This is a live mirror of the Perl 5 development currently hosted at https://github.com/perl/perl5
Change filter of problematic code points for EBCDIC
[perl5.git] / utf8.h
CommitLineData
a0ed51b3
LW
1/* utf8.h
2 *
f3cb6f94
KW
3 * This file contains definitions for use with the UTF-8 encoding. It
4 * actually also works with the variant UTF-8 encoding called UTF-EBCDIC, and
5 * hides almost all of the differences between these from the caller. In other
6 * words, someone should #include this file, and if the code is being compiled
7 * on an EBCDIC platform, things should mostly just work.
8 *
2eee27d7
SS
9 * Copyright (C) 2000, 2001, 2002, 2005, 2006, 2007, 2009,
10 * 2010, 2011 by Larry Wall and others
a0ed51b3
LW
11 *
12 * You may distribute under the terms of either the GNU General Public
13 * License or the Artistic License, as specified in the README file.
14 *
15 */
16
57f0e7e2
KW
17#ifndef H_UTF8 /* Guard against recursive inclusion */
18#define H_UTF8 1
19
39e02b42 20/* Use UTF-8 as the default script encoding?
1e54db1a 21 * Turning this on will break scripts having non-UTF-8 binary
39e02b42
JH
22 * data (such as Latin-1) in string literals. */
23#ifdef USE_UTF8_SCRIPTS
24# define USE_UTF8_IN_NAMES (!IN_BYTES)
25#else
26# define USE_UTF8_IN_NAMES (PL_hints & HINT_UTF8)
27#endif
28
3cd96634
KW
29#include "regcharclass.h"
30#include "unicode_constants.h"
31
051a06d4 32/* For to_utf8_fold_flags, q.v. */
e4f4ef45
KW
33#define FOLD_FLAGS_LOCALE 0x1
34#define FOLD_FLAGS_FULL 0x2
35#define FOLD_FLAGS_NOMIX_ASCII 0x4
051a06d4 36
83199d38
KW
37/* For _core_swash_init(), internal core use only */
38#define _CORE_SWASH_INIT_USER_DEFINED_PROPERTY 0x1
5d3d13d1 39#define _CORE_SWASH_INIT_RETURN_IF_UNDEF 0x2
87367d5f 40#define _CORE_SWASH_INIT_ACCEPT_INVLIST 0x4
83199d38 41
7bbfa158
KW
42/*
43=head1 Unicode Support
8cca77bc
KW
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.
7bbfa158
KW
50
51=for apidoc is_ascii_string
52
53This is a misleadingly-named synonym for L</is_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_invariant_string> is preferred.
57
58=cut
59*/
60#define is_ascii_string(s, len) is_invariant_string(s, len)
61
de69f3af
KW
62#define uvchr_to_utf8(a,b) uvchr_to_utf8_flags(a,b,0)
63#define uvchr_to_utf8_flags(d,uv,flags) \
64 uvoffuni_to_utf8_flags(d,NATIVE_TO_UNI(uv),flags)
65#define utf8_to_uvchr_buf(s, e, lenp) \
842991ae 66 utf8n_to_uvchr(s, (U8*)(e) - (U8*)(s), lenp, \
de69f3af
KW
67 ckWARN_d(WARN_UTF8) ? 0 : UTF8_ALLOW_ANY)
68
a0270393 69#define to_uni_fold(c, p, lenp) _to_uni_fold_flags(c, p, lenp, FOLD_FLAGS_FULL)
445bf929
KW
70#define to_utf8_fold(c, p, lenp) _to_utf8_fold_flags(c, p, lenp, FOLD_FLAGS_FULL)
71#define to_utf8_lower(a,b,c) _to_utf8_lower_flags(a,b,c,0)
72#define to_utf8_upper(a,b,c) _to_utf8_upper_flags(a,b,c,0)
73#define to_utf8_title(a,b,c) _to_utf8_title_flags(a,b,c,0)
36bb2ab6 74
fd7cb289 75/* Source backward compatibility. */
fd7cb289
RGS
76#define is_utf8_string_loc(s, len, ep) is_utf8_string_loclen(s, len, ep, 0)
77
eda9cac1
KW
78#define foldEQ_utf8(s1, pe1, l1, u1, s2, pe2, l2, u2) \
79 foldEQ_utf8_flags(s1, pe1, l1, u1, s2, pe2, l2, u2, 0)
baa60164 80#define FOLDEQ_UTF8_NOMIX_ASCII (1 << 0)
cea315b6 81#define FOLDEQ_LOCALE (1 << 1)
18f762c3
KW
82#define FOLDEQ_S1_ALREADY_FOLDED (1 << 2)
83#define FOLDEQ_S2_ALREADY_FOLDED (1 << 3)
d635b710
KW
84#define FOLDEQ_S1_FOLDS_SANE (1 << 4)
85#define FOLDEQ_S2_FOLDS_SANE (1 << 5)
a33c29bc 86
e6226b18
KW
87#define ibcmp_utf8(s1, pe1, l1, u1, s2, pe2, l2, u2) \
88 cBOOL(! foldEQ_utf8(s1, pe1, l1, u1, s2, pe2, l2, u2))
89
1d72bdf6
NIS
90#ifdef EBCDIC
91/* The equivalent of these macros but implementing UTF-EBCDIC
92 are in the following header file:
93 */
94
95#include "utfebcdic.h"
fd7cb289 96
d06134e5 97#else /* ! EBCDIC */
73c4f7a1
GS
98START_EXTERN_C
99
a0ed51b3 100#ifdef DOINIT
6f06b55f 101EXTCONST unsigned char PL_utf8skip[] = {
b2635aa8
KW
102/* 0x00 */ 1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1, /* ascii */
103/* 0x10 */ 1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1, /* ascii */
104/* 0x20 */ 1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1, /* ascii */
105/* 0x30 */ 1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1, /* ascii */
106/* 0x40 */ 1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1, /* ascii */
107/* 0x50 */ 1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1, /* ascii */
108/* 0x60 */ 1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1, /* ascii */
