This is a live mirror of the Perl 5 development currently hosted at https://github.com/perl/perl5
https://perl5.git.perl.org / perl5.git / blame_incremental
? search:
summary | shortlog | log | commit | commitdiff | tree
blob | blame (non-incremental) | history | HEAD
Add new function utf8_from_bytes_loc()
[perl5.git] / utf8.h
... / ...
CommitLineData
1/* utf8.h
2 *
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 *
9 * Copyright (C) 2000, 2001, 2002, 2005, 2006, 2007, 2009,
10 * 2010, 2011 by Larry Wall and others
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
17#ifndef PERL_UTF8_H_ /* Guard against recursive inclusion */
18#define PERL_UTF8_H_ 1
19
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)
25#else
26# define USE_UTF8_IN_NAMES (PL_hints & HINT_UTF8)
27#endif
28
29#include "regcharclass.h"
30#include "unicode_constants.h"
31
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
36
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
41
42/*
43=head1 Unicode Support
44L<perlguts/Unicode Support> has an introduction to this API.
45
46See also L</Character classification>,
47and L</Character case changing>.
48Various functions outside this section also work specially with Unicode.
49Search for the string "utf8" in this document.
50
51=for apidoc is_ascii_string
52
53This is a misleadingly-named synonym for L</is_utf8_invariant_string>.
54On ASCII-ish platforms, the name isn't misleading: the ASCII-range characters
55are exactly the UTF-8 invariants. But EBCDIC machines have more invariants
56than just the ASCII characters, so C<is_utf8_invariant_string> is preferred.
57
58=for apidoc is_invariant_string
59
60This is a somewhat misleadingly-named synonym for L</is_utf8_invariant_string>.
61C<is_utf8_invariant_string> is preferred, as it indicates under what conditions
62the string is invariant.
63
64=cut
65*/
66#define is_ascii_string(s, len) is_utf8_invariant_string(s, len)
67#define is_invariant_string(s, len) is_utf8_invariant_string(s, len)
68
69#define uvchr_to_utf8(a,b) uvchr_to_utf8_flags(a,b,0)
70#define uvchr_to_utf8_flags(d,uv,flags) \
71 uvoffuni_to_utf8_flags(d,NATIVE_TO_UNI(uv),flags)
72#define utf8_to_uvchr_buf(s, e, lenp) \
73 utf8n_to_uvchr(s, (U8*)(e) - (U8*)(s), lenp, \
74 ckWARN_d(WARN_UTF8) ? 0 : UTF8_ALLOW_ANY)
75#define utf8n_to_uvchr(s, len, lenp, flags) \
76 utf8n_to_uvchr_error(s, len, lenp, flags, 0)
77
78#define to_uni_fold(c, p, lenp) _to_uni_fold_flags(c, p, lenp, FOLD_FLAGS_FULL)
79
80#define to_utf8_fold(s, r, lenr) \
81 _to_utf8_fold_flags (s, NULL, r, lenr, FOLD_FLAGS_FULL, __FILE__, __LINE__)
82#define to_utf8_lower(s, r, lenr) \
83 _to_utf8_lower_flags(s, NULL, r ,lenr, 0, __FILE__, __LINE__)
84#define to_utf8_upper(s, r, lenr) \
85 _to_utf8_upper_flags(s, NULL, r, lenr, 0, __FILE__, __LINE__)
86#define to_utf8_title(s, r, lenr) \
87 _to_utf8_title_flags(s, NULL, r, lenr ,0, __FILE__, __LINE__)
88
89#define foldEQ_utf8(s1, pe1, l1, u1, s2, pe2, l2, u2) \
90 foldEQ_utf8_flags(s1, pe1, l1, u1, s2, pe2, l2, u2, 0)
91#define FOLDEQ_UTF8_NOMIX_ASCII (1 << 0)
92#define FOLDEQ_LOCALE (1 << 1)
93#define FOLDEQ_S1_ALREADY_FOLDED (1 << 2)
94#define FOLDEQ_S2_ALREADY_FOLDED (1 << 3)
95#define FOLDEQ_S1_FOLDS_SANE (1 << 4)
96#define FOLDEQ_S2_FOLDS_SANE (1 << 5)
97
98#define ibcmp_utf8(s1, pe1, l1, u1, s2, pe2, l2, u2) \
99 cBOOL(! foldEQ_utf8(s1, pe1, l1, u1, s2, pe2, l2, u2))
100
101#ifdef EBCDIC
102/* The equivalent of these macros but implementing UTF-EBCDIC
103 are in the following header file:
104 */
105
106#include "utfebcdic.h"
107
108#else /* ! EBCDIC */
109START_EXTERN_C
110
111/* How wide can a single UTF-8 encoded character become in bytes. */
112/* NOTE: Strictly speaking Perl's UTF-8 should not be called UTF-8 since UTF-8
113 * is an encoding of Unicode, and Unicode's upper limit, 0x10FFFF, can be
114 * expressed with 4 bytes. However, Perl thinks of UTF-8 as a way to encode
115 * non-negative integers in a binary format, even those above Unicode */
116#define UTF8_MAXBYTES 13
117
118#ifdef DOINIT
119EXTCONST unsigned char PL_utf8skip[] = {
120/* 0x00 */ 1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1, /* ascii */
121/* 0x10 */ 1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1, /* ascii */
122/* 0x20 */ 1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1, /* ascii */
123/* 0x30 */ 1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1, /* ascii */
124/* 0x40 */ 1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1, /* ascii */
125/* 0x50 */ 1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1, /* ascii */
126/* 0x60 */ 1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1, /* ascii */
127/* 0x70 */ 1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1, /* ascii */
128/* 0x80 */ 1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1, /* bogus: continuation byte */
129/* 0x90 */ 1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1, /* bogus: continuation byte */
130/* 0xA0 */ 1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1, /* bogus: continuation byte */
131/* 0xB0 */ 1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1, /* bogus: continuation byte */
132/* 0xC0 */ 2,2, /* overlong */
133/* 0xC2 */ 2,2,2,2,2,2,2,2,2,2,2,2,2,2, /* U+0080 to U+03FF */
134/* 0xD0 */ 2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2, /* U+0400 to U+07FF */
135/* 0xE0 */ 3,3,3,3,3,3,3,3,3,3,3,3,3,3,3,3, /* U+0800 to U+FFFF */
136/* 0xF0 */ 4,4,4,4,4,4,4,4,5,5,5,5,6,6, /* above BMP to 2**31 - 1 */
137 /* Perl extended (never was official UTF-8). Up to 36 bit */
138/* 0xFE */ 7,
139 /* More extended, Up to 72 bits (64-bit + reserved) */
140/* 0xFF */ UTF8_MAXBYTES
141};
142#else
143EXTCONST unsigned char PL_utf8skip[];
144#endif
145
146END_EXTERN_C
147
148#if defined(_MSC_VER) && _MSC_VER < 1400
149/* older MSVC versions have a smallish macro buffer */
150#define PERL_SMALL_MACRO_BUFFER
151#endif
152
153/* Native character to/from iso-8859-1. Are the identity functions on ASCII
154 * platforms */
155#ifdef PERL_SMALL_MACRO_BUFFER
156#define NATIVE_TO_LATIN1(ch) ((U8)(ch))
157#define LATIN1_TO_NATIVE(ch) ((U8)(ch))
158#else
159#define NATIVE_TO_LATIN1(ch) (__ASSERT_(FITS_IN_8_BITS(ch)) ((U8) (ch)))
160#define LATIN1_TO_NATIVE(ch) (__ASSERT_(FITS_IN_8_BITS(ch)) ((U8) (ch)))
161#endif
162
163/* I8 is an intermediate version of UTF-8 used only in UTF-EBCDIC. We thus
164 * consider it to be identical to UTF-8 on ASCII platforms. Strictly speaking
165 * UTF-8 and UTF-EBCDIC are two different things, but we often conflate them
166 * because they are 8-bit encodings that serve the same purpose in Perl, and
167 * rarely do we need to distinguish them. The term "NATIVE_UTF8" applies to
168 * whichever one is applicable on the current platform */
169#ifdef PERL_SMALL_MACRO_BUFFER
170#define NATIVE_UTF8_TO_I8(ch) ((U8) (ch))
171#define I8_TO_NATIVE_UTF8(ch) ((U8) (ch))
172#else
173#define NATIVE_UTF8_TO_I8(ch) (__ASSERT_(FITS_IN_8_BITS(ch)) ((U8) (ch)))
174#define I8_TO_NATIVE_UTF8(ch) (__ASSERT_(FITS_IN_8_BITS(ch)) ((U8) (ch)))
175#endif
176
177/* Transforms in wide UV chars */
178#define UNI_TO_NATIVE(ch) ((UV) (ch))
179#define NATIVE_TO_UNI(ch) ((UV) (ch))
180
181/*
182
183 The following table is from Unicode 3.2, plus the Perl extensions for above
184 U+10FFFF
185
186 Code Points 1st Byte 2nd Byte 3rd 4th 5th 6th 7th 8th-13th
187
188 U+0000..U+007F 00..7F
189 U+0080..U+07FF * C2..DF 80..BF
190 U+0800..U+0FFF E0 * A0..BF 80..BF
191 U+1000..U+CFFF E1..EC 80..BF 80..BF
192 U+D000..U+D7FF ED 80..9F 80..BF
193 U+D800..U+DFFF ED A0..BF 80..BF (surrogates)
194 U+E000..U+FFFF EE..EF 80..BF 80..BF
195 U+10000..U+3FFFF F0 * 90..BF 80..BF 80..BF
196 U+40000..U+FFFFF F1..F3 80..BF 80..BF 80..BF
197 U+100000..U+10FFFF F4 80..8F 80..BF 80..BF
198 Below are above-Unicode code points
199 U+110000..U+13FFFF F4 90..BF 80..BF 80..BF
200 U+110000..U+1FFFFF F5..F7 80..BF 80..BF 80..BF
201 U+200000..U+FFFFFF F8 * 88..BF 80..BF 80..BF 80..BF
202U+1000000..U+3FFFFFF F9..FB 80..BF 80..BF 80..BF 80..BF
203U+4000000..U+3FFFFFFF FC * 84..BF 80..BF 80..BF 80..BF 80..BF
204U+40000000..U+7FFFFFFF FD 80..BF 80..BF 80..BF 80..BF 80..BF
205U+80000000..U+FFFFFFFFF FE * 82..BF 80..BF 80..BF 80..BF 80..BF 80..BF
206U+1000000000.. FF 80..BF 80..BF 80..BF 80..BF 80..BF * 81..BF 80..BF
207
208Note the gaps before several of the byte entries above marked by '*'. These are
209caused by legal UTF-8 avoiding non-shortest encodings: it is technically
210possible to UTF-8-encode a single code point in different ways, but that is
211explicitly forbidden, and the shortest possible encoding should always be used
212(and that is what Perl does). The non-shortest ones are called 'overlongs'.
