3 | 524 | |
3c0792e4 KW |
525 | /* Longer, but more accurate name */ |
526 | #define UTF8_IS_ABOVE_LATIN1_START(c) UTF8_IS_ABOVE_LATIN1(c) | |
527 | ||
a62b247b KW |
528 | /* Convert a UTF-8 variant Latin1 character to a native code point value. |
529 | * Needs just one iteration of accumulate. Should be used only if it is known | |
530 | * that the code point is < 256, and is not UTF-8 invariant. Use the slower | |
531 | * but more general TWO_BYTE_UTF8_TO_NATIVE() which handles any code point | |
532 | * representable by two bytes (which turns out to be up through | |
533 | * MAX_PORTABLE_UTF8_TWO_BYTE). The two parameters are: | |
534 | * HI: a downgradable start byte; | |
535 | * LO: continuation. | |
536 | * */ | |
537 | #define EIGHT_BIT_UTF8_TO_NATIVE(HI, LO) \ | |
538 | ( __ASSERT_(UTF8_IS_DOWNGRADEABLE_START(HI)) \ | |
539 | __ASSERT_(UTF8_IS_CONTINUATION(LO)) \ | |
540 | LATIN1_TO_NATIVE(UTF8_ACCUMULATE(( \ | |
541 | NATIVE_UTF8_TO_I8(HI) & UTF_START_MASK(2)), (LO)))) | |
542 | ||
94bb8c36 | 543 | /* Convert a two (not one) byte utf8 character to a native code point value. |
2950f2a7 KW |
544 | * Needs just one iteration of accumulate. Should not be used unless it is |
545 | * known that the two bytes are legal: 1) two-byte start, and 2) continuation. | |
546 | * Note that the result can be larger than 255 if the input character is not | |
547 | * downgradable */ | |
94bb8c36 | 548 | #define TWO_BYTE_UTF8_TO_NATIVE(HI, LO) \ |
a6951642 KW |
549 | (__ASSERT_(FITS_IN_8_BITS(HI)) \ |
550 | __ASSERT_(FITS_IN_8_BITS(LO)) \ | |
551 | __ASSERT_(PL_utf8skip[HI] == 2) \ | |
552 | __ASSERT_(UTF8_IS_CONTINUATION(LO)) \ | |
94bb8c36 | 553 | UNI_TO_NATIVE(UTF8_ACCUMULATE((NATIVE_UTF8_TO_I8(HI) & UTF_START_MASK(2)), \ |
635e76f5 | 554 | (LO)))) |
94bb8c36 KW |
555 | |
556 | /* Should never be used, and be deprecated */ | |
557 | #define TWO_BYTE_UTF8_TO_UNI(HI, LO) NATIVE_TO_UNI(TWO_BYTE_UTF8_TO_NATIVE(HI, LO)) | |
2950f2a7 | 558 | |
bd18bd40 KW |
559 | /* |
560 | ||
561 | =for apidoc Am|STRLEN|UTF8SKIP|char* s | |
ee0ff0f5 KW |
562 | returns the number of bytes a non-malformed UTF-8 encoded character whose first |
563 | (perhaps only) byte is pointed to by C<s>. | |
564 | ||
565 | If there is a possibility of malformed input, use instead: | |
566 | ||
567 | =over | |
568 | ||
eb992c6f | 569 | =item C<L</UTF8_SAFE_SKIP>> if you know the maximum ending pointer in the |
ee0ff0f5 KW |
570 | buffer pointed to by C<s>; or |
571 | ||
eb992c6f | 572 | =item C<L</UTF8_CHK_SKIP>> if you don't know it. |
ee0ff0f5 KW |
573 | |
574 | =back | |
575 | ||
576 | It is better to restructure your code so the end pointer is passed down so that | |
577 | you know what it actually is at the point of this call, but if that isn't | |
eb992c6f | 578 | possible, C<L</UTF8_CHK_SKIP>> can minimize the chance of accessing beyond the end |
ee0ff0f5 | 579 | of the input buffer. |
bd18bd40 KW |
580 | |
581 | =cut | |
582 | */ | |
2a70536e | 583 | #define UTF8SKIP(s) PL_utf8skip[*(const U8*)(s)] |
a281f16c KW |
584 | |
585 | /* | |
586 | =for apidoc Am|STRLEN|UTF8_SKIP|char* s | |
eb992c6f | 587 | This is a synonym for C<L</UTF8SKIP>> |
a281f16c KW |
588 | |
589 | =cut | |
590 | */ | |
591 | ||
2a70536e | 592 | #define UTF8_SKIP(s) UTF8SKIP(s) |
d06134e5 | 593 | |
85fcc8f2 | 594 | /* |
ee0ff0f5 KW |
595 | =for apidoc Am|STRLEN|UTF8_CHK_SKIP|char* s |
596 | ||
eb992c6f KW |
597 | This is a safer version of C<L</UTF8SKIP>>, but still not as safe as |
598 | C<L</UTF8_SAFE_SKIP>>. This version doesn't blindly assume that the input | |
ee0ff0f5 KW |
599 | string pointed to by C<s> is well-formed, but verifies that there isn't a NUL |
600 | terminating character before the expected end of the next character in C<s>. | |
601 | The length C<UTF8_CHK_SKIP> returns stops just before any such NUL. | |
602 | ||
603 | Perl tends to add NULs, as an insurance policy, after the end of strings in | |
604 | SV's, so it is likely that using this macro will prevent inadvertent reading | |