109/* 0x70 */ 1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1, /* ascii */
110/* 0x80 */ 1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1, /* bogus: continuation byte */
111/* 0x90 */ 1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1, /* bogus: continuation byte */
112/* 0xA0 */ 1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1, /* bogus: continuation byte */
113/* 0xB0 */ 1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1, /* bogus: continuation byte */
114/* 0xC0 */ 2,2, /* overlong */
1ff3baa2 115/* 0xC2 */ 2,2,2,2,2,2,2,2,2,2,2,2,2,2, /* U+0080 to U+03FF */
b2635aa8
KW
116/* 0xD0 */ 2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2, /* U+0400 to U+07FF */
117/* 0xE0 */ 3,3,3,3,3,3,3,3,3,3,3,3,3,3,3,3, /* U+0800 to U+FFFF */
118/* 0xF0 */ 4,4,4,4,4,4,4,4,5,5,5,5,6,6, /* above BMP to 2**31 - 1 */
119/* 0xFE */ 7,13, /* Perl extended (never was official UTF-8). Up to 72bit
120 allowed (64-bit + reserved). */
a0ed51b3
LW
121};
122#else
6f06b55f 123EXTCONST unsigned char PL_utf8skip[];
a0ed51b3
LW
124#endif
125
73c4f7a1 126END_EXTERN_C
7e2040f0 127
59a449d5
KW
128/* Native character to/from iso-8859-1. Are the identity functions on ASCII
129 * platforms */
130#define NATIVE_TO_LATIN1(ch) (ch)
131#define LATIN1_TO_NATIVE(ch) (ch)
132
133/* I8 is an intermediate version of UTF-8 used only in UTF-EBCDIC. We thus
134 * consider it to be identical to UTF-8 on ASCII platforms. Strictly speaking
135 * UTF-8 and UTF-EBCDIC are two different things, but we often conflate them
136 * because they are 8-bit encodings that serve the same purpose in Perl, and
137 * rarely do we need to distinguish them. The term "NATIVE_UTF8" applies to
138 * whichever one is applicable on the current platform */
139#define NATIVE_UTF8_TO_I8(ch) (ch)
140#define I8_TO_NATIVE_UTF8(ch) (ch)
141
1d72bdf6
NIS
142/* Transforms in wide UV chars */
143#define UNI_TO_NATIVE(ch) (ch)
144#define NATIVE_TO_UNI(ch) (ch)
d7578b48 145
877d9f0d 146/*
9041c2e3 147
8c007b5a 148 The following table is from Unicode 3.2.
877d9f0d
JH
149
150 Code Points 1st Byte 2nd Byte 3rd Byte 4th Byte
151
375122d7 152 U+0000..U+007F 00..7F
e1b711da 153 U+0080..U+07FF * C2..DF 80..BF
37e2e78e 154 U+0800..U+0FFF E0 * A0..BF 80..BF
375122d7 155 U+1000..U+CFFF E1..EC 80..BF 80..BF
e1b711da 156 U+D000..U+D7FF ED 80..9F 80..BF
537124e4 157 U+D800..U+DFFF ED A0..BF 80..BF (surrogates)
375122d7 158 U+E000..U+FFFF EE..EF 80..BF 80..BF
37e2e78e 159 U+10000..U+3FFFF F0 * 90..BF 80..BF 80..BF
877d9f0d
JH
160 U+40000..U+FFFFF F1..F3 80..BF 80..BF 80..BF
161 U+100000..U+10FFFF F4 80..8F 80..BF 80..BF
b2635aa8
KW
162 Below are non-Unicode code points
163 U+110000..U+13FFFF F4 90..BF 80..BF 80..BF
164 U+110000..U+1FFFFF F5..F7 80..BF 80..BF 80..BF
537124e4 165 U+200000..: F8.. * 88..BF 80..BF 80..BF 80..BF
877d9f0d 166
e1b711da 167Note the gaps before several of the byte entries above marked by '*'. These are
37e2e78e
KW
168caused by legal UTF-8 avoiding non-shortest encodings: it is technically
169possible to UTF-8-encode a single code point in different ways, but that is
170explicitly forbidden, and the shortest possible encoding should always be used
15824458 171(and that is what Perl does). The non-shortest ones are called 'overlongs'.
8c007b5a 172
877d9f0d
JH
173 */
174
8c007b5a
JH
175/*
176 Another way to look at it, as bits:
177
b2635aa8 178 Code Points 1st Byte 2nd Byte 3rd Byte 4th Byte
8c007b5a 179
b2635aa8
KW
180 0aaa aaaa 0aaa aaaa
181 0000 0bbb bbaa aaaa 110b bbbb 10aa aaaa
182 cccc bbbb bbaa aaaa 1110 cccc 10bb bbbb 10aa aaaa
183 00 000d ddcc cccc bbbb bbaa aaaa 1111 0ddd 10cc cccc 10bb bbbb 10aa aaaa
8c007b5a
JH
184
185As you can see, the continuation bytes all begin with C<10>, and the
e1b711da 186leading bits of the start byte tell how many bytes there are in the
8c007b5a
JH
187encoded character.
188
65ab9279
TC
189Perl's extended UTF-8 means we can have start bytes up to FF.
190
8c007b5a
JH
191*/
192
15824458
KW
193/* Is the representation of the Unicode code point 'c' the same regardless of
194 * being encoded in UTF-8 or not? */
c4d5f83a 195#define UNI_IS_INVARIANT(c) (((UV)c) < 0x80)
15824458
KW
196
197/* Is the UTF8-encoded byte 'c' part of a variant sequence in UTF-8? This is
198 * the inverse of UTF8_IS_INVARIANT */
e4f4ef45 199#define UTF8_IS_CONTINUED(c) (((U8)c) & 0x80)
15824458
KW
200
201/* Is the byte 'c' the first byte of a multi-byte UTF8-8 encoded sequence?
