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
Rename internal macro and move to utf8.h
[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/*
38=for apidoc is_ascii_string
39
40This is a misleadingly-named synonym for L</is_utf8_invariant_string>.
41On ASCII-ish platforms, the name isn't misleading: the ASCII-range characters
42are exactly the UTF-8 invariants. But EBCDIC machines have more invariants
43than just the ASCII characters, so C<is_utf8_invariant_string> is preferred.
44
45=for apidoc is_invariant_string
46
47This is a somewhat misleadingly-named synonym for L</is_utf8_invariant_string>.
48C<is_utf8_invariant_string> is preferred, as it indicates under what conditions
49the string is invariant.
50
51=cut
52*/
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)
55
56#define uvoffuni_to_utf8_flags(d,uv,flags) \
57 uvoffuni_to_utf8_flags_msgs(d, uv, flags, 0)
58#define uvchr_to_utf8(a,b) uvchr_to_utf8_flags(a,b,0)
59#define uvchr_to_utf8_flags(d,uv,flags) \
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)
63#define utf8_to_uvchr_buf(s, e, lenp) \
64 utf8_to_uvchr_buf_helper((const U8 *) (s), (const U8 *) e, lenp)
65#define utf8n_to_uvchr(s, len, lenp, flags) \
66 utf8n_to_uvchr_error(s, len, lenp, flags, 0)
67#define utf8n_to_uvchr_error(s, len, lenp, flags, errors) \
68 utf8n_to_uvchr_msgs(s, len, lenp, flags, errors, 0)
69
70#define to_uni_fold(c, p, lenp) _to_uni_fold_flags(c, p, lenp, FOLD_FLAGS_FULL)
71
72#define foldEQ_utf8(s1, pe1, l1, u1, s2, pe2, l2, u2) \
73 foldEQ_utf8_flags(s1, pe1, l1, u1, s2, pe2, l2, u2, 0)
74#define FOLDEQ_UTF8_NOMIX_ASCII (1 << 0)
75#define FOLDEQ_LOCALE (1 << 1)
76#define FOLDEQ_S1_ALREADY_FOLDED (1 << 2)
77#define FOLDEQ_S2_ALREADY_FOLDED (1 << 3)
78#define FOLDEQ_S1_FOLDS_SANE (1 << 4)
79#define FOLDEQ_S2_FOLDS_SANE (1 << 5)
80
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
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"
96
97#else /* ! EBCDIC */
98START_EXTERN_C
99
100/*
101
102=for apidoc AmnU|STRLEN|UTF8_MAXBYTES
103
104The maximum width of a single UTF-8 encoded character, in bytes.
105
106NOTE: Strictly speaking Perl's UTF-8 should not be called UTF-8 since UTF-8
107is an encoding of Unicode, and Unicode's upper limit, 0x10FFFF, can be
108expressed with 4 bytes. However, Perl thinks of UTF-8 as a way to encode
109non-negative integers in a binary format, even those above Unicode.
110
111=cut
112 */
113#define UTF8_MAXBYTES 13
114
115#ifdef DOINIT
116EXTCONST unsigned char PL_utf8skip[] = {
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 */
130/* 0xC2 */ 2,2,2,2,2,2,2,2,2,2,2,2,2,2, /* U+0080 to U+03FF */
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 */
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) */
137/* 0xFF */ UTF8_MAXBYTES
138};
139#else
140EXTCONST unsigned char PL_utf8skip[];
141#endif
142
143END_EXTERN_C
144
145/*
146
147=for apidoc Am|U8|NATIVE_TO_LATIN1|U8 ch
148
149Returns the Latin-1 (including ASCII and control characters) equivalent of the
150input native code point given by C<ch>. Thus, C<NATIVE_TO_LATIN1(193)> on
151EBCDIC platforms returns 65. These each represent the character C<"A"> on
152their respective platforms. On ASCII platforms no conversion is needed, so
153this macro expands to just its input, adding no time nor space requirements to
154the implementation.
155
156For conversion of code points potentially larger than will fit in a character,
157use L</NATIVE_TO_UNI>.