605 | beyond the end of the input buffer, even if it is malformed UTF-8. | |
606 | ||
607 | This macro is intended to be used by XS modules where the inputs could be | |
608 | malformed, and it isn't feasible to restructure to use the safer | |
eb992c6f | 609 | C<L</UTF8_SAFE_SKIP>>, for example when interfacing with a C library. |
ee0ff0f5 KW |
610 | |
611 | =cut | |
612 | */ | |
613 | ||
614 | #define UTF8_CHK_SKIP(s) \ | |
8974941d | 615 | (UNLIKELY(s[0] == '\0') ? 1 : MIN(UTF8SKIP(s), \ |
f87d8789 | 616 | my_strnlen((char *) (s), UTF8SKIP(s)))) |
ee0ff0f5 | 617 | /* |
85fcc8f2 KW |
618 | |
619 | =for apidoc Am|STRLEN|UTF8_SAFE_SKIP|char* s|char* e | |
45671da2 KW |
620 | returns 0 if S<C<s E<gt>= e>>; otherwise returns the number of bytes in the |
621 | UTF-8 encoded character whose first byte is pointed to by C<s>. But it never | |
622 | returns beyond C<e>. On DEBUGGING builds, it asserts that S<C<s E<lt>= e>>. | |
85fcc8f2 KW |
623 | |
624 | =cut | |
625 | */ | |
45671da2 | 626 | #define UTF8_SAFE_SKIP(s, e) (__ASSERT_((e) >= (s)) \ |
8974941d | 627 | UNLIKELY(((e) - (s)) <= 0) \ |
45671da2 KW |
628 | ? 0 \ |
629 | : MIN(((e) - (s)), UTF8_SKIP(s))) | |
85fcc8f2 | 630 | |
2d1545e5 KW |
631 | /* Most code that says 'UNI_' really means the native value for code points up |
632 | * through 255 */ | |
633 | #define UNI_IS_INVARIANT(cp) UVCHR_IS_INVARIANT(cp) | |
634 | ||
c2b32798 KW |
635 | /* |
636 | =for apidoc Am|bool|UTF8_IS_INVARIANT|char c | |
637 | ||
638 | Evaluates to 1 if the byte C<c> represents the same character when encoded in | |
639 | UTF-8 as when not; otherwise evaluates to 0. UTF-8 invariant characters can be | |
640 | copied as-is when converting to/from UTF-8, saving time. | |
641 | ||
642 | In spite of the name, this macro gives the correct result if the input string | |
643 | from which C<c> comes is not encoded in UTF-8. | |
644 | ||
645 | See C<L</UVCHR_IS_INVARIANT>> for checking if a UV is invariant. | |
646 | ||
647 | =cut | |
648 | ||
649 | The reason it works on both UTF-8 encoded strings and non-UTF-8 encoded, is | |
650 | that it returns TRUE in each for the exact same set of bit patterns. It is | |
651 | valid on a subset of what UVCHR_IS_INVARIANT is valid on, so can just use that; | |
652 | and the compiler should optimize out anything extraneous given the | |
653 | implementation of the latter. The |0 makes sure this isn't mistakenly called | |
654 | with a ptr argument. | |
655 | */ | |
5c06326b | 656 | #define UTF8_IS_INVARIANT(c) UVCHR_IS_INVARIANT((c) | 0) |
5fc230f1 KW |
657 | |
658 | /* Like the above, but its name implies a non-UTF8 input, which as the comments | |
659 | * above show, doesn't matter as to its implementation */ | |
38953e5a | 660 | #define NATIVE_BYTE_IS_INVARIANT(c) UVCHR_IS_INVARIANT(c) |
d06134e5 | 661 | |
2c03e801 KW |
662 | /* Misleadingly named: is the UTF8-encoded byte 'c' part of a variant sequence |
663 | * in UTF-8? This is the inverse of UTF8_IS_INVARIANT. */ | |
664 | #define UTF8_IS_CONTINUED(c) (__ASSERT_(FITS_IN_8_BITS(c)) \ | |
665 | (! UTF8_IS_INVARIANT(c))) | |
666 | ||
48ccf5e1 KW |
667 | /* The macros in the next 4 sets are used to generate the two utf8 or utfebcdic |
668 | * bytes from an ordinal that is known to fit into exactly two (not one) bytes; | |
669 | * it must be less than 0x3FF to work across both encodings. */ | |
670 | ||
671 | /* These two are helper macros for the other three sets, and should not be used | |
672 | * directly anywhere else. 'translate_function' is either NATIVE_TO_LATIN1 | |
1ff3baa2 KW |
673 | * (which works for code points up through 0xFF) or NATIVE_TO_UNI which works |
674 | * for any code point */ | |
48ccf5e1 | 675 | #define __BASE_TWO_BYTE_HI(c, translate_function) \ |
2863dafa | 676 | (__ASSERT_(! UVCHR_IS_INVARIANT(c)) \ |
48ccf5e1 | 677 | I8_TO_NATIVE_UTF8((translate_function(c) >> UTF_ACCUMULATION_SHIFT) \ |
2863dafa | 678 | | UTF_START_MARK(2))) |
48ccf5e1 | 679 | #define __BASE_TWO_BYTE_LO(c, translate_function) \ |