202 * This doesn't catch invariants (they are single-byte). It also excludes the
203 * illegal overlong sequences that begin with C0 and C1. */
e4f4ef45 204#define UTF8_IS_START(c) (((U8)c) >= 0xc2)
15824458
KW
205
206/* Is the byte 'c' part of a multi-byte UTF8-8 encoded sequence, and not the
207 * first byte thereof? */
e4f4ef45 208#define UTF8_IS_CONTINUATION(c) ((((U8)c) & 0xC0) == 0x80)
0ae1fa71 209
15824458
KW
210/* Is the UTF8-encoded byte 'c' the first byte of a two byte sequence? Use
211 * UTF8_IS_NEXT_CHAR_DOWNGRADEABLE() instead if the input isn't known to
212 * be well-formed. Masking with 0xfe allows the low bit to be 0 or 1; thus
213 * this matches 0xc[23]. */
559c7f10 214#define UTF8_IS_DOWNGRADEABLE_START(c) (((U8)(c) & 0xfe) == 0xc2)
4ab10950 215
15824458
KW
216/* Is the UTF8-encoded byte 'c' the first byte of a sequence of bytes that
217 * represent a code point > 255? */
e4f4ef45 218#define UTF8_IS_ABOVE_LATIN1(c) ((U8)(c) >= 0xc4)
8850bf83 219
15824458
KW
220/* This defines the 1-bits that are to be in the first byte of a multi-byte
221 * UTF-8 encoded character that give the number of bytes that comprise the
f710bf0c 222 * character. 'len' is the number of bytes in the multi-byte sequence. */
ee372ee9 223#define UTF_START_MARK(len) (((len) > 7) ? 0xFF : (0xFF & (0xFE << (7-(len)))))
b2635aa8
KW
224
225/* Masks out the initial one bits in a start byte, leaving the real data ones.
f710bf0c
KW
226 * Doesn't work on an invariant byte. 'len' is the number of bytes in the
227 * multi-byte sequence that comprises the character. */
22901f30 228#define UTF_START_MASK(len) (((len) >= 7) ? 0x00 : (0x1F >> ((len)-2)))
1d72bdf6 229
15824458
KW
230/* This defines the bits that are to be in the continuation bytes of a multi-byte
231 * UTF-8 encoded character that indicate it is a continuation byte. */
1d72bdf6 232#define UTF_CONTINUATION_MARK 0x80
15824458
KW
233
234/* This is the number of low-order bits a continuation byte in a UTF-8 encoded
235 * sequence contributes to the specification of the code point. In the bit
236 * maps above, you see that the first 2 bits are a constant '10', leaving 6 of
237 * real information */
1d72bdf6 238#define UTF_ACCUMULATION_SHIFT 6
b2635aa8
KW
239
240/* 2**UTF_ACCUMULATION_SHIFT - 1 */
1d72bdf6 241#define UTF_CONTINUATION_MASK ((U8)0x3f)
c512ce4f 242
15824458
KW
243/* If a value is anded with this, and the result is non-zero, then using the
244 * original value in UTF8_ACCUMULATE will overflow, shifting bits off the left
245 * */
eb83ed87 246#define UTF_ACCUMULATION_OVERFLOW_MASK \
bb88be5f
KW
247 (((UV) UTF_CONTINUATION_MASK) << ((sizeof(UV) * CHARBITS) \
248 - UTF_ACCUMULATION_SHIFT))
eb83ed87 249
6588300d
NC
250#if UVSIZE >= 8
251# define UTF8_QUAD_MAX UINT64_C(0x1000000000)
252
5aaebcb3
KW
253/* Input is a true Unicode (not-native) code point */
254#define OFFUNISKIP(uv) ( (uv) < 0x80 ? 1 : \
1d68d6cd
SC
255 (uv) < 0x800 ? 2 : \
256 (uv) < 0x10000 ? 3 : \
257 (uv) < 0x200000 ? 4 : \
258 (uv) < 0x4000000 ? 5 : \
259 (uv) < 0x80000000 ? 6 : \
9041c2e3 260 (uv) < UTF8_QUAD_MAX ? 7 : 13 )
1d68d6cd
SC
261#else
262/* No, I'm not even going to *TRY* putting #ifdef inside a #define */
5aaebcb3 263#define OFFUNISKIP(uv) ( (uv) < 0x80 ? 1 : \
1d68d6cd
SC
264 (uv) < 0x800 ? 2 : \
265 (uv) < 0x10000 ? 3 : \
266 (uv) < 0x200000 ? 4 : \
267 (uv) < 0x4000000 ? 5 : \
268 (uv) < 0x80000000 ? 6 : 7 )
269#endif
270
03c76984
KW
271/* How wide can a single UTF-8 encoded character become in bytes. */
272/* NOTE: Strictly speaking Perl's UTF-8 should not be called UTF-8 since UTF-8
273 * is an encoding of Unicode, and Unicode's upper limit, 0x10FFFF, can be
274 * expressed with 4 bytes. However, Perl thinks of UTF-8 as a way to encode
275 * non-negative integers in a binary format, even those above Unicode */
276#define UTF8_MAXBYTES 13
277
278/* The maximum number of UTF-8 bytes a single Unicode character can
279 * uppercase/lowercase/fold into. Unicode guarantees that the maximum
280 * expansion is 3 characters. On ASCIIish platforms, the highest Unicode
281 * character occupies 4 bytes, therefore this number would be 12, but this is
282 * smaller than the maximum width a single above-Unicode character can occupy,
283 * so use that instead */
284#if UTF8_MAXBYTES < 12
285#error UTF8_MAXBYTES must be at least 12
286#endif
287
0ed2b00b
KW
288/* ^? is defined to be DEL on ASCII systems. See the definition of toCTRL()
289 * for more */
290#define QUESTION_MARK_CTRL DEL_NATIVE
291
843a4590
KW
292#define MAX_UTF8_TWO_BYTE 0x7FF