213
214 */
215
216/*
217 Another way to look at it, as bits:
218
219 Code Points 1st Byte 2nd Byte 3rd Byte 4th Byte
220
221 0aaa aaaa 0aaa aaaa
222 0000 0bbb bbaa aaaa 110b bbbb 10aa aaaa
223 cccc bbbb bbaa aaaa 1110 cccc 10bb bbbb 10aa aaaa
224 00 000d ddcc cccc bbbb bbaa aaaa 1111 0ddd 10cc cccc 10bb bbbb 10aa aaaa
225
226As you can see, the continuation bytes all begin with C<10>, and the
227leading bits of the start byte tell how many bytes there are in the
228encoded character.
229
230Perl's extended UTF-8 means we can have start bytes up through FF, though any
231beginning with FF yields a code point that is too large for 32-bit ASCII
232platforms. FF signals to use 13 bytes for the encoded character. This breaks
233the paradigm that the number of leading bits gives how many total bytes there
234are in the character.
235
236*/
237
238/* Anything larger than this will overflow the word if it were converted into a UV */
239#if defined(UV_IS_QUAD)
240# define HIGHEST_REPRESENTABLE_UTF8 "\xFF\x80\x8F\xBF\xBF\xBF\xBF\xBF\xBF\xBF\xBF\xBF\xBF"
241#else
242# define HIGHEST_REPRESENTABLE_UTF8 "\xFE\x83\xBF\xBF\xBF\xBF\xBF"
243#endif
244
245/* Is the representation of the Unicode code point 'cp' the same regardless of
246 * being encoded in UTF-8 or not? */
247#define OFFUNI_IS_INVARIANT(cp) isASCII(cp)
248
249/*
250=for apidoc Am|bool|UVCHR_IS_INVARIANT|UV cp
251
252Evaluates to 1 if the representation of code point C<cp> is the same whether or
253not it is encoded in UTF-8; otherwise evaluates to 0. UTF-8 invariant
254characters can be copied as-is when converting to/from UTF-8, saving time.
255C<cp> is Unicode if above 255; otherwise is platform-native.
256
257=cut
258 */
259
260#define UVCHR_IS_INVARIANT(cp) OFFUNI_IS_INVARIANT(cp)
261
262/* This defines the bits that are to be in the continuation bytes of a multi-byte
263 * UTF-8 encoded character that mark it is a continuation byte. */
264#define UTF_CONTINUATION_MARK 0x80
265
266/* Misleadingly named: is the UTF8-encoded byte 'c' part of a variant sequence
267 * in UTF-8? This is the inverse of UTF8_IS_INVARIANT. The |0 makes sure this
268 * isn't mistakenly called with a ptr argument */
269#define UTF8_IS_CONTINUED(c) (__ASSERT_(FITS_IN_8_BITS(c)) \
270 ((U8)((c) | 0)) & UTF_CONTINUATION_MARK)
271
272/* Is the byte 'c' the first byte of a multi-byte UTF8-8 encoded sequence?
273 * This doesn't catch invariants (they are single-byte). It also excludes the
274 * illegal overlong sequences that begin with C0 and C1. The |0 makes sure
275 * this isn't mistakenly called with a ptr argument */
276#define UTF8_IS_START(c) (__ASSERT_(FITS_IN_8_BITS(c)) \
277 ((U8)((c) | 0)) >= 0xc2)
278
279/* For use in UTF8_IS_CONTINUATION() below */
280#define UTF_IS_CONTINUATION_MASK 0xC0
281
282/* Is the byte 'c' part of a multi-byte UTF8-8 encoded sequence, and not the
283 * first byte thereof? The |0 makes sure this isn't mistakenly called with a
284 * ptr argument */
285#define UTF8_IS_CONTINUATION(c) (__ASSERT_(FITS_IN_8_BITS(c)) \
286 (((U8)((c) | 0)) & UTF_IS_CONTINUATION_MASK) == UTF_CONTINUATION_MARK)
287
288/* Is the UTF8-encoded byte 'c' the first byte of a two byte sequence? Use
289 * UTF8_IS_NEXT_CHAR_DOWNGRADEABLE() instead if the input isn't known to
290 * be well-formed. Masking with 0xfe allows the low bit to be 0 or 1; thus
291 * this matches 0xc[23]. The |0 makes sure this isn't mistakenly called with a
292 * ptr argument */
293#define UTF8_IS_DOWNGRADEABLE_START(c) (__ASSERT_(FITS_IN_8_BITS(c)) \
294 (((U8)((c) | 0)) & 0xfe) == 0xc2)
295
296/* Is the UTF8-encoded byte 'c' the first byte of a sequence of bytes that
297 * represent a code point > 255? The |0 makes sure this isn't mistakenly
298 * called with a ptr argument */
299#define UTF8_IS_ABOVE_LATIN1(c) (__ASSERT_(FITS_IN_8_BITS(c)) \
300 ((U8)((c) | 0)) >= 0xc4)
301
302/* This is the number of low-order bits a continuation byte in a UTF-8 encoded
303 * sequence contributes to the specification of the code point. In the bit
304 * maps above, you see that the first 2 bits are a constant '10', leaving 6 of
305 * real information */
306#define UTF_ACCUMULATION_SHIFT 6
307
308/* ^? is defined to be DEL on ASCII systems. See the definition of toCTRL()
309 * for more */
310#define QUESTION_MARK_CTRL DEL_NATIVE
311
312/* Surrogates, non-character code points and above-Unicode code points are
313 * problematic in some contexts. This allows code that needs to check for
314 * those to to quickly exclude the vast majority of code points it will
315 * encounter */
316#define isUTF8_POSSIBLY_PROBLEMATIC(c) (__ASSERT_(FITS_IN_8_BITS(c)) \
317 (U8) c >= 0xED)
318
319/* A helper macro for isUTF8_CHAR, so use that one instead of this. This was
320 * generated by regen/regcharclass.pl, and then moved here. Then it was
321 * hand-edited to add some LIKELY() calls, presuming that malformations are
322 * unlikely. The lines that generated it were then commented out. This was
323 * done because it takes on the order of 10 minutes to generate, and is never
324 * going to change, unless the generated code is improved, and figuring out
325 * the LIKELYs there would be hard.
326 *
327 UTF8_CHAR: Matches legal UTF-8 variant code points up through 0x1FFFFFF
328
329 0x80 - 0x1FFFFF
330*/
331/*** GENERATED CODE ***/
332#define is_UTF8_CHAR_utf8_no_length_checks(s) \
333( ( 0xC2 <= ((const U8*)s)[0] && ((const U8*)s)[0] <= 0xDF ) ? \
334 ( LIKELY( ( ((const U8*)s)[1] & 0xC0 ) == 0x80 ) ? 2 : 0 ) \
335: ( 0xE0 == ((const U8*)s)[0] ) ? \
336 ( LIKELY( ( ( ((const U8*)s)[1] & 0xE0 ) == 0xA0 ) && ( ( ((const U8*)s)[2] & 0xC0 ) == 0x80 ) ) ? 3 : 0 )\
337: ( 0xE1 <= ((const U8*)s)[0] && ((const U8*)s)[0] <= 0xEF ) ? \
338 ( LIKELY( ( ( ((const U8*)s)[1] & 0xC0 ) == 0x80 ) && ( ( ((const U8*)s)[2] & 0xC0 ) == 0x80 ) ) ? 3 : 0 )\
339: ( 0xF0 == ((const U8*)s)[0] ) ? \
340 ( LIKELY( ( ( 0x90 <= ((const U8*)s)[1] && ((const U8*)s)[1] <= 0xBF ) && ( ( ((const U8*)s)[2] & 0xC0 ) == 0x80 ) ) && ( ( ((const U8*)s)[3] & 0xC0 ) == 0x80 ) ) ? 4 : 0 )\
341: ( ( ( ( 0xF1 <= ((const U8*)s)[0] && ((const U8*)s)[0] <= 0xF7 ) && LIKELY( ( ((const U8*)s)[1] & 0xC0 ) == 0x80 ) ) && LIKELY( ( ((const U8*)s)[2] & 0xC0 ) == 0x80 ) ) && LIKELY( ( ((const U8*)s)[3] & 0xC0 ) == 0x80 ) ) ? 4 : 0 )
342
343/* The above macro handles UTF-8 that has this start byte as the maximum */
344#define _IS_UTF8_CHAR_HIGHEST_START_BYTE 0xF7
345
346/* A helper macro for isSTRICT_UTF8_CHAR, so use that one instead of this.