2863dafa | 680 | (__ASSERT_(! UVCHR_IS_INVARIANT(c)) \ |
48ccf5e1 | 681 | I8_TO_NATIVE_UTF8((translate_function(c) & UTF_CONTINUATION_MASK) \ |
2863dafa | 682 | | UTF_CONTINUATION_MARK)) |
48ccf5e1 | 683 | |
48ccf5e1 KW |
684 | /* The next two macros should not be used. They were designed to be usable as |
685 | * the case label of a switch statement, but this doesn't work for EBCDIC. Use | |
9d0d3a03 | 686 | * regen/unicode_constants.pl instead */ |
48ccf5e1 KW |
687 | #define UTF8_TWO_BYTE_HI_nocast(c) __BASE_TWO_BYTE_HI(c, NATIVE_TO_UNI) |
688 | #define UTF8_TWO_BYTE_LO_nocast(c) __BASE_TWO_BYTE_LO(c, NATIVE_TO_UNI) | |
689 | ||
690 | /* The next two macros are used when the source should be a single byte | |
691 | * character; checked for under DEBUGGING */ | |
692 | #define UTF8_EIGHT_BIT_HI(c) (__ASSERT_(FITS_IN_8_BITS(c)) \ | |
4c8cd605 | 693 | ( __BASE_TWO_BYTE_HI(c, NATIVE_TO_LATIN1))) |
48ccf5e1 | 694 | #define UTF8_EIGHT_BIT_LO(c) (__ASSERT_(FITS_IN_8_BITS(c)) \ |
4c8cd605 | 695 | (__BASE_TWO_BYTE_LO(c, NATIVE_TO_LATIN1))) |
48ccf5e1 KW |
696 | |
697 | /* These final two macros in the series are used when the source can be any | |
698 | * code point whose UTF-8 is known to occupy 2 bytes; they are less efficient | |
699 | * than the EIGHT_BIT versions on EBCDIC platforms. We use the logical '~' | |
700 | * operator instead of "<=" to avoid getting compiler warnings. | |
d52b8576 | 701 | * MAX_UTF8_TWO_BYTE should be exactly all one bits in the lower few |
48ccf5e1 KW |
702 | * places, so the ~ works */ |
703 | #define UTF8_TWO_BYTE_HI(c) \ | |
704 | (__ASSERT_((sizeof(c) == 1) \ | |
d52b8576 | 705 | || !(((WIDEST_UTYPE)(c)) & ~MAX_UTF8_TWO_BYTE)) \ |
4c8cd605 | 706 | (__BASE_TWO_BYTE_HI(c, NATIVE_TO_UNI))) |
48ccf5e1 KW |
707 | #define UTF8_TWO_BYTE_LO(c) \ |
708 | (__ASSERT_((sizeof(c) == 1) \ | |
d52b8576 | 709 | || !(((WIDEST_UTYPE)(c)) & ~MAX_UTF8_TWO_BYTE)) \ |
4c8cd605 | 710 | (__BASE_TWO_BYTE_LO(c, NATIVE_TO_UNI))) |
d06134e5 | 711 | |
e7214ce8 KW |
712 | /* This is illegal in any well-formed UTF-8 in both EBCDIC and ASCII |
713 | * as it is only in overlongs. */ | |
714 | #define ILLEGAL_UTF8_BYTE I8_TO_NATIVE_UTF8(0xC1) | |
715 | ||
7e2040f0 | 716 | /* |
e3036cf4 | 717 | * 'UTF' is whether or not p is encoded in UTF8. The names 'foo_lazy_if' stem |
20df05f4 KW |
718 | * from an earlier version of these macros in which they didn't call the |
719 | * foo_utf8() macros (i.e. were 'lazy') unless they decided that *p is the | |
720 | * beginning of a utf8 character. Now that foo_utf8() determines that itself, | |
721 | * no need to do it again here | |
7e2040f0 | 722 | */ |
da8c1a98 KW |
723 | #define isIDFIRST_lazy_if_safe(p, e, UTF) \ |
724 | ((IN_BYTES || !UTF) \ | |
725 | ? isIDFIRST(*(p)) \ | |
726 | : isIDFIRST_utf8_safe(p, e)) | |
da8c1a98 KW |
727 | #define isWORDCHAR_lazy_if_safe(p, e, UTF) \ |
728 | ((IN_BYTES || !UTF) \ | |
729 | ? isWORDCHAR(*(p)) \ | |
730 | : isWORDCHAR_utf8_safe((U8 *) p, (U8 *) e)) | |
4c1d9526 | 731 | #define isALNUM_lazy_if_safe(p, e, UTF) isWORDCHAR_lazy_if_safe(p, e, UTF) |
da8c1a98 | 732 | |
89ebb4a3 JH |
733 | #define UTF8_MAXLEN UTF8_MAXBYTES |
734 | ||
8cb75cc8 KW |
735 | /* A Unicode character can fold to up to 3 characters */ |
736 | #define UTF8_MAX_FOLD_CHAR_EXPAND 3 | |
737 | ||
d3481830 | 738 | #define IN_BYTES UNLIKELY(CopHINTS_get(PL_curcop) & HINT_BYTES) |
bd18bd40 KW |
739 | |
740 | /* | |
741 | ||
742 | =for apidoc Am|bool|DO_UTF8|SV* sv | |
743 | Returns a bool giving whether or not the PV in C<sv> is to be treated as being | |
744 | encoded in UTF-8. | |
745 | ||
746 | You should use this I<after> a call to C<SvPV()> or one of its variants, in | |
747 | case any call to string overloading updates the internal UTF-8 encoding flag. | |
748 | ||
749 | =cut | |
750 | */ | |
0064a8a9 | 751 | #define DO_UTF8(sv) (SvUTF8(sv) && !IN_BYTES) |
1ff3baa2 KW |
752 | |
753 | /* Should all strings be treated as Unicode, and not just UTF-8 encoded ones? | |
754 | * Is so within 'feature unicode_strings' or 'locale :not_characters', and not | |