293
03c76984
KW
294#define UTF8_MAXBYTES_CASE UTF8_MAXBYTES
295
d06134e5
KW
296#endif /* EBCDIC vs ASCII */
297
298/* Rest of these are attributes of Unicode and perl's internals rather than the
299 * encoding, or happen to be the same in both ASCII and EBCDIC (at least at
300 * this level; the macros that some of these call may have different
301 * definitions in the two encodings */
302
59a449d5
KW
303/* In domain restricted to ASCII, these may make more sense to the reader than
304 * the ones with Latin1 in the name */
305#define NATIVE_TO_ASCII(ch) NATIVE_TO_LATIN1(ch)
306#define ASCII_TO_NATIVE(ch) LATIN1_TO_NATIVE(ch)
307
308/* More or less misleadingly-named defines, retained for back compat */
309#define NATIVE_TO_UTF(ch) NATIVE_UTF8_TO_I8(ch)
310#define NATIVE_TO_I8(ch) NATIVE_UTF8_TO_I8(ch)
311#define UTF_TO_NATIVE(ch) I8_TO_NATIVE_UTF8(ch)
312#define I8_TO_NATIVE(ch) I8_TO_NATIVE_UTF8(ch)
313#define NATIVE8_TO_UNI(ch) NATIVE_TO_LATIN1(ch)
d06134e5 314
537124e4
KW
315/* Adds a UTF8 continuation byte 'new' of information to a running total code
316 * point 'old' of all the continuation bytes so far. This is designed to be
155d2738
KW
317 * used in a loop to convert from UTF-8 to the code point represented. Note
318 * that this is asymmetric on EBCDIC platforms, in that the 'new' parameter is
319 * the UTF-EBCDIC byte, whereas the 'old' parameter is a Unicode (not EBCDIC)
320 * code point in process of being generated */
321#define UTF8_ACCUMULATE(old, new) (((old) << UTF_ACCUMULATION_SHIFT) \
322 | ((NATIVE_UTF8_TO_I8((U8)new)) \
323 & UTF_CONTINUATION_MASK))
d06134e5 324
4ab10950
KW
325/* This works in the face of malformed UTF-8. */
326#define UTF8_IS_NEXT_CHAR_DOWNGRADEABLE(s, e) (UTF8_IS_DOWNGRADEABLE_START(*s) \
327 && ( (e) - (s) > 1) \
328 && UTF8_IS_CONTINUATION(*((s)+1)))
329
5aaebcb3
KW
330/* Number of bytes a code point occupies in UTF-8. */
331#define NATIVE_SKIP(uv) OFFUNISKIP(NATIVE_TO_UNI(uv))
332
bd18bd40
KW
333/*
334
335=for apidoc Am|STRLEN|UVCHR_SKIP|UV cp
336returns the number of bytes required to represent the code point C<cp> when
337encoded as UTF-8. C<cp> is a native (ASCII or EBCDIC) code point if less than
338255; a Unicode code point otherwise.
339
340=cut
341 */
342
5aaebcb3
KW
343/* Most code which says UNISKIP is really thinking in terms of native code
344 * points (0-255) plus all those beyond. This is an imprecise term, but having
2accb712
KW
345 * it means existing code continues to work. For precision, use UVCHR_SKIP,
346 * NATIVE_SKIP, and OFFUNISKIP */
5aaebcb3 347#define UNISKIP(uv) NATIVE_SKIP(uv)
2accb712 348#define UVCHR_SKIP(uv) NATIVE_SKIP(uv)
5aaebcb3 349
3c0792e4
KW
350/* Longer, but more accurate name */
351#define UTF8_IS_ABOVE_LATIN1_START(c) UTF8_IS_ABOVE_LATIN1(c)
352
a62b247b
KW
353/* Convert a UTF-8 variant Latin1 character to a native code point value.
354 * Needs just one iteration of accumulate. Should be used only if it is known
355 * that the code point is < 256, and is not UTF-8 invariant. Use the slower
356 * but more general TWO_BYTE_UTF8_TO_NATIVE() which handles any code point
357 * representable by two bytes (which turns out to be up through
358 * MAX_PORTABLE_UTF8_TWO_BYTE). The two parameters are:
359 * HI: a downgradable start byte;
360 * LO: continuation.
361 * */
362#define EIGHT_BIT_UTF8_TO_NATIVE(HI, LO) \
363 ( __ASSERT_(UTF8_IS_DOWNGRADEABLE_START(HI)) \
364 __ASSERT_(UTF8_IS_CONTINUATION(LO)) \
365 LATIN1_TO_NATIVE(UTF8_ACCUMULATE(( \
366 NATIVE_UTF8_TO_I8(HI) & UTF_START_MASK(2)), (LO))))
367
94bb8c36 368/* Convert a two (not one) byte utf8 character to a native code point value.
2950f2a7
KW
369 * Needs just one iteration of accumulate. Should not be used unless it is
370 * known that the two bytes are legal: 1) two-byte start, and 2) continuation.
371 * Note that the result can be larger than 255 if the input character is not
372 * downgradable */
94bb8c36 373#define TWO_BYTE_UTF8_TO_NATIVE(HI, LO) \
635e76f5
KW
374 ( __ASSERT_(UTF8SKIP(HI) == 2) \
375 __ASSERT_(UTF8_IS_CONTINUATION(LO)) \
94bb8c36 376 UNI_TO_NATIVE(UTF8_ACCUMULATE((NATIVE_UTF8_TO_I8(HI) & UTF_START_MASK(2)), \
635e76f5 377 (LO))))
94bb8c36
KW
378
379/* Should never be used, and be deprecated */
380#define TWO_BYTE_UTF8_TO_UNI(HI, LO) NATIVE_TO_UNI(TWO_BYTE_UTF8_TO_NATIVE(HI, LO))
2950f2a7 381
bd18bd40
KW
382/*
383
384=for apidoc Am|STRLEN|UTF8SKIP|char* s
385returns the number of bytes in the UTF-8 encoded character whose first (perhaps
386only) byte is pointed to by C<s>.