158
159=for apidoc Am|U8|LATIN1_TO_NATIVE|U8 ch
160
161Returns the native equivalent of the input Latin-1 code point (including ASCII
162and control characters) given by C<ch>. Thus, C<LATIN1_TO_NATIVE(66)> on
163EBCDIC platforms returns 194. These each represent the character C<"B"> on
164their respective platforms. On ASCII platforms no conversion is needed, so
165this macro expands to just its input, adding no time nor space requirements to
166the implementation.
167
168For conversion of code points potentially larger than will fit in a character,
169use L</UNI_TO_NATIVE>.
170
171=for apidoc Am|UV|NATIVE_TO_UNI|UV ch
172
173Returns the Unicode equivalent of the input native code point given by C<ch>.
174Thus, C<NATIVE_TO_UNI(195)> on EBCDIC platforms returns 67. These each
175represent the character C<"C"> on their respective platforms. On ASCII
176platforms no conversion is needed, so this macro expands to just its input,
177adding no time nor space requirements to the implementation.
178
179=for apidoc Am|UV|UNI_TO_NATIVE|UV ch
180
181Returns the native equivalent of the input Unicode code point given by C<ch>.
182Thus, C<UNI_TO_NATIVE(68)> on EBCDIC platforms returns 196. These each
183represent the character C<"D"> on their respective platforms. On ASCII
184platforms no conversion is needed, so this macro expands to just its input,
185adding no time nor space requirements to the implementation.
186
187=cut
188*/
189
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)))
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 */
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)))
201
202#define UNI_TO_NATIVE(ch) ((UV) ((ch) | 0))
203#define NATIVE_TO_UNI(ch) ((UV) ((ch) | 0))
204
205/*
206
207 The following table is from Unicode 3.2, plus the Perl extensions for above
208 U+10FFFF
209
210 Code Points 1st Byte 2nd Byte 3rd 4th 5th 6th 7th 8th-13th
211
212 U+0000..U+007F 00..7F
213 U+0080..U+07FF * C2..DF 80..BF
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
226U+1000000..U+3FFFFFF F9..FB 80..BF 80..BF 80..BF 80..BF
227U+4000000..U+3FFFFFFF FC * 84..BF 80..BF 80..BF 80..BF 80..BF
228U+40000000..U+7FFFFFFF FD 80..BF 80..BF 80..BF 80..BF 80..BF
229U+80000000..U+FFFFFFFFF FE * 82..BF 80..BF 80..BF 80..BF 80..BF 80..BF
230U+1000000000.. FF 80..BF 80..BF 80..BF 80..BF 80..BF * 81..BF 80..BF
231
232Note the gaps before several of the byte entries above marked by '*'. These are
233caused by legal UTF-8 avoiding non-shortest encodings: it is technically
234possible to UTF-8-encode a single code point in different ways, but that is
235explicitly forbidden, and the shortest possible encoding should always be used
236(and that is what Perl does). The non-shortest ones are called 'overlongs'.
237
238 */
239
240/*
241 Another way to look at it, as bits:
242
243 Code Points 1st Byte 2nd Byte 3rd Byte 4th Byte
244
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
249
250As you can see, the continuation bytes all begin with C<10>, and the
251leading bits of the start byte tell how many bytes there are in the
252encoded character.
253
254Perl's extended UTF-8 means we can have start bytes up through FF, though any
255beginning with FF yields a code point that is too large for 32-bit ASCII
256platforms. FF signals to use 13 bytes for the encoded character. This breaks
257the paradigm that the number of leading bits gives how many total bytes there
258are in the character. */
259
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 */
264# define UTF_CONTINUATION_BYTE_INFO_BITS 6
265
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
272 * those to quickly exclude the vast majority of code points it will
273 * encounter */
274#define isUTF8_POSSIBLY_PROBLEMATIC(c) (__ASSERT_(FITS_IN_8_BITS(c)) \
275 (U8) c >= 0xED)
276
277#endif /* EBCDIC vs ASCII */
278
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
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
291 * real information in a continuation byte. This turns out to be 0x3F in
292 * UTF-8, 0x1F in UTF-EBCDIC. */
293#define UTF_CONTINUATION_MASK \
294 ((U8) nBIT_MASK(UTF_CONTINUATION_BYTE_INFO_BITS))
295
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
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
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.) */
305#define UTF_CONTINUATION_MARK (UTF_IS_CONTINUATION_MASK & 0xB0)
306
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
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
318/*
319=for apidoc Am|bool|UVCHR_IS_INVARIANT|UV cp
320
321Evaluates to 1 if the representation of code point C<cp> is the same whether or
322not it is encoded in UTF-8; otherwise evaluates to 0. UTF-8 invariant
323characters can be copied as-is when converting to/from UTF-8, saving time.