347 * Like is_UTF8_CHAR_utf8_no_length_checks(), this was moved here and LIKELYs
348 * added manually.
349 *
350 STRICT_UTF8_CHAR: Matches legal Unicode UTF-8 variant code points, no
351 surrrogates nor non-character code points
352*/
353/*** GENERATED CODE ***/
354#define is_STRICT_UTF8_CHAR_utf8_no_length_checks(s) \
355( ( 0xC2 <= ((const U8*)s)[0] && ((const U8*)s)[0] <= 0xDF ) ? \
356 ( LIKELY( ( ((const U8*)s)[1] & 0xC0 ) == 0x80 ) ? 2 : 0 ) \
357: ( 0xE0 == ((const U8*)s)[0] ) ? \
358 ( LIKELY( ( ( ((const U8*)s)[1] & 0xE0 ) == 0xA0 ) && ( ( ((const U8*)s)[2] & 0xC0 ) == 0x80 ) ) ? 3 : 0 )\
359: ( ( 0xE1 <= ((const U8*)s)[0] && ((const U8*)s)[0] <= 0xEC ) || 0xEE == ((const U8*)s)[0] ) ?\
360 ( ( ( ( ((const U8*)s)[1] & 0xC0 ) == 0x80 ) && ( ( ((const U8*)s)[2] & 0xC0 ) == 0x80 ) ) ? 3 : 0 )\
361: ( 0xED == ((const U8*)s)[0] ) ? \
362 ( LIKELY( ( ( ((const U8*)s)[1] & 0xE0 ) == 0x80 ) && ( ( ((const U8*)s)[2] & 0xC0 ) == 0x80 ) ) ? 3 : 0 )\
363: ( 0xEF == ((const U8*)s)[0] ) ? \
364 ( ( ( 0x80 <= ((const U8*)s)[1] && ((const U8*)s)[1] <= 0xB6 ) || ( 0xB8 <= ((const U8*)s)[1] && ((const U8*)s)[1] <= 0xBE ) ) ?\
365 ( LIKELY( ( ((const U8*)s)[2] & 0xC0 ) == 0x80 ) ? 3 : 0 ) \
366 : ( 0xB7 == ((const U8*)s)[1] ) ? \
367 ( LIKELY( ( ((const U8*)s)[2] & 0xF0 ) == 0x80 || ( ((const U8*)s)[2] & 0xF0 ) == 0xB0 ) ? 3 : 0 )\
368 : ( ( 0xBF == ((const U8*)s)[1] ) && ( 0x80 <= ((const U8*)s)[2] && ((const U8*)s)[2] <= 0xBD ) ) ? 3 : 0 )\
369: ( 0xF0 == ((const U8*)s)[0] ) ? \
370 ( ( ( 0x90 <= ((const U8*)s)[1] && ((const U8*)s)[1] <= 0x9E ) || ( 0xA0 <= ((const U8*)s)[1] && ((const U8*)s)[1] <= 0xAE ) || ( 0xB0 <= ((const U8*)s)[1] && ((const U8*)s)[1] <= 0xBE ) ) ?\
371 ( LIKELY( ( ( ((const U8*)s)[2] & 0xC0 ) == 0x80 ) && ( ( ((const U8*)s)[3] & 0xC0 ) == 0x80 ) ) ? 4 : 0 )\
372 : ( ((const U8*)s)[1] == 0x9F || ( ( ((const U8*)s)[1] & 0xEF ) == 0xAF ) ) ? \
373 ( ( 0x80 <= ((const U8*)s)[2] && ((const U8*)s)[2] <= 0xBE ) ? \
374 ( LIKELY( ( ((const U8*)s)[3] & 0xC0 ) == 0x80 ) ? 4 : 0 ) \
375 : LIKELY( ( 0xBF == ((const U8*)s)[2] ) && ( 0x80 <= ((const U8*)s)[3] && ((const U8*)s)[3] <= 0xBD ) ) ? 4 : 0 )\
376 : 0 ) \
377: ( 0xF1 <= ((const U8*)s)[0] && ((const U8*)s)[0] <= 0xF3 ) ? \
378 ( ( ( ( ((const U8*)s)[1] & 0xC8 ) == 0x80 ) || ( ( ((const U8*)s)[1] & 0xCC ) == 0x88 ) || ( ( ((const U8*)s)[1] & 0xCE ) == 0x8C ) || ( ( ((const U8*)s)[1] & 0xCF ) == 0x8E ) ) ?\
379 ( LIKELY( ( ( ((const U8*)s)[2] & 0xC0 ) == 0x80 ) && ( ( ((const U8*)s)[3] & 0xC0 ) == 0x80 ) ) ? 4 : 0 )\
380 : ( ( ((const U8*)s)[1] & 0xCF ) == 0x8F ) ? \
381 ( ( 0x80 <= ((const U8*)s)[2] && ((const U8*)s)[2] <= 0xBE ) ? \
382 ( LIKELY( ( ((const U8*)s)[3] & 0xC0 ) == 0x80 ) ? 4 : 0 ) \
383 : LIKELY( ( 0xBF == ((const U8*)s)[2] ) && ( 0x80 <= ((const U8*)s)[3] && ((const U8*)s)[3] <= 0xBD ) ) ? 4 : 0 )\
384 : 0 ) \
385: ( 0xF4 == ((const U8*)s)[0] ) ? \
386 ( ( 0x80 <= ((const U8*)s)[1] && ((const U8*)s)[1] <= 0x8E ) ? \
387 ( LIKELY( ( ( ((const U8*)s)[2] & 0xC0 ) == 0x80 ) && ( ( ((const U8*)s)[3] & 0xC0 ) == 0x80 ) ) ? 4 : 0 )\
388 : ( 0x8F == ((const U8*)s)[1] ) ? \
389 ( ( 0x80 <= ((const U8*)s)[2] && ((const U8*)s)[2] <= 0xBE ) ? \
390 ( LIKELY( ( ((const U8*)s)[3] & 0xC0 ) == 0x80 ) ? 4 : 0 ) \
391 : LIKELY( ( 0xBF == ((const U8*)s)[2] ) && ( 0x80 <= ((const U8*)s)[3] && ((const U8*)s)[3] <= 0xBD ) ) ? 4 : 0 )\
392 : 0 ) \
393: 0 )
394
395/* Similarly,
396 C9_STRICT_UTF8_CHAR: Matches legal Unicode UTF-8 variant code
397 points, no surrogates
398 0x0080 - 0xD7FF
399 0xE000 - 0x10FFFF
400*/
401/*** GENERATED CODE ***/
402#define is_C9_STRICT_UTF8_CHAR_utf8_no_length_checks(s) \
403( ( 0xC2 <= ((const U8*)s)[0] && ((const U8*)s)[0] <= 0xDF ) ? \
404 ( LIKELY( ( ((const U8*)s)[1] & 0xC0 ) == 0x80 ) ? 2 : 0 ) \
405: ( 0xE0 == ((const U8*)s)[0] ) ? \
406 ( LIKELY( ( ( ((const U8*)s)[1] & 0xE0 ) == 0xA0 ) && ( ( ((const U8*)s)[2] & 0xC0 ) == 0x80 ) ) ? 3 : 0 )\
407: ( ( 0xE1 <= ((const U8*)s)[0] && ((const U8*)s)[0] <= 0xEC ) || ( ((const U8*)s)[0] & 0xFE ) == 0xEE ) ?\
408 ( LIKELY( ( ( ((const U8*)s)[1] & 0xC0 ) == 0x80 ) && ( ( ((const U8*)s)[2] & 0xC0 ) == 0x80 ) ) ? 3 : 0 )\
409: ( 0xED == ((const U8*)s)[0] ) ? \
410 ( LIKELY( ( ( ((const U8*)s)[1] & 0xE0 ) == 0x80 ) && ( ( ((const U8*)s)[2] & 0xC0 ) == 0x80 ) ) ? 3 : 0 )\
411: ( 0xF0 == ((const U8*)s)[0] ) ? \
412 ( LIKELY( ( ( 0x90 <= ((const U8*)s)[1] && ((const U8*)s)[1] <= 0xBF ) && ( ( ((const U8*)s)[2] & 0xC0 ) == 0x80 ) ) && ( ( ((const U8*)s)[3] & 0xC0 ) == 0x80 ) ) ? 4 : 0 )\
413: ( 0xF1 <= ((const U8*)s)[0] && ((const U8*)s)[0] <= 0xF3 ) ? \
414 ( LIKELY( ( ( ( ((const U8*)s)[1] & 0xC0 ) == 0x80 ) && ( ( ((const U8*)s)[2] & 0xC0 ) == 0x80 ) ) && ( ( ((const U8*)s)[3] & 0xC0 ) == 0x80 ) ) ? 4 : 0 )\
415: LIKELY( ( ( ( 0xF4 == ((const U8*)s)[0] ) && ( ( ((const U8*)s)[1] & 0xF0 ) == 0x80 ) ) && ( ( ((const U8*)s)[2] & 0xC0 ) == 0x80 ) ) && ( ( ((const U8*)s)[3] & 0xC0 ) == 0x80 ) ) ? 4 : 0 )
416
417#endif /* EBCDIC vs ASCII */
418
419/* 2**UTF_ACCUMULATION_SHIFT - 1 */
420#define UTF_CONTINUATION_MASK ((U8) ((1U << UTF_ACCUMULATION_SHIFT) - 1))
421
422/* Internal macro to be used only in this file to aid in constructing other
423 * publicly accessible macros.