387
388=cut
389 */
2a70536e
KW
390#define UTF8SKIP(s) PL_utf8skip[*(const U8*)(s)]
391#define UTF8_SKIP(s) UTF8SKIP(s)
d06134e5 392
15824458
KW
393/* Is the byte 'c' the same character when encoded in UTF-8 as when not. This
394 * works on both UTF-8 encoded strings and non-encoded, as it returns TRUE in
395 * each for the exact same set of bit patterns. (And it works on any byte in a
396 * UTF-8 encoded string) */
bc3632a8 397#define UTF8_IS_INVARIANT(c) UNI_IS_INVARIANT(NATIVE_UTF8_TO_I8(c))
15824458 398
6f2d5cbc
KW
399/* Like the above, but its name implies a non-UTF8 input, and is implemented
400 * differently (for no particular reason) */
401#define NATIVE_BYTE_IS_INVARIANT(c) UNI_IS_INVARIANT(NATIVE_TO_LATIN1(c))
402
403/* Like the above, but accepts any UV as input */
404#define UVCHR_IS_INVARIANT(uv) UNI_IS_INVARIANT(NATIVE_TO_UNI(uv))
d06134e5
KW
405
406#define MAX_PORTABLE_UTF8_TWO_BYTE 0x3FF /* constrained by EBCDIC */
407
48ccf5e1
KW
408/* The macros in the next 4 sets are used to generate the two utf8 or utfebcdic
409 * bytes from an ordinal that is known to fit into exactly two (not one) bytes;
410 * it must be less than 0x3FF to work across both encodings. */
411
412/* These two are helper macros for the other three sets, and should not be used
413 * directly anywhere else. 'translate_function' is either NATIVE_TO_LATIN1
1ff3baa2
KW
414 * (which works for code points up through 0xFF) or NATIVE_TO_UNI which works
415 * for any code point */
48ccf5e1
KW
416#define __BASE_TWO_BYTE_HI(c, translate_function) \
417 I8_TO_NATIVE_UTF8((translate_function(c) >> UTF_ACCUMULATION_SHIFT) \
418 | UTF_START_MARK(2))
419#define __BASE_TWO_BYTE_LO(c, translate_function) \
420 I8_TO_NATIVE_UTF8((translate_function(c) & UTF_CONTINUATION_MASK) \
421 | UTF_CONTINUATION_MARK)
422
48ccf5e1
KW
423/* The next two macros should not be used. They were designed to be usable as
424 * the case label of a switch statement, but this doesn't work for EBCDIC. Use
9d0d3a03 425 * regen/unicode_constants.pl instead */
48ccf5e1
KW
426#define UTF8_TWO_BYTE_HI_nocast(c) __BASE_TWO_BYTE_HI(c, NATIVE_TO_UNI)
427#define UTF8_TWO_BYTE_LO_nocast(c) __BASE_TWO_BYTE_LO(c, NATIVE_TO_UNI)
428
429/* The next two macros are used when the source should be a single byte
430 * character; checked for under DEBUGGING */
431#define UTF8_EIGHT_BIT_HI(c) (__ASSERT_(FITS_IN_8_BITS(c)) \
432 ((U8) __BASE_TWO_BYTE_HI(c, NATIVE_TO_LATIN1)))
433#define UTF8_EIGHT_BIT_LO(c) (__ASSERT_(FITS_IN_8_BITS(c)) \
434 ((U8) __BASE_TWO_BYTE_LO(c, NATIVE_TO_LATIN1)))
435
436/* These final two macros in the series are used when the source can be any
437 * code point whose UTF-8 is known to occupy 2 bytes; they are less efficient
438 * than the EIGHT_BIT versions on EBCDIC platforms. We use the logical '~'
439 * operator instead of "<=" to avoid getting compiler warnings.
d52b8576 440 * MAX_UTF8_TWO_BYTE should be exactly all one bits in the lower few
48ccf5e1
KW
441 * places, so the ~ works */
442#define UTF8_TWO_BYTE_HI(c) \
443 (__ASSERT_((sizeof(c) == 1) \
d52b8576 444 || !(((WIDEST_UTYPE)(c)) & ~MAX_UTF8_TWO_BYTE)) \
d9759938 445 ((U8) __BASE_TWO_BYTE_HI(c, NATIVE_TO_UNI)))
48ccf5e1
KW
446#define UTF8_TWO_BYTE_LO(c) \
447 (__ASSERT_((sizeof(c) == 1) \
d52b8576 448 || !(((WIDEST_UTYPE)(c)) & ~MAX_UTF8_TWO_BYTE)) \
d9759938 449 ((U8) __BASE_TWO_BYTE_LO(c, NATIVE_TO_UNI)))
d06134e5 450
e7214ce8
KW
451/* This is illegal in any well-formed UTF-8 in both EBCDIC and ASCII
452 * as it is only in overlongs. */
453#define ILLEGAL_UTF8_BYTE I8_TO_NATIVE_UTF8(0xC1)
454
7e2040f0 455/*
e3036cf4 456 * 'UTF' is whether or not p is encoded in UTF8. The names 'foo_lazy_if' stem
20df05f4
KW
457 * from an earlier version of these macros in which they didn't call the
458 * foo_utf8() macros (i.e. were 'lazy') unless they decided that *p is the
459 * beginning of a utf8 character. Now that foo_utf8() determines that itself,
460 * no need to do it again here
7e2040f0 461 */
e3036cf4
KW
462#define isIDFIRST_lazy_if(p,UTF) ((IN_BYTES || !UTF ) \
463 ? isIDFIRST(*(p)) \
464 : isIDFIRST_utf8((const U8*)p))
32636478
KW
465#define isWORDCHAR_lazy_if(p,UTF) ((IN_BYTES || (!UTF )) \
466 ? isWORDCHAR(*(p)) \
467 : isWORDCHAR_utf8((const U8*)p))
468#define isALNUM_lazy_if(p,UTF) isWORDCHAR_lazy_if(p,UTF)
1d72bdf6 469
89ebb4a3
JH
470#define UTF8_MAXLEN UTF8_MAXBYTES
471
8cb75cc8
KW
472/* A Unicode character can fold to up to 3 characters */
473#define UTF8_MAX_FOLD_CHAR_EXPAND 3
474
a98fe34d 475#define IN_BYTES (CopHINTS_get(PL_curcop) & HINT_BYTES)
bd18bd40
KW
476
477/*
478
479=for apidoc Am|bool|DO_UTF8|SV* sv
480Returns a bool giving whether or not the PV in C<sv> is to be treated as being
481encoded in UTF-8.
482
483You should use this I<after> a call to C<SvPV()> or one of its variants, in
484case any call to string overloading updates the internal UTF-8 encoding flag.
485
486=cut
487*/
0064a8a9 488#define DO_UTF8(sv) (SvUTF8(sv) && !IN_BYTES)
1ff3baa2
KW
489
490/* Should all strings be treated as Unicode, and not just UTF-8 encoded ones?