755 | * within 'use bytes'. UTF-8 locales are not tested for here, but perhaps | |
756 | * could be */ | |
70844984 | 757 | #define IN_UNI_8_BIT \ |
1604cfb0 | 758 | (( ( (CopHINTS_get(PL_curcop) & HINT_UNI_8_BIT)) \ |
70844984 KW |
759 | || ( CopHINTS_get(PL_curcop) & HINT_LOCALE_PARTIAL \ |
760 | /* -1 below is for :not_characters */ \ | |
761 | && _is_in_locale_category(FALSE, -1))) \ | |
762 | && (! IN_BYTES)) | |
b36bf33f | 763 | |
1d72bdf6 | 764 | |
99904f65 KW |
765 | /* Perl extends Unicode so that it is possible to encode (as extended UTF-8 or |
766 | * UTF-EBCDIC) any 64-bit value. No standard known to khw ever encoded higher | |
767 | * than a 31 bit value. On ASCII platforms this just meant arbitrarily saying | |
768 | * nothing could be higher than this. On these the start byte FD gets you to | |
769 | * 31 bits, and FE and FF are forbidden as start bytes. On EBCDIC platforms, | |
770 | * FD gets you only to 26 bits; adding FE to mean 7 total bytes gets you to 30 | |
771 | * bits. To get to 31 bits, they treated an initial FF byte idiosyncratically. | |
772 | * It was considered to be the start byte FE meaning it had 7 total bytes, and | |
773 | * the final 1 was treated as an information bit, getting you to 31 bits. | |
774 | * | |
775 | * Perl used to accept this idiosyncratic interpretation of FF, but now rejects | |
776 | * it in order to get to being able to encode 64 bits. The bottom line is that | |
777 | * it is a Perl extension to use the start bytes FE and FF on ASCII platforms, | |
778 | * and the start byte FF on EBCDIC ones. That translates into that it is a | |
779 | * Perl extension to represent anything occupying more than 31 bits on ASCII | |
780 | * platforms; 30 bits on EBCDIC. */ | |
781 | #define UNICODE_IS_PERL_EXTENDED(uv) \ | |
782 | UNLIKELY((UV) (uv) > nBIT_UMAX(31 - ONE_IF_EBCDIC_ZERO_IF_NOT)) | |
43732c4f KW |
783 | #define UTF8_IS_PERL_EXTENDED(s) \ |
784 | (UTF8SKIP(s) > 6 + ONE_IF_EBCDIC_ZERO_IF_NOT) | |
99904f65 | 785 | |
c76687c5 | 786 | #define UTF8_ALLOW_EMPTY 0x0001 /* Allow a zero length string */ |
2b5e7bc2 | 787 | #define UTF8_GOT_EMPTY UTF8_ALLOW_EMPTY |
c76687c5 KW |
788 | |
789 | /* Allow first byte to be a continuation byte */ | |
1d72bdf6 | 790 | #define UTF8_ALLOW_CONTINUATION 0x0002 |
2b5e7bc2 | 791 | #define UTF8_GOT_CONTINUATION UTF8_ALLOW_CONTINUATION |
c76687c5 | 792 | |
cd01d3b1 | 793 | /* Unexpected non-continuation byte */ |
1d72bdf6 | 794 | #define UTF8_ALLOW_NON_CONTINUATION 0x0004 |
2b5e7bc2 | 795 | #define UTF8_GOT_NON_CONTINUATION UTF8_ALLOW_NON_CONTINUATION |
949cf498 KW |
796 | |
797 | /* expecting more bytes than were available in the string */ | |
798 | #define UTF8_ALLOW_SHORT 0x0008 | |
2b5e7bc2 | 799 | #define UTF8_GOT_SHORT UTF8_ALLOW_SHORT |
949cf498 | 800 | |
94953955 KW |
801 | /* Overlong sequence; i.e., the code point can be specified in fewer bytes. |
802 | * First one will convert the overlong to the REPLACEMENT CHARACTER; second | |
803 | * will return what the overlong evaluates to */ | |
949cf498 | 804 | #define UTF8_ALLOW_LONG 0x0010 |
94953955 | 805 | #define UTF8_ALLOW_LONG_AND_ITS_VALUE (UTF8_ALLOW_LONG|0x0020) |
2b5e7bc2 KW |
806 | #define UTF8_GOT_LONG UTF8_ALLOW_LONG |
807 | ||
d60baaa7 KW |
808 | #define UTF8_ALLOW_OVERFLOW 0x0080 |
809 | #define UTF8_GOT_OVERFLOW UTF8_ALLOW_OVERFLOW | |
949cf498 | 810 | |
f180b292 | 811 | #define UTF8_DISALLOW_SURROGATE 0x0100 /* Unicode surrogates */ |
2b5e7bc2 | 812 | #define UTF8_GOT_SURROGATE UTF8_DISALLOW_SURROGATE |
f180b292 | 813 | #define UTF8_WARN_SURROGATE 0x0200 |
949cf498 | 814 | |
c4e96019 KW |
815 | /* Unicode non-character code points */ |
816 | #define UTF8_DISALLOW_NONCHAR 0x0400 | |
2b5e7bc2 | 817 | #define UTF8_GOT_NONCHAR UTF8_DISALLOW_NONCHAR |
c4e96019 | 818 | #define UTF8_WARN_NONCHAR 0x0800 |
949cf498 | 819 | |
c4e96019 KW |
820 | /* Super-set of Unicode: code points above the legal max */ |
821 | #define UTF8_DISALLOW_SUPER 0x1000 | |
2b5e7bc2 | 822 | #define UTF8_GOT_SUPER UTF8_DISALLOW_SUPER |