424 * The number of bytes required to express this uv in UTF-8, for just those
425 * uv's requiring 2 through 6 bytes, as these are common to all platforms and
426 * word sizes. The number of bytes needed is given by the number of leading 1
427 * bits in the start byte. There are 32 start bytes that have 2 initial 1 bits
428 * (C0-DF); there are 16 that have 3 initial 1 bits (E0-EF); 8 that have 4
429 * initial 1 bits (F0-F8); 4 that have 5 initial 1 bits (F9-FB), and 2 that
430 * have 6 initial 1 bits (FC-FD). The largest number a string of n bytes can
431 * represent is (the number of possible start bytes for 'n')
432 * * (the number of possiblities for each start byte
433 * The latter in turn is
434 * 2 ** ( (how many continuation bytes there are)
435 * * (the number of bits of information each
436 * continuation byte holds))
437 *
438 * If we were on a platform where we could use a fast find first set bit
439 * instruction (or count leading zeros instruction) this could be replaced by
440 * using that to find the log2 of the uv, and divide that by the number of bits
441 * of information in each continuation byte, adjusting for large cases and how
442 * much information is in a start byte for that length */
443#define __COMMON_UNI_SKIP(uv) \
444 (UV) (uv) < (32 * (1U << ( UTF_ACCUMULATION_SHIFT))) ? 2 : \
445 (UV) (uv) < (16 * (1U << (2 * UTF_ACCUMULATION_SHIFT))) ? 3 : \
446 (UV) (uv) < ( 8 * (1U << (3 * UTF_ACCUMULATION_SHIFT))) ? 4 : \
447 (UV) (uv) < ( 4 * (1U << (4 * UTF_ACCUMULATION_SHIFT))) ? 5 : \
448 (UV) (uv) < ( 2 * (1U << (5 * UTF_ACCUMULATION_SHIFT))) ? 6 :
449
450/* Internal macro to be used only in this file.
451 * This adds to __COMMON_UNI_SKIP the details at this platform's upper range.
452 * For any-sized EBCDIC platforms, or 64-bit ASCII ones, we need one more test
453 * to see if just 7 bytes is needed, or if the maximum is needed. For 32-bit
454 * ASCII platforms, everything is representable by 7 bytes */
455#if defined(UV_IS_QUAD) || defined(EBCDIC)
456# define __BASE_UNI_SKIP(uv) (__COMMON_UNI_SKIP(uv) \
457 (UV) (uv) < ((UV) 1U << (6 * UTF_ACCUMULATION_SHIFT)) ? 7 : UTF8_MAXBYTES)
458#else
459# define __BASE_UNI_SKIP(uv) (__COMMON_UNI_SKIP(uv) 7)
460#endif
461
462/* The next two macros use the base macro defined above, and add in the tests
463 * at the low-end of the range, for just 1 byte, yielding complete macros,
464 * publicly accessible. */
465
466/* Input is a true Unicode (not-native) code point */
467#define OFFUNISKIP(uv) (OFFUNI_IS_INVARIANT(uv) ? 1 : __BASE_UNI_SKIP(uv))
468
469/*
470
471=for apidoc Am|STRLEN|UVCHR_SKIP|UV cp
472returns the number of bytes required to represent the code point C<cp> when
473encoded as UTF-8. C<cp> is a native (ASCII or EBCDIC) code point if less than
474255; a Unicode code point otherwise.
475
476=cut
477 */
478#define UVCHR_SKIP(uv) ( UVCHR_IS_INVARIANT(uv) ? 1 : __BASE_UNI_SKIP(uv))
479
480/* The largest code point representable by two UTF-8 bytes on this platform.
481 * As explained in the comments for __COMMON_UNI_SKIP, 32 start bytes with
482 * UTF_ACCUMULATION_SHIFT bits of information each */
483#define MAX_UTF8_TWO_BYTE (32 * (1U << UTF_ACCUMULATION_SHIFT) - 1)
484
485/* The largest code point representable by two UTF-8 bytes on any platform that
486 * Perl runs on. This value is constrained by EBCDIC which has 5 bits per
487 * continuation byte */
488#define MAX_PORTABLE_UTF8_TWO_BYTE (32 * (1U << 5) - 1)
489
490/* The maximum number of UTF-8 bytes a single Unicode character can
491 * uppercase/lowercase/fold into. Unicode guarantees that the maximum
492 * expansion is UTF8_MAX_FOLD_CHAR_EXPAND characters, but any above-Unicode
493 * code point will fold to itself, so we only have to look at the expansion of
494 * the maximum Unicode code point. But this number may be less than the space
495 * occupied by a very large code point under Perl's extended UTF-8. We have to
496 * make it large enough to fit any single character. (It turns out that ASCII
497 * and EBCDIC differ in which is larger) */
498#define UTF8_MAXBYTES_CASE \
499 (UTF8_MAXBYTES >= (UTF8_MAX_FOLD_CHAR_EXPAND * OFFUNISKIP(0x10FFFF)) \
500 ? UTF8_MAXBYTES \
501 : (UTF8_MAX_FOLD_CHAR_EXPAND * OFFUNISKIP(0x10FFFF)))
502
503/* Rest of these are attributes of Unicode and perl's internals rather than the
504 * encoding, or happen to be the same in both ASCII and EBCDIC (at least at
505 * this level; the macros that some of these call may have different
506 * definitions in the two encodings */
507
508/* In domain restricted to ASCII, these may make more sense to the reader than
509 * the ones with Latin1 in the name */
510#define NATIVE_TO_ASCII(ch) NATIVE_TO_LATIN1(ch)
511#define ASCII_TO_NATIVE(ch) LATIN1_TO_NATIVE(ch)
512
513/* More or less misleadingly-named defines, retained for back compat */
514#define NATIVE_TO_UTF(ch) NATIVE_UTF8_TO_I8(ch)
515#define NATIVE_TO_I8(ch) NATIVE_UTF8_TO_I8(ch)
516#define UTF_TO_NATIVE(ch) I8_TO_NATIVE_UTF8(ch)
517#define I8_TO_NATIVE(ch) I8_TO_NATIVE_UTF8(ch)
518#define NATIVE8_TO_UNI(ch) NATIVE_TO_LATIN1(ch)
519
520/* This defines the 1-bits that are to be in the first byte of a multi-byte
521 * UTF-8 encoded character that mark it as a start byte and give the number of
522 * bytes that comprise the character. 'len' is the number of bytes in the
523 * multi-byte sequence. */
524#define UTF_START_MARK(len) (((len) > 7) ? 0xFF : (0xFF & (0xFE << (7-(len)))))
525
526/* Masks out the initial one bits in a start byte, leaving the real data ones.
527 * Doesn't work on an invariant byte. 'len' is the number of bytes in the
528 * multi-byte sequence that comprises the character. */
529#define UTF_START_MASK(len) (((len) >= 7) ? 0x00 : (0x1F >> ((len)-2)))
530
531/* Adds a UTF8 continuation byte 'new' of information to a running total code
532 * point 'old' of all the continuation bytes so far. This is designed to be
533 * used in a loop to convert from UTF-8 to the code point represented. Note
534 * that this is asymmetric on EBCDIC platforms, in that the 'new' parameter is
535 * the UTF-EBCDIC byte, whereas the 'old' parameter is a Unicode (not EBCDIC)
536 * code point in process of being generated */
537#define UTF8_ACCUMULATE(old, new) (__ASSERT_(FITS_IN_8_BITS(new)) \
538 ((old) << UTF_ACCUMULATION_SHIFT) \
539 | ((NATIVE_UTF8_TO_I8((U8)new)) \
540 & UTF_CONTINUATION_MASK))
541
542/* This works in the face of malformed UTF-8. */
543#define UTF8_IS_NEXT_CHAR_DOWNGRADEABLE(s, e) \
544 ( UTF8_IS_DOWNGRADEABLE_START(*(s)) \
545 && ( (e) - (s) > 1) \
546 && UTF8_IS_CONTINUATION(*((s)+1)))
547
548/* Number of bytes a code point occupies in UTF-8. */
549#define NATIVE_SKIP(uv) UVCHR_SKIP(uv)
550
551/* Most code which says UNISKIP is really thinking in terms of native code
552 * points (0-255) plus all those beyond. This is an imprecise term, but having
553 * it means existing code continues to work. For precision, use UVCHR_SKIP,
554 * NATIVE_SKIP, or OFFUNISKIP */
555#define UNISKIP(uv) UVCHR_SKIP(uv)
556
557/* Longer, but more accurate name */
558#define UTF8_IS_ABOVE_LATIN1_START(c) UTF8_IS_ABOVE_LATIN1(c)
559
560/* Convert a UTF-8 variant Latin1 character to a native code point value.