491 * Is so within 'feature unicode_strings' or 'locale :not_characters', and not
492 * within 'use bytes'. UTF-8 locales are not tested for here, but perhaps
493 * could be */
494#define IN_UNI_8_BIT \
d6ded950
KW
495 (((CopHINTS_get(PL_curcop) & (HINT_UNI_8_BIT)) \
496 || (CopHINTS_get(PL_curcop) & HINT_LOCALE_PARTIAL \
497 /* -1 below is for :not_characters */ \
498 && _is_in_locale_category(FALSE, -1))) \
499 && ! IN_BYTES)
b36bf33f 500
1d72bdf6 501
c76687c5
KW
502#define UTF8_ALLOW_EMPTY 0x0001 /* Allow a zero length string */
503
504/* Allow first byte to be a continuation byte */
1d72bdf6 505#define UTF8_ALLOW_CONTINUATION 0x0002
c76687c5
KW
506
507/* Allow second... bytes to be non-continuation bytes */
1d72bdf6 508#define UTF8_ALLOW_NON_CONTINUATION 0x0004
949cf498
KW
509
510/* expecting more bytes than were available in the string */
511#define UTF8_ALLOW_SHORT 0x0008
512
513/* Overlong sequence; i.e., the code point can be specified in fewer bytes. */
514#define UTF8_ALLOW_LONG 0x0010
515
516#define UTF8_DISALLOW_SURROGATE 0x0020 /* Unicode surrogates */
517#define UTF8_WARN_SURROGATE 0x0040
518
519#define UTF8_DISALLOW_NONCHAR 0x0080 /* Unicode non-character */
520#define UTF8_WARN_NONCHAR 0x0100 /* code points */
521
522#define UTF8_DISALLOW_SUPER 0x0200 /* Super-set of Unicode: code */
523#define UTF8_WARN_SUPER 0x0400 /* points above the legal max */
524
525/* Code points which never were part of the original UTF-8 standard, the first
ea5ced44
KW
526 * byte of which is a FE or FF on ASCII platforms. If the first byte is FF, it
527 * will overflow a 32-bit word. If the first byte is FE, it will overflow a
528 * signed 32-bit word. */
949cf498
KW
529#define UTF8_DISALLOW_FE_FF 0x0800
530#define UTF8_WARN_FE_FF 0x1000
531
532#define UTF8_CHECK_ONLY 0x2000
533
534/* For backwards source compatibility. They do nothing, as the default now
535 * includes what they used to mean. The first one's meaning was to allow the
536 * just the single non-character 0xFFFF */
537#define UTF8_ALLOW_FFFF 0
538#define UTF8_ALLOW_SURROGATE 0
539
33d9abfb 540#define UTF8_DISALLOW_ILLEGAL_INTERCHANGE (UTF8_DISALLOW_SUPER|UTF8_DISALLOW_NONCHAR|UTF8_DISALLOW_SURROGATE|UTF8_DISALLOW_FE_FF)
949cf498 541#define UTF8_WARN_ILLEGAL_INTERCHANGE \
33d9abfb 542 (UTF8_WARN_SUPER|UTF8_WARN_NONCHAR|UTF8_WARN_SURROGATE|UTF8_WARN_FE_FF)
949cf498
KW
543#define UTF8_ALLOW_ANY \
544 (~(UTF8_DISALLOW_ILLEGAL_INTERCHANGE|UTF8_WARN_ILLEGAL_INTERCHANGE))
545#define UTF8_ALLOW_ANYUV \
546 (UTF8_ALLOW_EMPTY \
547 & ~(UTF8_DISALLOW_ILLEGAL_INTERCHANGE|UTF8_WARN_ILLEGAL_INTERCHANGE))
9f7f3913
ST
548#define UTF8_ALLOW_DEFAULT (ckWARN(WARN_UTF8) ? 0 : \
549 UTF8_ALLOW_ANYUV)
1d72bdf6 550
7131f24d
KW
551/* Surrogates, non-character code points and above-Unicode code points are
552 * problematic in some contexts. This allows code that needs to check for
553 * those to to quickly exclude the vast majority of code points it will
554 * encounter */
ac6f1fbe
KW
555#ifdef EBCDIC
556# define isUTF8_POSSIBLY_PROBLEMATIC(c) (NATIVE_UTF8_TO_I8(c) >= 0xF1)
557#else
558# define isUTF8_POSSIBLY_PROBLEMATIC(c) ((U8) c >= 0xED)
559#endif
7131f24d 560
0c58a72b
KW
561/* Several of the macros below have a second parameter that is currently
562 * unused; but could be used in the future to make sure that the input is
563 * well-formed. */
564
565#define UTF8_IS_SURROGATE(s, e) cBOOL(is_SURROGATE_utf8(s))
b96a92fb 566#define UTF8_IS_REPLACEMENT(s, send) cBOOL(is_REPLACEMENT_utf8_safe(s,send))
7131f24d
KW
567
568/* ASCII EBCDIC I8
569 * U+10FFFF: \xF4\x8F\xBF\xBF \xF9\xA1\xBF\xBF\xBF max legal Unicode
570 * U+110000: \xF4\x90\x80\x80 \xF9\xA2\xA0\xA0\xA0
571 * U+110001: \xF4\x90\x80\x81 \xF9\xA2\xA0\xA0\xA1
a1776718
KW
572 *
573 * BE AWARE that this test doesn't rule out malformed code points, in
574 * particular overlongs */
7131f24d 575#ifdef EBCDIC /* Both versions assume well-formed UTF8 */
0c58a72b 576# define UTF8_IS_SUPER(s, e) (NATIVE_UTF8_TO_I8(* (U8*) (s)) >= 0xF9 \
bc3632a8 577 && (NATIVE_UTF8_TO_I8(* (U8*) (s)) > 0xF9 \
a3481822 578 || (NATIVE_UTF8_TO_I8(* ((U8*) (s) + 1)) >= 0xA2)))
7131f24d 579#else
0c58a72b
KW
580# define UTF8_IS_SUPER(s, e) (*(U8*) (s) >= 0xF4 \
581 && (*(U8*) (s) > 0xF4 || (*((U8*) (s) + 1) >= 0x90)))
7131f24d
KW
582#endif
583
b96a92fb
KW
584/* These are now machine generated, and the 'given' clause is no longer
585 * applicable */
0c58a72b 586#define UTF8_IS_NONCHAR_GIVEN_THAT_NON_SUPER_AND_GE_PROBLEMATIC(s, e) \