c4e96019 KW |
823 | #define UTF8_WARN_SUPER 0x2000 |
824 | ||
825 | /* The original UTF-8 standard did not define UTF-8 with start bytes of 0xFE or | |
826 | * 0xFF, though UTF-EBCDIC did. This allowed both versions to represent code | |
827 | * points up to 2 ** 31 - 1. Perl extends UTF-8 so that 0xFE and 0xFF are | |
828 | * usable on ASCII platforms, and 0xFF means something different than | |
829 | * UTF-EBCDIC defines. These changes allow code points of 64 bits (actually | |
830 | * somewhat more) to be represented on both platforms. But these are Perl | |
831 | * extensions, and not likely to be interchangeable with other languages. Note | |
832 | * that on ASCII platforms, FE overflows a signed 32-bit word, and FF an | |
833 | * unsigned one. */ | |
d044b7a7 KW |
834 | #define UTF8_DISALLOW_PERL_EXTENDED 0x4000 |
835 | #define UTF8_GOT_PERL_EXTENDED UTF8_DISALLOW_PERL_EXTENDED | |
836 | #define UTF8_WARN_PERL_EXTENDED 0x8000 | |
d35f2ca5 | 837 | |
57ff5f59 KW |
838 | /* For back compat, these old names are misleading for overlongs and |
839 | * UTF_EBCDIC. */ | |
d044b7a7 KW |
840 | #define UTF8_DISALLOW_ABOVE_31_BIT UTF8_DISALLOW_PERL_EXTENDED |
841 | #define UTF8_GOT_ABOVE_31_BIT UTF8_GOT_PERL_EXTENDED | |
842 | #define UTF8_WARN_ABOVE_31_BIT UTF8_WARN_PERL_EXTENDED | |
843 | #define UTF8_DISALLOW_FE_FF UTF8_DISALLOW_PERL_EXTENDED | |
844 | #define UTF8_WARN_FE_FF UTF8_WARN_PERL_EXTENDED | |
949cf498 | 845 | |
f180b292 | 846 | #define UTF8_CHECK_ONLY 0x10000 |
99a765e9 | 847 | #define _UTF8_NO_CONFIDENCE_IN_CURLEN 0x20000 /* Internal core use only */ |
949cf498 KW |
848 | |
849 | /* For backwards source compatibility. They do nothing, as the default now | |
850 | * includes what they used to mean. The first one's meaning was to allow the | |
851 | * just the single non-character 0xFFFF */ | |
852 | #define UTF8_ALLOW_FFFF 0 | |
c825ef8c | 853 | #define UTF8_ALLOW_FE_FF 0 |
949cf498 KW |
854 | #define UTF8_ALLOW_SURROGATE 0 |
855 | ||
ecc1615f KW |
856 | /* C9 refers to Unicode Corrigendum #9: allows but discourages non-chars */ |
857 | #define UTF8_DISALLOW_ILLEGAL_C9_INTERCHANGE \ | |
858 | (UTF8_DISALLOW_SUPER|UTF8_DISALLOW_SURROGATE) | |
859 | #define UTF8_WARN_ILLEGAL_C9_INTERCHANGE (UTF8_WARN_SUPER|UTF8_WARN_SURROGATE) | |
860 | ||
d35f2ca5 | 861 | #define UTF8_DISALLOW_ILLEGAL_INTERCHANGE \ |
ecc1615f | 862 | (UTF8_DISALLOW_ILLEGAL_C9_INTERCHANGE|UTF8_DISALLOW_NONCHAR) |
949cf498 | 863 | #define UTF8_WARN_ILLEGAL_INTERCHANGE \ |
ecc1615f KW |
864 | (UTF8_WARN_ILLEGAL_C9_INTERCHANGE|UTF8_WARN_NONCHAR) |
865 | ||
0eb3d6a0 KW |
866 | /* This is typically used for code that processes UTF-8 input and doesn't want |
867 | * to have to deal with any malformations that might be present. All such will | |
868 | * be safely replaced by the REPLACEMENT CHARACTER, unless other flags | |
869 | * overriding this are also present. */ | |
2d532c27 KW |
870 | #define UTF8_ALLOW_ANY ( UTF8_ALLOW_CONTINUATION \ |
871 | |UTF8_ALLOW_NON_CONTINUATION \ | |
872 | |UTF8_ALLOW_SHORT \ | |
d60baaa7 KW |
873 | |UTF8_ALLOW_LONG \ |
874 | |UTF8_ALLOW_OVERFLOW) | |
2d532c27 KW |
875 | |
876 | /* Accept any Perl-extended UTF-8 that evaluates to any UV on the platform, but | |
cd01d3b1 | 877 | * not any malformed. This is the default. */ |
2d532c27 KW |
878 | #define UTF8_ALLOW_ANYUV 0 |
879 | #define UTF8_ALLOW_DEFAULT UTF8_ALLOW_ANYUV | |
1d72bdf6 | 880 | |
89d986df KW |
881 | /* |
882 | =for apidoc Am|bool|UTF8_IS_SURROGATE|const U8 *s|const U8 *e | |
883 | ||
884 | Evaluates to non-zero if the first few bytes of the string starting at C<s> and | |
885 | looking no further than S<C<e - 1>> are well-formed UTF-8 that represents one | |
886 | of the Unicode surrogate code points; otherwise it evaluates to 0. If | |
887 | non-zero, the value gives how many bytes starting at C<s> comprise the code | |
888 | point's representation. | |
889 | ||
890 | =cut | |
891 | */ | |
892 | #define UTF8_IS_SURROGATE(s, e) is_SURROGATE_utf8_safe(s, e) | |