561 * Needs just one iteration of accumulate. Should be used only if it is known
562 * that the code point is < 256, and is not UTF-8 invariant. Use the slower
563 * but more general TWO_BYTE_UTF8_TO_NATIVE() which handles any code point
564 * representable by two bytes (which turns out to be up through
565 * MAX_PORTABLE_UTF8_TWO_BYTE). The two parameters are:
566 * HI: a downgradable start byte;
567 * LO: continuation.
568 * */
569#define EIGHT_BIT_UTF8_TO_NATIVE(HI, LO) \
570 ( __ASSERT_(UTF8_IS_DOWNGRADEABLE_START(HI)) \
571 __ASSERT_(UTF8_IS_CONTINUATION(LO)) \
572 LATIN1_TO_NATIVE(UTF8_ACCUMULATE(( \
573 NATIVE_UTF8_TO_I8(HI) & UTF_START_MASK(2)), (LO))))
574
575/* Convert a two (not one) byte utf8 character to a native code point value.
576 * Needs just one iteration of accumulate. Should not be used unless it is
577 * known that the two bytes are legal: 1) two-byte start, and 2) continuation.
578 * Note that the result can be larger than 255 if the input character is not
579 * downgradable */
580#define TWO_BYTE_UTF8_TO_NATIVE(HI, LO) \
581 (__ASSERT_(FITS_IN_8_BITS(HI)) \
582 __ASSERT_(FITS_IN_8_BITS(LO)) \
583 __ASSERT_(PL_utf8skip[HI] == 2) \
584 __ASSERT_(UTF8_IS_CONTINUATION(LO)) \
585 UNI_TO_NATIVE(UTF8_ACCUMULATE((NATIVE_UTF8_TO_I8(HI) & UTF_START_MASK(2)), \
586 (LO))))
587
588/* Should never be used, and be deprecated */
589#define TWO_BYTE_UTF8_TO_UNI(HI, LO) NATIVE_TO_UNI(TWO_BYTE_UTF8_TO_NATIVE(HI, LO))
590
591/*
592
593=for apidoc Am|STRLEN|UTF8SKIP|char* s
594returns the number of bytes in the UTF-8 encoded character whose first (perhaps
595only) byte is pointed to by C<s>.
596
597=cut
598 */
599#define UTF8SKIP(s) PL_utf8skip[*(const U8*)(s)]
600#define UTF8_SKIP(s) UTF8SKIP(s)
601
602/* Most code that says 'UNI_' really means the native value for code points up
603 * through 255 */
604#define UNI_IS_INVARIANT(cp) UVCHR_IS_INVARIANT(cp)
605
606/*
607=for apidoc Am|bool|UTF8_IS_INVARIANT|char c
608
609Evaluates to 1 if the byte C<c> represents the same character when encoded in
610UTF-8 as when not; otherwise evaluates to 0. UTF-8 invariant characters can be
611copied as-is when converting to/from UTF-8, saving time.
612
613In spite of the name, this macro gives the correct result if the input string
614from which C<c> comes is not encoded in UTF-8.
615
616See C<L</UVCHR_IS_INVARIANT>> for checking if a UV is invariant.
617
618=cut
619
620The reason it works on both UTF-8 encoded strings and non-UTF-8 encoded, is
621that it returns TRUE in each for the exact same set of bit patterns. It is
622valid on a subset of what UVCHR_IS_INVARIANT is valid on, so can just use that;
623and the compiler should optimize out anything extraneous given the
624implementation of the latter. The |0 makes sure this isn't mistakenly called
625with a ptr argument.
626*/
627#define UTF8_IS_INVARIANT(c) UVCHR_IS_INVARIANT((c) | 0)
628
629/* Like the above, but its name implies a non-UTF8 input, which as the comments
630 * above show, doesn't matter as to its implementation */
631#define NATIVE_BYTE_IS_INVARIANT(c) UVCHR_IS_INVARIANT(c)
632
633/* The macros in the next 4 sets are used to generate the two utf8 or utfebcdic
634 * bytes from an ordinal that is known to fit into exactly two (not one) bytes;
635 * it must be less than 0x3FF to work across both encodings. */
636
637/* These two are helper macros for the other three sets, and should not be used
638 * directly anywhere else. 'translate_function' is either NATIVE_TO_LATIN1
639 * (which works for code points up through 0xFF) or NATIVE_TO_UNI which works
640 * for any code point */
641#define __BASE_TWO_BYTE_HI(c, translate_function) \
642 (__ASSERT_(! UVCHR_IS_INVARIANT(c)) \
643 I8_TO_NATIVE_UTF8((translate_function(c) >> UTF_ACCUMULATION_SHIFT) \
644 | UTF_START_MARK(2)))
645#define __BASE_TWO_BYTE_LO(c, translate_function) \
646 (__ASSERT_(! UVCHR_IS_INVARIANT(c)) \
647 I8_TO_NATIVE_UTF8((translate_function(c) & UTF_CONTINUATION_MASK) \
648 | UTF_CONTINUATION_MARK))
649
650/* The next two macros should not be used. They were designed to be usable as
651 * the case label of a switch statement, but this doesn't work for EBCDIC. Use
652 * regen/unicode_constants.pl instead */
653#define UTF8_TWO_BYTE_HI_nocast(c) __BASE_TWO_BYTE_HI(c, NATIVE_TO_UNI)
654#define UTF8_TWO_BYTE_LO_nocast(c) __BASE_TWO_BYTE_LO(c, NATIVE_TO_UNI)
655
656/* The next two macros are used when the source should be a single byte
657 * character; checked for under DEBUGGING */
658#define UTF8_EIGHT_BIT_HI(c) (__ASSERT_(FITS_IN_8_BITS(c)) \
659 ( __BASE_TWO_BYTE_HI(c, NATIVE_TO_LATIN1)))
660#define UTF8_EIGHT_BIT_LO(c) (__ASSERT_(FITS_IN_8_BITS(c)) \
661 (__BASE_TWO_BYTE_LO(c, NATIVE_TO_LATIN1)))
662
663/* These final two macros in the series are used when the source can be any
664 * code point whose UTF-8 is known to occupy 2 bytes; they are less efficient
665 * than the EIGHT_BIT versions on EBCDIC platforms. We use the logical '~'
666 * operator instead of "<=" to avoid getting compiler warnings.
667 * MAX_UTF8_TWO_BYTE should be exactly all one bits in the lower few
668 * places, so the ~ works */
669#define UTF8_TWO_BYTE_HI(c) \
670 (__ASSERT_((sizeof(c) == 1) \
671 || !(((WIDEST_UTYPE)(c)) & ~MAX_UTF8_TWO_BYTE)) \
672 (__BASE_TWO_BYTE_HI(c, NATIVE_TO_UNI)))
673#define UTF8_TWO_BYTE_LO(c) \
674 (__ASSERT_((sizeof(c) == 1) \
675 || !(((WIDEST_UTYPE)(c)) & ~MAX_UTF8_TWO_BYTE)) \
676 (__BASE_TWO_BYTE_LO(c, NATIVE_TO_UNI)))
677
678/* This is illegal in any well-formed UTF-8 in both EBCDIC and ASCII
679 * as it is only in overlongs. */
680#define ILLEGAL_UTF8_BYTE I8_TO_NATIVE_UTF8(0xC1)
681
682/*
683 * 'UTF' is whether or not p is encoded in UTF8. The names 'foo_lazy_if' stem
684 * from an earlier version of these macros in which they didn't call the
685 * foo_utf8() macros (i.e. were 'lazy') unless they decided that *p is the
686 * beginning of a utf8 character. Now that foo_utf8() determines that itself,
687 * no need to do it again here
688 */
689#define isIDFIRST_lazy_if(p,UTF) \
690 _is_utf8_FOO(_CC_IDFIRST, (const U8 *) p, "isIDFIRST_lazy_if", \
691 "isIDFIRST_lazy_if_safe", \
692 cBOOL(UTF && ! IN_BYTES), 0, __FILE__,__LINE__)
693
694#define isIDFIRST_lazy_if_safe(p, e, UTF) \
695 ((IN_BYTES || !UTF) \
696 ? isIDFIRST(*(p)) \
697 : isIDFIRST_utf8_safe(p, e))
698
699#define isWORDCHAR_lazy_if(p,UTF) \
700 _is_utf8_FOO(_CC_IDFIRST, (const U8 *) p, "isWORDCHAR_lazy_if", \
701 "isWORDCHAR_lazy_if_safe", \
702 cBOOL(UTF && ! IN_BYTES), 0, __FILE__,__LINE__)
703
704#define isWORDCHAR_lazy_if_safe(p, e, UTF) \
705 ((IN_BYTES || !UTF) \
706 ? isWORDCHAR(*(p)) \
707 : isWORDCHAR_utf8_safe((U8 *) p, (U8 *) e))
708
709#define isALNUM_lazy_if(p,UTF) \
710 _is_utf8_FOO(_CC_IDFIRST, (const U8 *) p, "isALNUM_lazy_if", \
711 "isWORDCHAR_lazy_if_safe", \
712 cBOOL(UTF && ! IN_BYTES), 0, __FILE__,__LINE__)
713
714#define UTF8_MAXLEN UTF8_MAXBYTES
715
716/* A Unicode character can fold to up to 3 characters */
717#define UTF8_MAX_FOLD_CHAR_EXPAND 3
718
719#define IN_BYTES UNLIKELY(CopHINTS_get(PL_curcop) & HINT_BYTES)
720
721/*
722
723=for apidoc Am|bool|DO_UTF8|SV* sv
724Returns a bool giving whether or not the PV in C<sv> is to be treated as being
725encoded in UTF-8.