b96a92fb 587 cBOOL(is_NONCHAR_utf8(s))
0c58a72b
KW
588#define UTF8_IS_NONCHAR(s, e) \
589 UTF8_IS_NONCHAR_GIVEN_THAT_NON_SUPER_AND_GE_PROBLEMATIC(s, e)
7131f24d 590
c867b360
JH
591#define UNICODE_SURROGATE_FIRST 0xD800
592#define UNICODE_SURROGATE_LAST 0xDFFF
593#define UNICODE_REPLACEMENT 0xFFFD
594#define UNICODE_BYTE_ORDER_MARK 0xFEFF
1d72bdf6 595
b851fbc1 596/* Though our UTF-8 encoding can go beyond this,
c76687c5 597 * let's be conservative and do as Unicode says. */
b851fbc1
JH
598#define PERL_UNICODE_MAX 0x10FFFF
599
949cf498
KW
600#define UNICODE_WARN_SURROGATE 0x0001 /* UTF-16 surrogates */
601#define UNICODE_WARN_NONCHAR 0x0002 /* Non-char code points */
602#define UNICODE_WARN_SUPER 0x0004 /* Above 0x10FFFF */
603#define UNICODE_WARN_FE_FF 0x0008 /* Above 0x10FFFF */
604#define UNICODE_DISALLOW_SURROGATE 0x0010
605#define UNICODE_DISALLOW_NONCHAR 0x0020
606#define UNICODE_DISALLOW_SUPER 0x0040
607#define UNICODE_DISALLOW_FE_FF 0x0080
bb88be5f
KW
608#define UNICODE_WARN_ILLEGAL_INTERCHANGE \
609 (UNICODE_WARN_SURROGATE|UNICODE_WARN_NONCHAR|UNICODE_WARN_SUPER)
610#define UNICODE_DISALLOW_ILLEGAL_INTERCHANGE \
611 (UNICODE_DISALLOW_SURROGATE|UNICODE_DISALLOW_NONCHAR|UNICODE_DISALLOW_SUPER)
949cf498
KW
612
613/* For backward source compatibility, as are now the default */
614#define UNICODE_ALLOW_SURROGATE 0
615#define UNICODE_ALLOW_SUPER 0
616#define UNICODE_ALLOW_ANY 0
b851fbc1 617
1d72bdf6
NIS
618#define UNICODE_IS_SURROGATE(c) ((c) >= UNICODE_SURROGATE_FIRST && \
619 (c) <= UNICODE_SURROGATE_LAST)
a10ec373 620#define UNICODE_IS_REPLACEMENT(c) ((c) == UNICODE_REPLACEMENT)
872c91ae 621#define UNICODE_IS_BYTE_ORDER_MARK(c) ((c) == UNICODE_BYTE_ORDER_MARK)
7131f24d
KW
622#define UNICODE_IS_NONCHAR(c) ((c >= 0xFDD0 && c <= 0xFDEF) \
623 /* The other noncharacters end in FFFE or FFFF, which \
624 * the mask below catches both of, but beyond the last \
625 * official unicode code point, they aren't \
626 * noncharacters, since those aren't Unicode \
627 * characters at all */ \
628 || ((((c & 0xFFFE) == 0xFFFE)) && ! UNICODE_IS_SUPER(c)))
629#define UNICODE_IS_SUPER(c) ((c) > PERL_UNICODE_MAX)
630#define UNICODE_IS_FE_FF(c) ((c) > 0x7FFFFFFF)
1d72bdf6 631
ec34087a
KW
632#define LATIN_SMALL_LETTER_SHARP_S LATIN_SMALL_LETTER_SHARP_S_NATIVE
633#define LATIN_SMALL_LETTER_Y_WITH_DIAERESIS \
634 LATIN_SMALL_LETTER_Y_WITH_DIAERESIS_NATIVE
635#define MICRO_SIGN MICRO_SIGN_NATIVE
636#define LATIN_CAPITAL_LETTER_A_WITH_RING_ABOVE \
637 LATIN_CAPITAL_LETTER_A_WITH_RING_ABOVE_NATIVE
638#define LATIN_SMALL_LETTER_A_WITH_RING_ABOVE \
639 LATIN_SMALL_LETTER_A_WITH_RING_ABOVE_NATIVE
09091399
JH
640#define UNICODE_GREEK_CAPITAL_LETTER_SIGMA 0x03A3
641#define UNICODE_GREEK_SMALL_LETTER_FINAL_SIGMA 0x03C2
642#define UNICODE_GREEK_SMALL_LETTER_SIGMA 0x03C3
9dcbe121 643#define GREEK_SMALL_LETTER_MU 0x03BC
9e682c18
KW
644#define GREEK_CAPITAL_LETTER_MU 0x039C /* Upper and title case
645 of MICRON */
646#define LATIN_CAPITAL_LETTER_Y_WITH_DIAERESIS 0x0178 /* Also is title case */
0766489e
KW
647#ifdef LATIN_CAPITAL_LETTER_SHARP_S_UTF8
648# define LATIN_CAPITAL_LETTER_SHARP_S 0x1E9E
649#endif
74894415
KW
650#define LATIN_CAPITAL_LETTER_I_WITH_DOT_ABOVE 0x130
651#define LATIN_SMALL_LETTER_DOTLESS_I 0x131
9e682c18 652#define LATIN_SMALL_LETTER_LONG_S 0x017F
a9f50d33
KW
653#define LATIN_SMALL_LIGATURE_LONG_S_T 0xFB05
654#define LATIN_SMALL_LIGATURE_ST 0xFB06
9e682c18
KW
655#define KELVIN_SIGN 0x212A
656#define ANGSTROM_SIGN 0x212B
09091399 657
9e55ce06 658#define UNI_DISPLAY_ISPRINT 0x0001
c728cb41
JH
659#define UNI_DISPLAY_BACKSLASH 0x0002
660#define UNI_DISPLAY_QQ (UNI_DISPLAY_ISPRINT|UNI_DISPLAY_BACKSLASH)
661#define UNI_DISPLAY_REGEX (UNI_DISPLAY_ISPRINT|UNI_DISPLAY_BACKSLASH)
9e55ce06 662
5cd46e1f
KW
663#define ANYOF_FOLD_SHARP_S(node, input, end) \
664 (ANYOF_BITMAP_TEST(node, LATIN_SMALL_LETTER_SHARP_S) && \
137165a6 665 (ANYOF_NONBITMAP(node)) && \
39065660 666 (ANYOF_FLAGS(node) & ANYOF_LOC_NONBITMAP_FOLD) && \
07b6858f 667 ((end) > (input) + 1) && \
305b8651 668 isALPHA_FOLD_EQ((input)[0], 's'))
6302f837 669
ebc501f0 670#define SHARP_S_SKIP 2
3b0fc154 671
a4f7a67c
KW
672/* If you want to exclude surrogates, and beyond legal Unicode, see the blame
673 * log for earlier versions which gave details for these */
4d646140 674
6302f837
KW
675/* A helper macro for isUTF8_CHAR, so use that one, and not this one. This is
676 * retained solely for backwards compatibility and may be deprecated and
677 * removed in a future Perl version.