893 | ||
894 | ||
895 | #define UTF8_IS_REPLACEMENT(s, send) is_REPLACEMENT_utf8_safe(s,send) | |
896 | ||
285aa1f0 | 897 | #define MAX_LEGAL_CP ((UV)IV_MAX) |
40606899 | 898 | |
89d986df KW |
899 | /* |
900 | =for apidoc Am|bool|UTF8_IS_SUPER|const U8 *s|const U8 *e | |
901 | ||
902 | Recall that Perl recognizes an extension to UTF-8 that can encode code | |
903 | points larger than the ones defined by Unicode, which are 0..0x10FFFF. | |
904 | ||
905 | This macro evaluates to non-zero if the first few bytes of the string starting | |
906 | at C<s> and looking no further than S<C<e - 1>> are from this UTF-8 extension; | |
907 | otherwise it evaluates to 0. If non-zero, the value gives how many bytes | |
908 | starting at C<s> comprise the code point's representation. | |
909 | ||
910 | 0 is returned if the bytes are not well-formed extended UTF-8, or if they | |
911 | represent a code point that cannot fit in a UV on the current platform. Hence | |
912 | this macro can give different results when run on a 64-bit word machine than on | |
913 | one with a 32-bit word size. | |
0c58a72b | 914 | |
891fd405 | 915 | Note that it is illegal to have code points that are larger than what can |
89d986df | 916 | fit in an IV on the current machine. |
7131f24d | 917 | |
89d986df KW |
918 | =cut |
919 | ||
920 | * ASCII EBCDIC I8 | |
7131f24d KW |
921 | * U+10FFFF: \xF4\x8F\xBF\xBF \xF9\xA1\xBF\xBF\xBF max legal Unicode |
922 | * U+110000: \xF4\x90\x80\x80 \xF9\xA2\xA0\xA0\xA0 | |
923 | * U+110001: \xF4\x90\x80\x81 \xF9\xA2\xA0\xA0\xA1 | |
89d986df KW |
924 | */ |
925 | #ifdef EBCDIC | |
a14e0a36 | 926 | # define UTF8_IS_SUPER(s, e) \ |
8974941d KW |
927 | (( ((e) > (s) + 4) \ |
928 | && (NATIVE_UTF8_TO_I8(*(s)) >= 0xF9) \ | |
929 | && UNLIKELY( NATIVE_UTF8_TO_I8(*(s)) > 0xF9 \ | |
930 | || (NATIVE_UTF8_TO_I8(*((s) + 1)) >= 0xA2)) \ | |
931 | && LIKELY((s) + UTF8SKIP(s) <= (e))) \ | |
0cf25474 | 932 | ? is_utf8_char_helper(s, s + UTF8SKIP(s), 0) : 0) |
7131f24d | 933 | #else |
a14e0a36 | 934 | # define UTF8_IS_SUPER(s, e) \ |
8974941d | 935 | (( ((e) > (s) + 3) \ |
89d986df | 936 | && (*(U8*) (s)) >= 0xF4 \ |
8974941d KW |
937 | && (UNLIKELY( ((*(U8*) (s)) > 0xF4) \ |
938 | || (*((U8*) (s) + 1) >= 0x90))) \ | |
89d986df | 939 | && LIKELY((s) + UTF8SKIP(s) <= (e))) \ |
1376b35c | 940 | ? is_utf8_char_helper(s, s + UTF8SKIP(s), 0) : 0) |
7131f24d KW |
941 | #endif |
942 | ||
b96a92fb KW |
943 | /* These are now machine generated, and the 'given' clause is no longer |
944 | * applicable */ | |
0c58a72b | 945 | #define UTF8_IS_NONCHAR_GIVEN_THAT_NON_SUPER_AND_GE_PROBLEMATIC(s, e) \ |
89d986df KW |
946 | cBOOL(is_NONCHAR_utf8_safe(s,e)) |
947 | ||
948 | /* | |
949 | =for apidoc Am|bool|UTF8_IS_NONCHAR|const U8 *s|const U8 *e | |
950 | ||
951 | Evaluates to non-zero if the first few bytes of the string starting at C<s> and | |
952 | looking no further than S<C<e - 1>> are well-formed UTF-8 that represents one | |
953 | of the Unicode non-character code points; otherwise it evaluates to 0. If | |
954 | non-zero, the value gives how many bytes starting at C<s> comprise the code | |
955 | point's representation. | |
956 | ||
d296fe14 KW |
957 | =for apidoc AmnU|UV|UNICODE_REPLACEMENT |
958 | ||
959 | Evaluates to 0xFFFD, the code point of the Unicode REPLACEMENT CHARACTER | |
960 | ||
89d986df KW |
961 | =cut |
962 | */ | |
0c58a72b KW |
963 | #define UTF8_IS_NONCHAR(s, e) \ |
964 | UTF8_IS_NONCHAR_GIVEN_THAT_NON_SUPER_AND_GE_PROBLEMATIC(s, e) | |
7131f24d | 965 | |
c867b360 JH |
966 | #define UNICODE_SURROGATE_FIRST 0xD800 |
967 | #define UNICODE_SURROGATE_LAST 0xDFFF | |
968 | #define UNICODE_REPLACEMENT 0xFFFD | |
969 | #define UNICODE_BYTE_ORDER_MARK 0xFEFF | |
1d72bdf6 | 970 | |
b851fbc1 | 971 | /* Though our UTF-8 encoding can go beyond this, |
c76687c5 | 972 | * let's be conservative and do as Unicode says. */ |
b851fbc1 JH |
973 | #define PERL_UNICODE_MAX 0x10FFFF |
974 | ||
d044b7a7 KW |
975 | #define UNICODE_WARN_SURROGATE 0x0001 /* UTF-16 surrogates */ |