324C<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
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
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
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 */
378#define __COMMON_UNI_SKIP(uv) \
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 :
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.
387 * For any-sized EBCDIC platforms, or 64-bit ASCII ones, we need one more test
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 */
390#if defined(UV_IS_QUAD) || defined(EBCDIC)
391# define __BASE_UNI_SKIP(uv) (__COMMON_UNI_SKIP(uv) \
392 LIKELY((UV) (uv) < ((UV) 1U << (6 * UTF_ACCUMULATION_SHIFT))) \
393 ? 7 \
394 : UTF8_MAXBYTES)
395#else
396# define __BASE_UNI_SKIP(uv) (__COMMON_UNI_SKIP(uv) 7)
397#endif
398
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))
405
406/*
407
408=for apidoc Am|STRLEN|UVCHR_SKIP|UV cp
409returns the number of bytes required to represent the code point C<cp> when
410encoded as UTF-8. C<cp> is a native (ASCII or EBCDIC) code point if less than
411255; a Unicode code point otherwise.
412
413=cut
414 */
415#define UVCHR_SKIP(uv) ( UVCHR_IS_INVARIANT(uv) ? 1 : __BASE_UNI_SKIP(uv))
416
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?
421 * This excludes invariants (they are single-byte). It also excludes the
422 * illegal overlong sequences that begin with C0 and C1 on ASCII platforms, and
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)) \
427 (NATIVE_UTF8_TO_I8(c) >= UTF_MIN_START_BYTE))
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
434
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
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)) \
447 inRANGE_helper_(U8, NATIVE_UTF8_TO_I8(c), \
448 UTF_MIN_START_BYTE, UTF_MIN_ABOVE_LATIN1_BYTE - 1))
449
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
452 * UTF_ACCUMULATION_SHIFT bits of information each */
453#define MAX_UTF8_TWO_BYTE (32 * (1U << UTF_ACCUMULATION_SHIFT) - 1)
454
455/* The largest code point representable by two UTF-8 bytes on any platform that
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))
460
461/*
462
463=for apidoc AmnU|STRLEN|UTF8_MAXBYTES_CASE
464
465The maximum number of UTF-8 bytes a single Unicode character can
466uppercase/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*/
479#define UTF8_MAXBYTES_CASE \
480 MAX(UTF8_MAXBYTES, UTF8_MAX_FOLD_CHAR_EXPAND * OFFUNISKIP(0x10FFFF))
481
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
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)
498
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
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 */
505#define UTF8_ACCUMULATE(old, new) (__ASSERT_(FITS_IN_8_BITS(new)) \
506 ((old) << UTF_ACCUMULATION_SHIFT) \
507 | ((NATIVE_UTF8_TO_I8(new)) \
508 & UTF_CONTINUATION_MASK))
509
510/* This works in the face of malformed UTF-8. */
511#define UTF8_IS_NEXT_CHAR_DOWNGRADEABLE(s, e) \
512 ( ( (e) - (s) > 1) \
513 && UTF8_IS_DOWNGRADEABLE_START(*(s)) \
514 && UTF8_IS_CONTINUATION(*((s)+1)))
515
516/* Number of bytes a code point occupies in UTF-8. */
517#define NATIVE_SKIP(uv) UVCHR_SKIP(uv)
518
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
521 * it means existing code continues to work. For precision, use UVCHR_SKIP,
522 * NATIVE_SKIP, or OFFUNISKIP */
523#define UNISKIP(uv) UVCHR_SKIP(uv)
524
525/* Longer, but more accurate name */
526#define UTF8_IS_ABOVE_LATIN1_START(c) UTF8_IS_ABOVE_LATIN1(c)
527
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
543/* Convert a two (not one) byte utf8 character to a native code point value.