678 *
679 * regen/regcharclass.pl generates is_UTF8_CHAR_utf8() macros for up to these
d9f92374
KW
680 * number of bytes. So this has to be coordinated with that file */
681#ifdef EBCDIC
682# define IS_UTF8_CHAR_FAST(n) ((n) <= 3)
683#else
684# define IS_UTF8_CHAR_FAST(n) ((n) <= 4)
685#endif
686
4d646140 687#ifndef EBCDIC
6302f837
KW
688/* A helper macro for isUTF8_CHAR, so use that one instead of this. This was
689 * generated by regen/regcharclass.pl, and then moved here. The lines that
690 * generated it were then commented out. This was done solely because it takes
691 * on the order of 10 minutes to generate, and is never going to change, unless
692 * the generated code is improved.
693 *
1ff3baa2
KW
694 * The EBCDIC versions have been cut to not cover all of legal Unicode,
695 * otherwise they take too long to generate; besides there is a separate one
696 * for each code page, so they are in regcharclass.h instead of here */
39a0f513 697/*
5dca9278 698 UTF8_CHAR: Matches legal UTF-8 encoded characters from 2 through 4 bytes
39a0f513 699
5dca9278 700 0x80 - 0x1FFFFF
39a0f513 701*/
4d646140 702/*** GENERATED CODE ***/
5dca9278
KW
703#define is_UTF8_CHAR_utf8_no_length_checks(s) \
704( ( 0xC2 <= ((U8*)s)[0] && ((U8*)s)[0] <= 0xDF ) ? \
705 ( ( ( ((U8*)s)[1] & 0xC0 ) == 0x80 ) ? 2 : 0 ) \
706: ( 0xE0 == ((U8*)s)[0] ) ? \
707 ( ( ( ( ((U8*)s)[1] & 0xE0 ) == 0xA0 ) && ( ( ((U8*)s)[2] & 0xC0 ) == 0x80 ) ) ? 3 : 0 )\
708: ( 0xE1 <= ((U8*)s)[0] && ((U8*)s)[0] <= 0xEF ) ? \
709 ( ( ( ( ((U8*)s)[1] & 0xC0 ) == 0x80 ) && ( ( ((U8*)s)[2] & 0xC0 ) == 0x80 ) ) ? 3 : 0 )\
710: ( 0xF0 == ((U8*)s)[0] ) ? \
711 ( ( ( ( 0x90 <= ((U8*)s)[1] && ((U8*)s)[1] <= 0xBF ) && ( ( ((U8*)s)[2] & 0xC0 ) == 0x80 ) ) && ( ( ((U8*)s)[3] & 0xC0 ) == 0x80 ) ) ? 4 : 0 )\
712: ( ( ( ( 0xF1 <= ((U8*)s)[0] && ((U8*)s)[0] <= 0xF7 ) && ( ( ((U8*)s)[1] & 0xC0 ) == 0x80 ) ) && ( ( ((U8*)s)[2] & 0xC0 ) == 0x80 ) ) && ( ( ((U8*)s)[3] & 0xC0 ) == 0x80 ) ) ? 4 : 0 )
4d646140 713#endif
3b0fc154 714
6302f837 715/*
5dca9278
KW
716
717=for apidoc Am|STRLEN|isUTF8_CHAR|const U8 *s|const U8 *e
718
719Returns the number of bytes beginning at C<s> which form a legal UTF-8 (or
61b16eb9
KW
720UTF-EBCDIC) encoded character, looking no further than S<C<e - s>> bytes into
721C<s>. Returns 0 if the sequence starting at C<s> through S<C<e - 1>> is not
722well-formed UTF-8.
6302f837
KW
723
724Note that an INVARIANT character (i.e. ASCII on non-EBCDIC
5dca9278
KW
725machines) is a valid UTF-8 character.
726
727=cut
728*/
6302f837 729
dd9bc2b0 730#define isUTF8_CHAR(s, e) (UNLIKELY((e) <= (s)) \
6302f837
KW
731 ? 0 \
732 : (UTF8_IS_INVARIANT(*s)) \
733 ? 1 \
dd9bc2b0 734 : UNLIKELY(((e) - (s)) < UTF8SKIP(s)) \
6302f837 735 ? 0 \
dd9bc2b0 736 : LIKELY(IS_UTF8_CHAR_FAST(UTF8SKIP(s))) \
5dca9278 737 ? is_UTF8_CHAR_utf8_no_length_checks(s) \
6302f837
KW
738 : _is_utf8_char_slow(s, e))
739
3cedd9d9
KW
740#define is_utf8_char_buf(buf, buf_end) isUTF8_CHAR(buf, buf_end)
741
6302f837
KW
742/* Do not use; should be deprecated. Use isUTF8_CHAR() instead; this is
743 * retained solely for backwards compatibility */
744#define IS_UTF8_CHAR(p, n) (isUTF8_CHAR(p, (p) + (n)) == n)
e9a8c099 745
57f0e7e2
KW
746#endif /* H_UTF8 */
747
e9a8c099 748/*
14d04a33 749 * ex: set ts=8 sts=4 sw=4 et:
e9a8c099 750 */