726
727You should use this I<after> a call to C<SvPV()> or one of its variants, in
728case any call to string overloading updates the internal UTF-8 encoding flag.
729
730=cut
731*/
732#define DO_UTF8(sv) (SvUTF8(sv) && !IN_BYTES)
733
734/* Should all strings be treated as Unicode, and not just UTF-8 encoded ones?
735 * Is so within 'feature unicode_strings' or 'locale :not_characters', and not
736 * within 'use bytes'. UTF-8 locales are not tested for here, but perhaps
737 * could be */
738#define IN_UNI_8_BIT \
739 (( ( (CopHINTS_get(PL_curcop) & HINT_UNI_8_BIT)) \
740 || ( CopHINTS_get(PL_curcop) & HINT_LOCALE_PARTIAL \
741 /* -1 below is for :not_characters */ \
742 && _is_in_locale_category(FALSE, -1))) \
743 && (! IN_BYTES))
744
745
746#define UTF8_ALLOW_EMPTY 0x0001 /* Allow a zero length string */
747#define UTF8_GOT_EMPTY UTF8_ALLOW_EMPTY
748
749/* Allow first byte to be a continuation byte */
750#define UTF8_ALLOW_CONTINUATION 0x0002
751#define UTF8_GOT_CONTINUATION UTF8_ALLOW_CONTINUATION
752
753/* Unexpected continuation byte */
754#define UTF8_ALLOW_NON_CONTINUATION 0x0004
755#define UTF8_GOT_NON_CONTINUATION UTF8_ALLOW_NON_CONTINUATION
756
757/* expecting more bytes than were available in the string */
758#define UTF8_ALLOW_SHORT 0x0008
759#define UTF8_GOT_SHORT UTF8_ALLOW_SHORT
760
761/* Overlong sequence; i.e., the code point can be specified in fewer bytes.
762 * First one will convert the overlong to the REPLACEMENT CHARACTER; second
763 * will return what the overlong evaluates to */
764#define UTF8_ALLOW_LONG 0x0010
765#define UTF8_ALLOW_LONG_AND_ITS_VALUE (UTF8_ALLOW_LONG|0x0020)
766#define UTF8_GOT_LONG UTF8_ALLOW_LONG
767
768#define UTF8_ALLOW_OVERFLOW 0x0080
769#define UTF8_GOT_OVERFLOW UTF8_ALLOW_OVERFLOW
770
771#define UTF8_DISALLOW_SURROGATE 0x0100 /* Unicode surrogates */
772#define UTF8_GOT_SURROGATE UTF8_DISALLOW_SURROGATE
773#define UTF8_WARN_SURROGATE 0x0200
774
775#define UTF8_DISALLOW_NONCHAR 0x0400 /* Unicode non-character */
776#define UTF8_GOT_NONCHAR UTF8_DISALLOW_NONCHAR
777#define UTF8_WARN_NONCHAR 0x0800 /* code points */
778
779#define UTF8_DISALLOW_SUPER 0x1000 /* Super-set of Unicode: code */
780#define UTF8_GOT_SUPER UTF8_DISALLOW_SUPER
781#define UTF8_WARN_SUPER 0x2000 /* points above the legal max */
782
783/* Code points which never were part of the original UTF-8 standard, which only
784 * went up to 2 ** 31 - 1. Note that these all overflow a signed 32-bit word,
785 * The first byte of these code points is FE or FF on ASCII platforms. If the
786 * first byte is FF, it will overflow a 32-bit word. */
787#define UTF8_DISALLOW_ABOVE_31_BIT 0x4000
788#define UTF8_GOT_ABOVE_31_BIT UTF8_DISALLOW_ABOVE_31_BIT
789#define UTF8_WARN_ABOVE_31_BIT 0x8000
790
791/* For back compat, these old names are misleading for UTF_EBCDIC */
792#define UTF8_DISALLOW_FE_FF UTF8_DISALLOW_ABOVE_31_BIT
793#define UTF8_WARN_FE_FF UTF8_WARN_ABOVE_31_BIT
794
795#define UTF8_CHECK_ONLY 0x10000
796#define _UTF8_NO_CONFIDENCE_IN_CURLEN 0x20000 /* Internal core use only */
797
798/* For backwards source compatibility. They do nothing, as the default now
799 * includes what they used to mean. The first one's meaning was to allow the
800 * just the single non-character 0xFFFF */
801#define UTF8_ALLOW_FFFF 0
802#define UTF8_ALLOW_SURROGATE 0
803
804/* C9 refers to Unicode Corrigendum #9: allows but discourages non-chars */
805#define UTF8_DISALLOW_ILLEGAL_C9_INTERCHANGE \
806 (UTF8_DISALLOW_SUPER|UTF8_DISALLOW_SURROGATE)
807#define UTF8_WARN_ILLEGAL_C9_INTERCHANGE (UTF8_WARN_SUPER|UTF8_WARN_SURROGATE)
808
809#define UTF8_DISALLOW_ILLEGAL_INTERCHANGE \
810 (UTF8_DISALLOW_ILLEGAL_C9_INTERCHANGE|UTF8_DISALLOW_NONCHAR)
811#define UTF8_WARN_ILLEGAL_INTERCHANGE \
812 (UTF8_WARN_ILLEGAL_C9_INTERCHANGE|UTF8_WARN_NONCHAR)
813
814/* This is typically used for code that processes UTF-8 input and doesn't want
815 * to have to deal with any malformations that might be present. All such will
816 * be safely replaced by the REPLACEMENT CHARACTER, unless other flags
817 * overriding this are also present. */
818#define UTF8_ALLOW_ANY ( UTF8_ALLOW_CONTINUATION \
819 |UTF8_ALLOW_NON_CONTINUATION \
820 |UTF8_ALLOW_SHORT \
821 |UTF8_ALLOW_LONG \
822 |UTF8_ALLOW_OVERFLOW)
823
824/* Accept any Perl-extended UTF-8 that evaluates to any UV on the platform, but
825 * not any malformed. This is the default. (Note that UVs above IV_MAX are
826 * deprecated. */
827#define UTF8_ALLOW_ANYUV 0
828#define UTF8_ALLOW_DEFAULT UTF8_ALLOW_ANYUV
829
830/*
831=for apidoc Am|bool|UTF8_IS_SURROGATE|const U8 *s|const U8 *e
832
833Evaluates to non-zero if the first few bytes of the string starting at C<s> and
834looking no further than S<C<e - 1>> are well-formed UTF-8 that represents one
835of the Unicode surrogate code points; otherwise it evaluates to 0. If
836non-zero, the value gives how many bytes starting at C<s> comprise the code
837point's representation.
838
839=cut
840 */
841#define UTF8_IS_SURROGATE(s, e) is_SURROGATE_utf8_safe(s, e)
842
843
844#define UTF8_IS_REPLACEMENT(s, send) is_REPLACEMENT_utf8_safe(s,send)
845
846/*
847=for apidoc Am|bool|UTF8_IS_SUPER|const U8 *s|const U8 *e
848
849Recall that Perl recognizes an extension to UTF-8 that can encode code
850points larger than the ones defined by Unicode, which are 0..0x10FFFF.
851
852This macro evaluates to non-zero if the first few bytes of the string starting
853at C<s> and looking no further than S<C<e - 1>> are from this UTF-8 extension;
854otherwise it evaluates to 0. If non-zero, the value gives how many bytes
855starting at C<s> comprise the code point's representation.
856
8570 is returned if the bytes are not well-formed extended UTF-8, or if they
858represent a code point that cannot fit in a UV on the current platform. Hence
859this macro can give different results when run on a 64-bit word machine than on
860one with a 32-bit word size.
861
862Note that it is deprecated to have code points that are larger than what can
863fit in an IV on the current machine.