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 */
548#define TWO_BYTE_UTF8_TO_NATIVE(HI, LO) \
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)) \
553 UNI_TO_NATIVE(UTF8_ACCUMULATE((NATIVE_UTF8_TO_I8(HI) & UTF_START_MASK(2)), \
554 (LO))))
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))
558
559/*
560
561=for apidoc Am|STRLEN|UTF8SKIP|char* s
562returns the number of bytes a non-malformed UTF-8 encoded character whose first
563(perhaps only) byte is pointed to by C<s>.
564
565If there is a possibility of malformed input, use instead:
566
567=over
568
569=item C<L</UTF8_SAFE_SKIP>> if you know the maximum ending pointer in the
570buffer pointed to by C<s>; or
571
572=item C<L</UTF8_CHK_SKIP>> if you don't know it.
573
574=back
575
576It is better to restructure your code so the end pointer is passed down so that
577you know what it actually is at the point of this call, but if that isn't
578possible, C<L</UTF8_CHK_SKIP>> can minimize the chance of accessing beyond the end
579of the input buffer.
580
581=cut
582 */
583#define UTF8SKIP(s) PL_utf8skip[*(const U8*)(s)]
584
585/*
586=for apidoc Am|STRLEN|UTF8_SKIP|char* s
587This is a synonym for C<L</UTF8SKIP>>
588
589=cut
590*/
591
592#define UTF8_SKIP(s) UTF8SKIP(s)
593
594/*
595=for apidoc Am|STRLEN|UTF8_CHK_SKIP|char* s
596
597This is a safer version of C<L</UTF8SKIP>>, but still not as safe as
598C<L</UTF8_SAFE_SKIP>>. This version doesn't blindly assume that the input
599string pointed to by C<s> is well-formed, but verifies that there isn't a NUL
600terminating character before the expected end of the next character in C<s>.
601The length C<UTF8_CHK_SKIP> returns stops just before any such NUL.
602
603Perl tends to add NULs, as an insurance policy, after the end of strings in
604SV's, so it is likely that using this macro will prevent inadvertent reading
605beyond the end of the input buffer, even if it is malformed UTF-8.
606
607This macro is intended to be used by XS modules where the inputs could be
608malformed, and it isn't feasible to restructure to use the safer
609C<L</UTF8_SAFE_SKIP>>, for example when interfacing with a C library.
610
611=cut
612*/
613
614#define UTF8_CHK_SKIP(s) \
615 (UNLIKELY(s[0] == '\0') ? 1 : MIN(UTF8SKIP(s), \
616 my_strnlen((char *) (s), UTF8SKIP(s))))
617/*
618
619=for apidoc Am|STRLEN|UTF8_SAFE_SKIP|char* s|char* e
620returns 0 if S<C<s E<gt>= e>>; otherwise returns the number of bytes in the
621UTF-8 encoded character whose first byte is pointed to by C<s>. But it never
622returns beyond C<e>. On DEBUGGING builds, it asserts that S<C<s E<lt>= e>>.
623
624=cut
625 */
626#define UTF8_SAFE_SKIP(s, e) (__ASSERT_((e) >= (s)) \
627 UNLIKELY(((e) - (s)) <= 0) \
628 ? 0 \
629 : MIN(((e) - (s)), UTF8_SKIP(s)))
630
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
635/*
636=for apidoc Am|bool|UTF8_IS_INVARIANT|char c
637
638Evaluates to 1 if the byte C<c> represents the same character when encoded in
639UTF-8 as when not; otherwise evaluates to 0. UTF-8 invariant characters can be
640copied as-is when converting to/from UTF-8, saving time.
641
642In spite of the name, this macro gives the correct result if the input string
643from which C<c> comes is not encoded in UTF-8.
644
645See C<L</UVCHR_IS_INVARIANT>> for checking if a UV is invariant.
646
647=cut
648
649The reason it works on both UTF-8 encoded strings and non-UTF-8 encoded, is
650that it returns TRUE in each for the exact same set of bit patterns. It is
651valid on a subset of what UVCHR_IS_INVARIANT is valid on, so can just use that;
652and the compiler should optimize out anything extraneous given the
653implementation of the latter. The |0 makes sure this isn't mistakenly called
654with a ptr argument.