976 | #define UNICODE_WARN_NONCHAR 0x0002 /* Non-char code points */ | |
977 | #define UNICODE_WARN_SUPER 0x0004 /* Above 0x10FFFF */ | |
978 | #define UNICODE_WARN_PERL_EXTENDED 0x0008 /* Above 0x7FFF_FFFF */ | |
979 | #define UNICODE_WARN_ABOVE_31_BIT UNICODE_WARN_PERL_EXTENDED | |
980 | #define UNICODE_DISALLOW_SURROGATE 0x0010 | |
981 | #define UNICODE_DISALLOW_NONCHAR 0x0020 | |
982 | #define UNICODE_DISALLOW_SUPER 0x0040 | |
983 | #define UNICODE_DISALLOW_PERL_EXTENDED 0x0080 | |
24b4c303 KW |
984 | |
985 | #ifdef PERL_CORE | |
986 | # define UNICODE_ALLOW_ABOVE_IV_MAX 0x0100 | |
987 | #endif | |
d044b7a7 | 988 | #define UNICODE_DISALLOW_ABOVE_31_BIT UNICODE_DISALLOW_PERL_EXTENDED |
33f38593 KW |
989 | |
990 | #define UNICODE_GOT_SURROGATE UNICODE_DISALLOW_SURROGATE | |
991 | #define UNICODE_GOT_NONCHAR UNICODE_DISALLOW_NONCHAR | |
992 | #define UNICODE_GOT_SUPER UNICODE_DISALLOW_SUPER | |
993 | #define UNICODE_GOT_PERL_EXTENDED UNICODE_DISALLOW_PERL_EXTENDED | |
994 | ||
ecc1615f KW |
995 | #define UNICODE_WARN_ILLEGAL_C9_INTERCHANGE \ |
996 | (UNICODE_WARN_SURROGATE|UNICODE_WARN_SUPER) | |
bb88be5f | 997 | #define UNICODE_WARN_ILLEGAL_INTERCHANGE \ |
ecc1615f KW |
998 | (UNICODE_WARN_ILLEGAL_C9_INTERCHANGE|UNICODE_WARN_NONCHAR) |
999 | #define UNICODE_DISALLOW_ILLEGAL_C9_INTERCHANGE \ | |
1000 | (UNICODE_DISALLOW_SURROGATE|UNICODE_DISALLOW_SUPER) | |
bb88be5f | 1001 | #define UNICODE_DISALLOW_ILLEGAL_INTERCHANGE \ |
ecc1615f | 1002 | (UNICODE_DISALLOW_ILLEGAL_C9_INTERCHANGE|UNICODE_DISALLOW_NONCHAR) |
949cf498 KW |
1003 | |
1004 | /* For backward source compatibility, as are now the default */ | |
1005 | #define UNICODE_ALLOW_SURROGATE 0 | |
1006 | #define UNICODE_ALLOW_SUPER 0 | |
1007 | #define UNICODE_ALLOW_ANY 0 | |
b851fbc1 | 1008 | |
08e4a1ef KW |
1009 | /* This matches the 2048 code points between these */ |
1010 | #define UNICODE_IS_SURROGATE(uv) UNLIKELY(inRANGE(uv, UNICODE_SURROGATE_FIRST, \ | |
1011 | UNICODE_SURROGATE_LAST)) | |
2d6b3d38 | 1012 | |
8974941d KW |
1013 | #define UNICODE_IS_REPLACEMENT(uv) UNLIKELY((UV) (uv) == UNICODE_REPLACEMENT) |
1014 | #define UNICODE_IS_BYTE_ORDER_MARK(uv) UNLIKELY((UV) (uv) \ | |
1015 | == UNICODE_BYTE_ORDER_MARK) | |
c149ab20 KW |
1016 | |
1017 | /* Is 'uv' one of the 32 contiguous-range noncharacters? */ | |
21e13fc5 KW |
1018 | #define UNICODE_IS_32_CONTIGUOUS_NONCHARS(uv) \ |
1019 | UNLIKELY(inRANGE(uv, 0xFDD0, 0xFDEF)) | |
c149ab20 KW |
1020 | |
1021 | /* Is 'uv' one of the 34 plane-ending noncharacters 0xFFFE, 0xFFFF, 0x1FFFE, | |
1022 | * 0x1FFFF, ... 0x10FFFE, 0x10FFFF, given that we know that 'uv' is not above | |
1023 | * the Unicode legal max */ | |
1024 | #define UNICODE_IS_END_PLANE_NONCHAR_GIVEN_NOT_SUPER(uv) \ | |
8974941d | 1025 | UNLIKELY(((UV) (uv) & 0xFFFE) == 0xFFFE) |
c149ab20 KW |
1026 | |
1027 | #define UNICODE_IS_NONCHAR(uv) \ | |
aacc849b KW |
1028 | ( UNLIKELY(UNICODE_IS_32_CONTIGUOUS_NONCHARS(uv)) \ |
1029 | || ( UNLIKELY(UNICODE_IS_END_PLANE_NONCHAR_GIVEN_NOT_SUPER(uv)) \ | |
1030 | && LIKELY(! UNICODE_IS_SUPER(uv)))) | |
c149ab20 | 1031 | |
8974941d | 1032 | #define UNICODE_IS_SUPER(uv) UNLIKELY((UV) (uv) > PERL_UNICODE_MAX) |
1d72bdf6 | 1033 | |
ec34087a KW |
1034 | #define LATIN_SMALL_LETTER_SHARP_S LATIN_SMALL_LETTER_SHARP_S_NATIVE |
1035 | #define LATIN_SMALL_LETTER_Y_WITH_DIAERESIS \ | |
1036 | LATIN_SMALL_LETTER_Y_WITH_DIAERESIS_NATIVE | |
1037 | #define MICRO_SIGN MICRO_SIGN_NATIVE | |
1038 | #define LATIN_CAPITAL_LETTER_A_WITH_RING_ABOVE \ | |
1039 | LATIN_CAPITAL_LETTER_A_WITH_RING_ABOVE_NATIVE | |
1040 | #define LATIN_SMALL_LETTER_A_WITH_RING_ABOVE \ | |
1041 | LATIN_SMALL_LETTER_A_WITH_RING_ABOVE_NATIVE | |
09091399 JH |
1042 | #define UNICODE_GREEK_CAPITAL_LETTER_SIGMA 0x03A3 |
1043 | #define UNICODE_GREEK_SMALL_LETTER_FINAL_SIGMA 0x03C2 | |
1044 | #define UNICODE_GREEK_SMALL_LETTER_SIGMA 0x03C3 | |
9dcbe121 | 1045 | #define GREEK_SMALL_LETTER_MU 0x03BC |
9e682c18 KW |
1046 | #define GREEK_CAPITAL_LETTER_MU 0x039C /* Upper and title case |