864
865=cut
866
867 * ASCII EBCDIC I8
868 * U+10FFFF: \xF4\x8F\xBF\xBF \xF9\xA1\xBF\xBF\xBF max legal Unicode
869 * U+110000: \xF4\x90\x80\x80 \xF9\xA2\xA0\xA0\xA0
870 * U+110001: \xF4\x90\x80\x81 \xF9\xA2\xA0\xA0\xA1
871 */
872#ifdef EBCDIC
873# define UTF8_IS_SUPER(s, e) \
874 (( LIKELY((e) > (s) + 4) \
875 && NATIVE_UTF8_TO_I8(*(s)) >= 0xF9 \
876 && ( NATIVE_UTF8_TO_I8(*(s)) > 0xF9 \
877 || (NATIVE_UTF8_TO_I8(*((s) + 1)) >= 0xA2)) \
878 && LIKELY((s) + UTF8SKIP(s) <= (e))) \
879 ? _is_utf8_char_helper(s, s + UTF8SKIP(s), 0) : 0)
880#else
881# define UTF8_IS_SUPER(s, e) \
882 (( LIKELY((e) > (s) + 3) \
883 && (*(U8*) (s)) >= 0xF4 \
884 && ((*(U8*) (s)) > 0xF4 || (*((U8*) (s) + 1) >= 0x90))\
885 && LIKELY((s) + UTF8SKIP(s) <= (e))) \
886 ? _is_utf8_char_helper(s, s + UTF8SKIP(s), 0) : 0)
887#endif
888
889/* These are now machine generated, and the 'given' clause is no longer
890 * applicable */
891#define UTF8_IS_NONCHAR_GIVEN_THAT_NON_SUPER_AND_GE_PROBLEMATIC(s, e) \
892 cBOOL(is_NONCHAR_utf8_safe(s,e))
893
894/*
895=for apidoc Am|bool|UTF8_IS_NONCHAR|const U8 *s|const U8 *e
896
897Evaluates to non-zero if the first few bytes of the string starting at C<s> and
898looking no further than S<C<e - 1>> are well-formed UTF-8 that represents one
899of the Unicode non-character code points; otherwise it evaluates to 0. If
900non-zero, the value gives how many bytes starting at C<s> comprise the code
901point's representation.
902
903=cut
904 */
905#define UTF8_IS_NONCHAR(s, e) \
906 UTF8_IS_NONCHAR_GIVEN_THAT_NON_SUPER_AND_GE_PROBLEMATIC(s, e)
907
908#define UNICODE_SURROGATE_FIRST 0xD800
909#define UNICODE_SURROGATE_LAST 0xDFFF
910#define UNICODE_REPLACEMENT 0xFFFD
911#define UNICODE_BYTE_ORDER_MARK 0xFEFF
912
913/* Though our UTF-8 encoding can go beyond this,
914 * let's be conservative and do as Unicode says. */
915#define PERL_UNICODE_MAX 0x10FFFF
916
917#define UNICODE_WARN_SURROGATE 0x0001 /* UTF-16 surrogates */
918#define UNICODE_WARN_NONCHAR 0x0002 /* Non-char code points */
919#define UNICODE_WARN_SUPER 0x0004 /* Above 0x10FFFF */
920#define UNICODE_WARN_ABOVE_31_BIT 0x0008 /* Above 0x7FFF_FFFF */
921#define UNICODE_DISALLOW_SURROGATE 0x0010
922#define UNICODE_DISALLOW_NONCHAR 0x0020
923#define UNICODE_DISALLOW_SUPER 0x0040
924#define UNICODE_DISALLOW_ABOVE_31_BIT 0x0080
925#define UNICODE_WARN_ILLEGAL_C9_INTERCHANGE \
926 (UNICODE_WARN_SURROGATE|UNICODE_WARN_SUPER)
927#define UNICODE_WARN_ILLEGAL_INTERCHANGE \
928 (UNICODE_WARN_ILLEGAL_C9_INTERCHANGE|UNICODE_WARN_NONCHAR)
929#define UNICODE_DISALLOW_ILLEGAL_C9_INTERCHANGE \
930 (UNICODE_DISALLOW_SURROGATE|UNICODE_DISALLOW_SUPER)
931#define UNICODE_DISALLOW_ILLEGAL_INTERCHANGE \
932 (UNICODE_DISALLOW_ILLEGAL_C9_INTERCHANGE|UNICODE_DISALLOW_NONCHAR)
933
934/* For backward source compatibility, as are now the default */
935#define UNICODE_ALLOW_SURROGATE 0
936#define UNICODE_ALLOW_SUPER 0
937#define UNICODE_ALLOW_ANY 0
938
939/* This matches the 2048 code points between UNICODE_SURROGATE_FIRST (0xD800) and
940 * UNICODE_SURROGATE_LAST (0xDFFF) */
941#define UNICODE_IS_SURROGATE(uv) (((UV) (uv) & (~0xFFFF | 0xF800)) \
942 == 0xD800)
943
944#define UNICODE_IS_REPLACEMENT(uv) ((UV) (uv) == UNICODE_REPLACEMENT)
945#define UNICODE_IS_BYTE_ORDER_MARK(uv) ((UV) (uv) == UNICODE_BYTE_ORDER_MARK)
946
947/* Is 'uv' one of the 32 contiguous-range noncharacters? */
948#define UNICODE_IS_32_CONTIGUOUS_NONCHARS(uv) ((UV) (uv) >= 0xFDD0 \
949 && (UV) (uv) <= 0xFDEF)
950
951/* Is 'uv' one of the 34 plane-ending noncharacters 0xFFFE, 0xFFFF, 0x1FFFE,
952 * 0x1FFFF, ... 0x10FFFE, 0x10FFFF, given that we know that 'uv' is not above
953 * the Unicode legal max */
954#define UNICODE_IS_END_PLANE_NONCHAR_GIVEN_NOT_SUPER(uv) \
955 (((UV) (uv) & 0xFFFE) == 0xFFFE)
956
957#define UNICODE_IS_NONCHAR(uv) \
958 ( UNICODE_IS_32_CONTIGUOUS_NONCHARS(uv) \
959 || ( LIKELY( ! UNICODE_IS_SUPER(uv)) \
960 && UNICODE_IS_END_PLANE_NONCHAR_GIVEN_NOT_SUPER(uv)))
961
962#define UNICODE_IS_SUPER(uv) ((UV) (uv) > PERL_UNICODE_MAX)
963#define UNICODE_IS_ABOVE_31_BIT(uv) ((UV) (uv) > 0x7FFFFFFF)
964
965#define LATIN_SMALL_LETTER_SHARP_S LATIN_SMALL_LETTER_SHARP_S_NATIVE
966#define LATIN_SMALL_LETTER_Y_WITH_DIAERESIS \
967 LATIN_SMALL_LETTER_Y_WITH_DIAERESIS_NATIVE
968#define MICRO_SIGN MICRO_SIGN_NATIVE
969#define LATIN_CAPITAL_LETTER_A_WITH_RING_ABOVE \
970 LATIN_CAPITAL_LETTER_A_WITH_RING_ABOVE_NATIVE
971#define LATIN_SMALL_LETTER_A_WITH_RING_ABOVE \
972 LATIN_SMALL_LETTER_A_WITH_RING_ABOVE_NATIVE
973#define UNICODE_GREEK_CAPITAL_LETTER_SIGMA 0x03A3
974#define UNICODE_GREEK_SMALL_LETTER_FINAL_SIGMA 0x03C2
975#define UNICODE_GREEK_SMALL_LETTER_SIGMA 0x03C3
976#define GREEK_SMALL_LETTER_MU 0x03BC
977#define GREEK_CAPITAL_LETTER_MU 0x039C /* Upper and title case
978 of MICRON */
979#define LATIN_CAPITAL_LETTER_Y_WITH_DIAERESIS 0x0178 /* Also is title case */
980#ifdef LATIN_CAPITAL_LETTER_SHARP_S_UTF8
981# define LATIN_CAPITAL_LETTER_SHARP_S 0x1E9E
982#endif
983#define LATIN_CAPITAL_LETTER_I_WITH_DOT_ABOVE 0x130
984#define LATIN_SMALL_LETTER_DOTLESS_I 0x131
985#define LATIN_SMALL_LETTER_LONG_S 0x017F
986#define LATIN_SMALL_LIGATURE_LONG_S_T 0xFB05
987#define LATIN_SMALL_LIGATURE_ST 0xFB06
988#define KELVIN_SIGN 0x212A
989#define ANGSTROM_SIGN 0x212B
990
991#define UNI_DISPLAY_ISPRINT 0x0001
992#define UNI_DISPLAY_BACKSLASH 0x0002
993#define UNI_DISPLAY_QQ (UNI_DISPLAY_ISPRINT|UNI_DISPLAY_BACKSLASH)
994#define UNI_DISPLAY_REGEX (UNI_DISPLAY_ISPRINT|UNI_DISPLAY_BACKSLASH)
995
996#define ANYOF_FOLD_SHARP_S(node, input, end) \
997 (ANYOF_BITMAP_TEST(node, LATIN_SMALL_LETTER_SHARP_S) && \
998 (ANYOF_NONBITMAP(node)) && \
999 (ANYOF_FLAGS(node) & ANYOF_LOC_NONBITMAP_FOLD) && \
1000 ((end) > (input) + 1) && \
1001 isALPHA_FOLD_EQ((input)[0], 's'))
1002
1003#define SHARP_S_SKIP 2
1004
1005/*
1006
1007=for apidoc Am|STRLEN|isUTF8_CHAR|const U8 *s|const U8 *e
1008
1009Evaluates to non-zero if the first few bytes of the string starting at C<s> and
1010looking no further than S<C<e - 1>> are well-formed UTF-8, as extended by Perl,
1011that represents some code point; otherwise it evaluates to 0. If non-zero, the
1012value gives how many bytes starting at C<s> comprise the code point's
1013representation. Any bytes remaining before C<e>, but beyond the ones needed to
1014form the first code point in C<s>, are not examined.