655*/
656#define UTF8_IS_INVARIANT(c) UVCHR_IS_INVARIANT((c) | 0)
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 */
660#define NATIVE_BYTE_IS_INVARIANT(c) UVCHR_IS_INVARIANT(c)
661
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
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
673 * (which works for code points up through 0xFF) or NATIVE_TO_UNI which works
674 * for any code point */
675#define __BASE_TWO_BYTE_HI(c, translate_function) \
676 (__ASSERT_(! UVCHR_IS_INVARIANT(c)) \
677 I8_TO_NATIVE_UTF8((translate_function(c) >> UTF_ACCUMULATION_SHIFT) \
678 | UTF_START_MARK(2)))
679#define __BASE_TWO_BYTE_LO(c, translate_function) \
680 (__ASSERT_(! UVCHR_IS_INVARIANT(c)) \
681 I8_TO_NATIVE_UTF8((translate_function(c) & UTF_CONTINUATION_MASK) \
682 | UTF_CONTINUATION_MARK))
683
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
686 * regen/unicode_constants.pl instead */
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)) \
693 ( __BASE_TWO_BYTE_HI(c, NATIVE_TO_LATIN1)))
694#define UTF8_EIGHT_BIT_LO(c) (__ASSERT_(FITS_IN_8_BITS(c)) \
695 (__BASE_TWO_BYTE_LO(c, NATIVE_TO_LATIN1)))
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.
701 * MAX_UTF8_TWO_BYTE should be exactly all one bits in the lower few
702 * places, so the ~ works */
703#define UTF8_TWO_BYTE_HI(c) \
704 (__ASSERT_((sizeof(c) == 1) \
705 || !(((WIDEST_UTYPE)(c)) & ~MAX_UTF8_TWO_BYTE)) \
706 (__BASE_TWO_BYTE_HI(c, NATIVE_TO_UNI)))
707#define UTF8_TWO_BYTE_LO(c) \
708 (__ASSERT_((sizeof(c) == 1) \
709 || !(((WIDEST_UTYPE)(c)) & ~MAX_UTF8_TWO_BYTE)) \
710 (__BASE_TWO_BYTE_LO(c, NATIVE_TO_UNI)))
711
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
716/*
717 * 'UTF' is whether or not p is encoded in UTF8. The names 'foo_lazy_if' stem
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
722 */
723#define isIDFIRST_lazy_if_safe(p, e, UTF) \
724 ((IN_BYTES || !UTF) \
725 ? isIDFIRST(*(p)) \
726 : isIDFIRST_utf8_safe(p, e))
727#define isWORDCHAR_lazy_if_safe(p, e, UTF) \
728 ((IN_BYTES || !UTF) \
729 ? isWORDCHAR(*(p)) \
730 : isWORDCHAR_utf8_safe((U8 *) p, (U8 *) e))
731#define isALNUM_lazy_if_safe(p, e, UTF) isWORDCHAR_lazy_if_safe(p, e, UTF)
732
733#define UTF8_MAXLEN UTF8_MAXBYTES
734
735/* A Unicode character can fold to up to 3 characters */
736#define UTF8_MAX_FOLD_CHAR_EXPAND 3
737
738#define IN_BYTES UNLIKELY(CopHINTS_get(PL_curcop) & HINT_BYTES)
739
740/*
741
742=for apidoc Am|bool|DO_UTF8|SV* sv
743Returns a bool giving whether or not the PV in C<sv> is to be treated as being
744encoded in UTF-8.
745
746You should use this I<after> a call to C<SvPV()> or one of its variants, in
747case any call to string overloading updates the internal UTF-8 encoding flag.