1047 | of MICRON */ | |
1048 | #define LATIN_CAPITAL_LETTER_Y_WITH_DIAERESIS 0x0178 /* Also is title case */ | |
0766489e KW |
1049 | #ifdef LATIN_CAPITAL_LETTER_SHARP_S_UTF8 |
1050 | # define LATIN_CAPITAL_LETTER_SHARP_S 0x1E9E | |
1051 | #endif | |
74894415 KW |
1052 | #define LATIN_CAPITAL_LETTER_I_WITH_DOT_ABOVE 0x130 |
1053 | #define LATIN_SMALL_LETTER_DOTLESS_I 0x131 | |
9e682c18 | 1054 | #define LATIN_SMALL_LETTER_LONG_S 0x017F |
a9f50d33 KW |
1055 | #define LATIN_SMALL_LIGATURE_LONG_S_T 0xFB05 |
1056 | #define LATIN_SMALL_LIGATURE_ST 0xFB06 | |
9e682c18 KW |
1057 | #define KELVIN_SIGN 0x212A |
1058 | #define ANGSTROM_SIGN 0x212B | |
09091399 | 1059 | |
9e55ce06 | 1060 | #define UNI_DISPLAY_ISPRINT 0x0001 |
c728cb41 | 1061 | #define UNI_DISPLAY_BACKSLASH 0x0002 |
daf6caf1 KW |
1062 | #define UNI_DISPLAY_BACKSPACE 0x0004 /* Allow \b when also |
1063 | UNI_DISPLAY_BACKSLASH */ | |
1064 | #define UNI_DISPLAY_QQ (UNI_DISPLAY_ISPRINT \ | |
1065 | |UNI_DISPLAY_BACKSLASH \ | |
1066 | |UNI_DISPLAY_BACKSPACE) | |
1067 | ||
1068 | /* Character classes could also allow \b, but not patterns in general */ | |
c728cb41 | 1069 | #define UNI_DISPLAY_REGEX (UNI_DISPLAY_ISPRINT|UNI_DISPLAY_BACKSLASH) |
9e55ce06 | 1070 | |
e0ffa6d6 | 1071 | /* Should be removed; maybe deprecated, but not used in CPAN */ |
ebc501f0 | 1072 | #define SHARP_S_SKIP 2 |
3b0fc154 | 1073 | |
3cedd9d9 | 1074 | #define is_utf8_char_buf(buf, buf_end) isUTF8_CHAR(buf, buf_end) |
976c1b08 KW |
1075 | #define bytes_from_utf8(s, lenp, is_utf8p) \ |
1076 | bytes_from_utf8_loc(s, lenp, is_utf8p, 0) | |
3cedd9d9 | 1077 | |
e23e8bc1 KW |
1078 | /* |
1079 | ||
25e3a4e0 KW |
1080 | =for apidoc Am|STRLEN|isUTF8_CHAR_flags|const U8 *s|const U8 *e| const U32 flags |
1081 | ||
1082 | Evaluates to non-zero if the first few bytes of the string starting at C<s> and | |
1083 | looking no further than S<C<e - 1>> are well-formed UTF-8, as extended by Perl, | |
1084 | that represents some code point, subject to the restrictions given by C<flags>; | |
1085 | otherwise it evaluates to 0. If non-zero, the value gives how many bytes | |
2717076a KW |
1086 | starting at C<s> comprise the code point's representation. Any bytes remaining |
1087 | before C<e>, but beyond the ones needed to form the first code point in C<s>, | |
1088 | are not examined. | |
25e3a4e0 KW |
1089 | |
1090 | If C<flags> is 0, this gives the same results as C<L</isUTF8_CHAR>>; | |
1091 | if C<flags> is C<UTF8_DISALLOW_ILLEGAL_INTERCHANGE>, this gives the same results | |
1092 | as C<L</isSTRICT_UTF8_CHAR>>; | |
1093 | and if C<flags> is C<UTF8_DISALLOW_ILLEGAL_C9_INTERCHANGE>, this gives | |
1094 | the same results as C<L</isC9_STRICT_UTF8_CHAR>>. | |
1095 | Otherwise C<flags> may be any combination of the C<UTF8_DISALLOW_I<foo>> flags | |
1096 | understood by C<L</utf8n_to_uvchr>>, with the same meanings. | |
1097 | ||
1098 | The three alternative macros are for the most commonly needed validations; they | |
1099 | are likely to run somewhat faster than this more general one, as they can be | |
1100 | inlined into your code. | |
1101 | ||
9f2abfde KW |
1102 | Use L</is_utf8_string_flags>, L</is_utf8_string_loc_flags>, and |
1103 | L</is_utf8_string_loclen_flags> to check entire strings. | |
1104 | ||
25e3a4e0 KW |
1105 | =cut |
1106 | */ | |
1107 | ||
1108 | #define isUTF8_CHAR_flags(s, e, flags) \ | |
1109 | (UNLIKELY((e) <= (s)) \ | |
1110 | ? 0 \ | |
1111 | : (UTF8_IS_INVARIANT(*s)) \ | |
1112 | ? 1 \ | |
1113 | : UNLIKELY(((e) - (s)) < UTF8SKIP(s)) \ | |
1114 | ? 0 \ | |
1376b35c | 1115 | : is_utf8_char_helper(s, e, flags)) |
25e3a4e0 | 1116 | |
6302f837 KW |
1117 | /* Do not use; should be deprecated. Use isUTF8_CHAR() instead; this is |
1118 | * retained solely for backwards compatibility */ | |
1119 | #define IS_UTF8_CHAR(p, n) (isUTF8_CHAR(p, (p) + (n)) == n) | |
e9a8c099 | 1120 | |
6a5bc5ac | 1121 | #endif /* PERL_UTF8_H_ */ |
57f0e7e2 | 1122 | |
e9a8c099 | 1123 | /* |
14d04a33 | 1124 | * ex: set ts=8 sts=4 sw=4 et: |
e9a8c099 | 1125 | */ |