1015
1016The code point can be any that will fit in a UV on this machine, using Perl's
1017extension to official UTF-8 to represent those higher than the Unicode maximum
1018of 0x10FFFF. That means that this macro is used to efficiently decide if the
1019next few bytes in C<s> is legal UTF-8 for a single character.
1020
1021Use C<L</isSTRICT_UTF8_CHAR>> to restrict the acceptable code points to those
1022defined by Unicode to be fully interchangeable across applications;
1023C<L</isC9_STRICT_UTF8_CHAR>> to use the L<Unicode Corrigendum
1024#9|http://www.unicode.org/versions/corrigendum9.html> definition of allowable
1025code points; and C<L</isUTF8_CHAR_flags>> for a more customized definition.
1026
1027Use C<L</is_utf8_string>>, C<L</is_utf8_string_loc>>, and
1028C<L</is_utf8_string_loclen>> to check entire strings.
1029
1030Note that it is deprecated to use code points higher than what will fit in an
1031IV. This macro does not raise any warnings for such code points, treating them
1032as valid.
1033
1034Note also that a UTF-8 INVARIANT character (i.e. ASCII on non-EBCDIC machines)
1035is a valid UTF-8 character.
1036
1037=cut
1038*/
1039
1040#define isUTF8_CHAR(s, e) \
1041 (UNLIKELY((e) <= (s)) \
1042 ? 0 \
1043 : (UTF8_IS_INVARIANT(*s)) \
1044 ? 1 \
1045 : UNLIKELY(((e) - (s)) < UTF8SKIP(s)) \
1046 ? 0 \
1047 : LIKELY(NATIVE_UTF8_TO_I8(*s) <= _IS_UTF8_CHAR_HIGHEST_START_BYTE) \
1048 ? is_UTF8_CHAR_utf8_no_length_checks(s) \
1049 : _is_utf8_char_helper(s, e, 0))
1050
1051#define is_utf8_char_buf(buf, buf_end) isUTF8_CHAR(buf, buf_end)
1052#define bytes_from_utf8(s, lenp, is_utf8p) \
1053 bytes_from_utf8_loc(s, lenp, is_utf8p, 0)
1054
1055/*
1056
1057=for apidoc Am|STRLEN|isSTRICT_UTF8_CHAR|const U8 *s|const U8 *e
1058
1059Evaluates to non-zero if the first few bytes of the string starting at C<s> and
1060looking no further than S<C<e - 1>> are well-formed UTF-8 that represents some
1061Unicode code point completely acceptable for open interchange between all
1062applications; otherwise it evaluates to 0. If non-zero, the value gives how
1063many bytes starting at C<s> comprise the code point's representation. Any
1064bytes remaining before C<e>, but beyond the ones needed to form the first code
1065point in C<s>, are not examined.
1066
1067The largest acceptable code point is the Unicode maximum 0x10FFFF, and must not
1068be a surrogate nor a non-character code point. Thus this excludes any code
1069point from Perl's extended UTF-8.
1070
1071This is used to efficiently decide if the next few bytes in C<s> is
1072legal Unicode-acceptable UTF-8 for a single character.
1073
1074Use C<L</isC9_STRICT_UTF8_CHAR>> to use the L<Unicode Corrigendum
1075#9|http://www.unicode.org/versions/corrigendum9.html> definition of allowable
1076code points; C<L</isUTF8_CHAR>> to check for Perl's extended UTF-8;
1077and C<L</isUTF8_CHAR_flags>> for a more customized definition.
1078
1079Use C<L</is_strict_utf8_string>>, C<L</is_strict_utf8_string_loc>>, and
1080C<L</is_strict_utf8_string_loclen>> to check entire strings.
1081
1082=cut
1083*/
1084
1085#define isSTRICT_UTF8_CHAR(s, e) \
1086 (UNLIKELY((e) <= (s)) \
1087 ? 0 \
1088 : (UTF8_IS_INVARIANT(*s)) \
1089 ? 1 \
1090 : UNLIKELY(((e) - (s)) < UTF8SKIP(s)) \
1091 ? 0 \
1092 : is_STRICT_UTF8_CHAR_utf8_no_length_checks(s))
1093
1094/*
1095
1096=for apidoc Am|STRLEN|isC9_STRICT_UTF8_CHAR|const U8 *s|const U8 *e
1097
1098Evaluates to non-zero if the first few bytes of the string starting at C<s> and
1099looking no further than S<C<e - 1>> are well-formed UTF-8 that represents some
1100Unicode non-surrogate code point; otherwise it evaluates to 0. If non-zero,
1101the value gives how many bytes starting at C<s> comprise the code point's
1102representation. Any bytes remaining before C<e>, but beyond the ones needed to
1103form the first code point in C<s>, are not examined.
1104
1105The largest acceptable code point is the Unicode maximum 0x10FFFF. This
1106differs from C<L</isSTRICT_UTF8_CHAR>> only in that it accepts non-character
1107code points. This corresponds to
1108L<Unicode Corrigendum #9|http://www.unicode.org/versions/corrigendum9.html>.
1109which said that non-character code points are merely discouraged rather than
1110completely forbidden in open interchange. See
1111L<perlunicode/Noncharacter code points>.
1112
1113Use C<L</isUTF8_CHAR>> to check for Perl's extended UTF-8; and
1114C<L</isUTF8_CHAR_flags>> for a more customized definition.
1115
1116Use C<L</is_c9strict_utf8_string>>, C<L</is_c9strict_utf8_string_loc>>, and
1117C<L</is_c9strict_utf8_string_loclen>> to check entire strings.
1118
1119=cut
1120*/
1121
1122#define isC9_STRICT_UTF8_CHAR(s, e) \
1123 (UNLIKELY((e) <= (s)) \
1124 ? 0 \
1125 : (UTF8_IS_INVARIANT(*s)) \
1126 ? 1 \
1127 : UNLIKELY(((e) - (s)) < UTF8SKIP(s)) \
1128 ? 0 \
1129 : is_C9_STRICT_UTF8_CHAR_utf8_no_length_checks(s))
1130
1131/*
1132
1133=for apidoc Am|STRLEN|isUTF8_CHAR_flags|const U8 *s|const U8 *e| const U32 flags
1134
1135Evaluates to non-zero if the first few bytes of the string starting at C<s> and
1136looking no further than S<C<e - 1>> are well-formed UTF-8, as extended by Perl,
1137that represents some code point, subject to the restrictions given by C<flags>;
1138otherwise it evaluates to 0. If non-zero, the value gives how many bytes
1139starting at C<s> comprise the code point's representation. Any bytes remaining
1140before C<e>, but beyond the ones needed to form the first code point in C<s>,
1141are not examined.
1142
1143If C<flags> is 0, this gives the same results as C<L</isUTF8_CHAR>>;
1144if C<flags> is C<UTF8_DISALLOW_ILLEGAL_INTERCHANGE>, this gives the same results
1145as C<L</isSTRICT_UTF8_CHAR>>;
1146and if C<flags> is C<UTF8_DISALLOW_ILLEGAL_C9_INTERCHANGE>, this gives
1147the same results as C<L</isC9_STRICT_UTF8_CHAR>>.
1148Otherwise C<flags> may be any combination of the C<UTF8_DISALLOW_I<foo>> flags
1149understood by C<L</utf8n_to_uvchr>>, with the same meanings.
1150
1151The three alternative macros are for the most commonly needed validations; they
1152are likely to run somewhat faster than this more general one, as they can be
1153inlined into your code.
1154
1155Use L</is_utf8_string_flags>, L</is_utf8_string_loc_flags>, and
1156L</is_utf8_string_loclen_flags> to check entire strings.
1157
1158=cut
1159*/
1160
1161#define isUTF8_CHAR_flags(s, e, flags) \
1162 (UNLIKELY((e) <= (s)) \
1163 ? 0 \
1164 : (UTF8_IS_INVARIANT(*s)) \
1165 ? 1 \
1166 : UNLIKELY(((e) - (s)) < UTF8SKIP(s)) \
1167 ? 0 \
1168 : _is_utf8_char_helper(s, e, flags))
1169
1170/* Do not use; should be deprecated. Use isUTF8_CHAR() instead; this is
1171 * retained solely for backwards compatibility */
1172#define IS_UTF8_CHAR(p, n) (isUTF8_CHAR(p, (p) + (n)) == n)
1173
1174#endif /* PERL_UTF8_H_ */
1175
1176/*
1177 * ex: set ts=8 sts=4 sw=4 et:
1178 */
Unnamed repository; edit this file 'description' to name the repository.