748
749=cut
750*/
751#define DO_UTF8(sv) (SvUTF8(sv) && !IN_BYTES)
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 */
757#define IN_UNI_8_BIT \
758 (( ( (CopHINTS_get(PL_curcop) & HINT_UNI_8_BIT)) \
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))
763
764
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))
783#define UTF8_IS_PERL_EXTENDED(s) \
784 (UTF8SKIP(s) > 6 + ONE_IF_EBCDIC_ZERO_IF_NOT)
785
786#define UTF8_ALLOW_EMPTY 0x0001 /* Allow a zero length string */
787#define UTF8_GOT_EMPTY UTF8_ALLOW_EMPTY
788
789/* Allow first byte to be a continuation byte */
790#define UTF8_ALLOW_CONTINUATION 0x0002
791#define UTF8_GOT_CONTINUATION UTF8_ALLOW_CONTINUATION
792
793/* Unexpected non-continuation byte */
794#define UTF8_ALLOW_NON_CONTINUATION 0x0004
795#define UTF8_GOT_NON_CONTINUATION UTF8_ALLOW_NON_CONTINUATION
796
797/* expecting more bytes than were available in the string */
798#define UTF8_ALLOW_SHORT 0x0008
799#define UTF8_GOT_SHORT UTF8_ALLOW_SHORT
800
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 */
804#define UTF8_ALLOW_LONG 0x0010
805#define UTF8_ALLOW_LONG_AND_ITS_VALUE (UTF8_ALLOW_LONG|0x0020)
806#define UTF8_GOT_LONG UTF8_ALLOW_LONG
807
808#define UTF8_ALLOW_OVERFLOW 0x0080
809#define UTF8_GOT_OVERFLOW UTF8_ALLOW_OVERFLOW
810
811#define UTF8_DISALLOW_SURROGATE 0x0100 /* Unicode surrogates */
812#define UTF8_GOT_SURROGATE UTF8_DISALLOW_SURROGATE
813#define UTF8_WARN_SURROGATE 0x0200
814
815/* Unicode non-character code points */
816#define UTF8_DISALLOW_NONCHAR 0x0400
817#define UTF8_GOT_NONCHAR UTF8_DISALLOW_NONCHAR
818#define UTF8_WARN_NONCHAR 0x0800
819
820/* Super-set of Unicode: code points above the legal max */
821#define UTF8_DISALLOW_SUPER 0x1000
822#define UTF8_GOT_SUPER UTF8_DISALLOW_SUPER
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. */
834#define UTF8_DISALLOW_PERL_EXTENDED 0x4000
835#define UTF8_GOT_PERL_EXTENDED UTF8_DISALLOW_PERL_EXTENDED
836#define UTF8_WARN_PERL_EXTENDED 0x8000
837
838/* For back compat, these old names are misleading for overlongs and
839 * UTF_EBCDIC. */
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
845
846#define UTF8_CHECK_ONLY 0x10000
847#define _UTF8_NO_CONFIDENCE_IN_CURLEN 0x20000 /* Internal core use only */
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
853#define UTF8_ALLOW_FE_FF 0
854#define UTF8_ALLOW_SURROGATE 0
855
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
861#define UTF8_DISALLOW_ILLEGAL_INTERCHANGE \
862 (UTF8_DISALLOW_ILLEGAL_C9_INTERCHANGE|UTF8_DISALLOW_NONCHAR)
863#define UTF8_WARN_ILLEGAL_INTERCHANGE \
864 (UTF8_WARN_ILLEGAL_C9_INTERCHANGE|UTF8_WARN_NONCHAR)
865
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. */
870#define UTF8_ALLOW_ANY ( UTF8_ALLOW_CONTINUATION \
871 |UTF8_ALLOW_NON_CONTINUATION \
872 |UTF8_ALLOW_SHORT \
873 |UTF8_ALLOW_LONG \
874 |UTF8_ALLOW_OVERFLOW)
875
876/* Accept any Perl-extended UTF-8 that evaluates to any UV on the platform, but
877 * not any malformed. This is the default. */
878#define UTF8_ALLOW_ANYUV 0
879#define UTF8_ALLOW_DEFAULT UTF8_ALLOW_ANYUV
880
881/*
882=for apidoc Am|bool|UTF8_IS_SURROGATE|const U8 *s|const U8 *e
883
884Evaluates to non-zero if the first few bytes of the string starting at C<s> and
885looking no further than S<C<e - 1>> are well-formed UTF-8 that represents one
886of the Unicode surrogate code points; otherwise it evaluates to 0. If
887non-zero, the value gives how many bytes starting at C<s> comprise the code
888point'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
897#define MAX_LEGAL_CP ((UV)IV_MAX)
898
899/*
900=for apidoc Am|bool|UTF8_IS_SUPER|const U8 *s|const U8 *e
901
902Recall that Perl recognizes an extension to UTF-8 that can encode code
903points larger than the ones defined by Unicode, which are 0..0x10FFFF.
904
905This macro evaluates to non-zero if the first few bytes of the string starting
906at C<s> and looking no further than S<C<e - 1>> are from this UTF-8 extension;
907otherwise it evaluates to 0. If non-zero, the value gives how many bytes
908starting at C<s> comprise the code point's representation.
909
9100 is returned if the bytes are not well-formed extended UTF-8, or if they
911represent a code point that cannot fit in a UV on the current platform. Hence
912this macro can give different results when run on a 64-bit word machine than on
913one with a 32-bit word size.
914
915Note that it is illegal to have code points that are larger than what can
916fit in an IV on the current machine.
917
918=cut
919
920 * ASCII EBCDIC I8
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
924 */
925#ifdef EBCDIC
926# define UTF8_IS_SUPER(s, e) \
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))) \
932 ? is_utf8_char_helper(s, s + UTF8SKIP(s), 0) : 0)
933#else
934# define UTF8_IS_SUPER(s, e) \
935 (( ((e) > (s) + 3) \
936 && (*(U8*) (s)) >= 0xF4 \
937 && (UNLIKELY( ((*(U8*) (s)) > 0xF4) \
938 || (*((U8*) (s) + 1) >= 0x90))) \
939 && LIKELY((s) + UTF8SKIP(s) <= (e))) \
940 ? is_utf8_char_helper(s, s + UTF8SKIP(s), 0) : 0)
941#endif
942
943/* These are now machine generated, and the 'given' clause is no longer
944 * applicable */
945#define UTF8_IS_NONCHAR_GIVEN_THAT_NON_SUPER_AND_GE_PROBLEMATIC(s, e) \
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
951Evaluates to non-zero if the first few bytes of the string starting at C<s> and
952looking no further than S<C<e - 1>> are well-formed UTF-8 that represents one
953of the Unicode non-character code points; otherwise it evaluates to 0. If
954non-zero, the value gives how many bytes starting at C<s> comprise the code
955point's representation.
956
957=for apidoc AmnU|UV|UNICODE_REPLACEMENT
958
959Evaluates to 0xFFFD, the code point of the Unicode REPLACEMENT CHARACTER
960
961=cut
962 */
963#define UTF8_IS_NONCHAR(s, e) \
964 UTF8_IS_NONCHAR_GIVEN_THAT_NON_SUPER_AND_GE_PROBLEMATIC(s, e)
965
966#define UNICODE_SURROGATE_FIRST 0xD800
967#define UNICODE_SURROGATE_LAST 0xDFFF
968#define UNICODE_REPLACEMENT 0xFFFD
969#define UNICODE_BYTE_ORDER_MARK 0xFEFF
970
971/* Though our UTF-8 encoding can go beyond this,
972 * let's be conservative and do as Unicode says. */
973#define PERL_UNICODE_MAX 0x10FFFF
974
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
984
985#ifdef PERL_CORE
986# define UNICODE_ALLOW_ABOVE_IV_MAX 0x0100
987#endif
988#define UNICODE_DISALLOW_ABOVE_31_BIT UNICODE_DISALLOW_PERL_EXTENDED
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
995#define UNICODE_WARN_ILLEGAL_C9_INTERCHANGE \
996 (UNICODE_WARN_SURROGATE|UNICODE_WARN_SUPER)
997#define UNICODE_WARN_ILLEGAL_INTERCHANGE \
998 (UNICODE_WARN_ILLEGAL_C9_INTERCHANGE|UNICODE_WARN_NONCHAR)
999#define UNICODE_DISALLOW_ILLEGAL_C9_INTERCHANGE \
1000 (UNICODE_DISALLOW_SURROGATE|UNICODE_DISALLOW_SUPER)
1001#define UNICODE_DISALLOW_ILLEGAL_INTERCHANGE \
1002 (UNICODE_DISALLOW_ILLEGAL_C9_INTERCHANGE|UNICODE_DISALLOW_NONCHAR)
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
1008
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))
1012
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)
1016
1017/* Is 'uv' one of the 32 contiguous-range noncharacters? */
1018#define UNICODE_IS_32_CONTIGUOUS_NONCHARS(uv) \
1019 UNLIKELY(inRANGE(uv, 0xFDD0, 0xFDEF))
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) \
1025 UNLIKELY(((UV) (uv) & 0xFFFE) == 0xFFFE)
1026
1027#define UNICODE_IS_NONCHAR(uv) \
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))))
1031
1032#define UNICODE_IS_SUPER(uv) UNLIKELY((UV) (uv) > PERL_UNICODE_MAX)
1033
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
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
1045#define GREEK_SMALL_LETTER_MU 0x03BC
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 */
1049#ifdef LATIN_CAPITAL_LETTER_SHARP_S_UTF8
1050# define LATIN_CAPITAL_LETTER_SHARP_S 0x1E9E
1051#endif
1052#define LATIN_CAPITAL_LETTER_I_WITH_DOT_ABOVE 0x130
1053#define LATIN_SMALL_LETTER_DOTLESS_I 0x131
1054#define LATIN_SMALL_LETTER_LONG_S 0x017F
1055#define LATIN_SMALL_LIGATURE_LONG_S_T 0xFB05
1056#define LATIN_SMALL_LIGATURE_ST 0xFB06
1057#define KELVIN_SIGN 0x212A
1058#define ANGSTROM_SIGN 0x212B
1059
1060#define UNI_DISPLAY_ISPRINT 0x0001
1061#define UNI_DISPLAY_BACKSLASH 0x0002
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 */
1069#define UNI_DISPLAY_REGEX (UNI_DISPLAY_ISPRINT|UNI_DISPLAY_BACKSLASH)
1070
1071/* Should be removed; maybe deprecated, but not used in CPAN */
1072#define SHARP_S_SKIP 2
1073
1074#define is_utf8_char_buf(buf, buf_end) isUTF8_CHAR(buf, buf_end)
1075#define bytes_from_utf8(s, lenp, is_utf8p) \
1076 bytes_from_utf8_loc(s, lenp, is_utf8p, 0)
1077
1078/*
1079
1080=for apidoc Am|STRLEN|isUTF8_CHAR_flags|const U8 *s|const U8 *e| const U32 flags
1081
1082Evaluates to non-zero if the first few bytes of the string starting at C<s> and
1083looking no further than S<C<e - 1>> are well-formed UTF-8, as extended by Perl,
1084that represents some code point, subject to the restrictions given by C<flags>;
1085otherwise it evaluates to 0. If non-zero, the value gives how many bytes
1086starting at C<s> comprise the code point's representation. Any bytes remaining
1087before C<e>, but beyond the ones needed to form the first code point in C<s>,
1088are not examined.
1089
1090If C<flags> is 0, this gives the same results as C<L</isUTF8_CHAR>>;
1091if C<flags> is C<UTF8_DISALLOW_ILLEGAL_INTERCHANGE>, this gives the same results
1092as C<L</isSTRICT_UTF8_CHAR>>;
1093and if C<flags> is C<UTF8_DISALLOW_ILLEGAL_C9_INTERCHANGE>, this gives
1094the same results as C<L</isC9_STRICT_UTF8_CHAR>>.
1095Otherwise C<flags> may be any combination of the C<UTF8_DISALLOW_I<foo>> flags
1096understood by C<L</utf8n_to_uvchr>>, with the same meanings.
1097
1098The three alternative macros are for the most commonly needed validations; they
1099are likely to run somewhat faster than this more general one, as they can be
1100inlined into your code.
1101
1102Use L</is_utf8_string_flags>, L</is_utf8_string_loc_flags>, and
1103L</is_utf8_string_loclen_flags> to check entire strings.
1104
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 \
1115 : is_utf8_char_helper(s, e, flags))
1116
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)
1120
1121#endif /* PERL_UTF8_H_ */
1122
1123/*
1124 * ex: set ts=8 sts=4 sw=4 et:
1125 */
Unnamed repository; edit this file 'description' to name the repository.