Commit | Line | Data |
---|---|---|
a0ed51b3 LW |
1 | /* utf8.c |
2 | * | |
1129b882 | 3 | * Copyright (C) 2000, 2001, 2002, 2003, 2004, 2005, 2006, 2007, 2008 |
b94e2f88 | 4 | * by Larry Wall and others |
a0ed51b3 LW |
5 | * |
6 | * You may distribute under the terms of either the GNU General Public | |
7 | * License or the Artistic License, as specified in the README file. | |
8 | * | |
9 | */ | |
10 | ||
11 | /* | |
4ac71550 TC |
12 | * 'What a fix!' said Sam. 'That's the one place in all the lands we've ever |
13 | * heard of that we don't want to see any closer; and that's the one place | |
14 | * we're trying to get to! And that's just where we can't get, nohow.' | |
15 | * | |
cdad3b53 | 16 | * [p.603 of _The Lord of the Rings_, IV/I: "The Taming of Sméagol"] |
a0ed51b3 LW |
17 | * |
18 | * 'Well do I understand your speech,' he answered in the same language; | |
19 | * 'yet few strangers do so. Why then do you not speak in the Common Tongue, | |
4ac71550 | 20 | * as is the custom in the West, if you wish to be answered?' |
cdad3b53 | 21 | * --Gandalf, addressing Théoden's door wardens |
4ac71550 TC |
22 | * |
23 | * [p.508 of _The Lord of the Rings_, III/vi: "The King of the Golden Hall"] | |
a0ed51b3 LW |
24 | * |
25 | * ...the travellers perceived that the floor was paved with stones of many | |
26 | * hues; branching runes and strange devices intertwined beneath their feet. | |
4ac71550 TC |
27 | * |
28 | * [p.512 of _The Lord of the Rings_, III/vi: "The King of the Golden Hall"] | |
a0ed51b3 LW |
29 | */ |
30 | ||
31 | #include "EXTERN.h" | |
864dbfa3 | 32 | #define PERL_IN_UTF8_C |
a0ed51b3 | 33 | #include "perl.h" |
81e983c1 | 34 | #include "inline_invlist.c" |
a0ed51b3 | 35 | |
27da23d5 JH |
36 | static const char unees[] = |
37 | "Malformed UTF-8 character (unexpected end of string)"; | |
901b21bf | 38 | |
48ef279e | 39 | /* |
ccfc67b7 | 40 | =head1 Unicode Support |
a0ed51b3 | 41 | |
166f8a29 DM |
42 | This file contains various utility functions for manipulating UTF8-encoded |
43 | strings. For the uninitiated, this is a method of representing arbitrary | |
61296642 | 44 | Unicode characters as a variable number of bytes, in such a way that |
56da48f7 DM |
45 | characters in the ASCII range are unmodified, and a zero byte never appears |
46 | within non-zero characters. | |
166f8a29 | 47 | |
eaf7a4d2 CS |
48 | =cut |
49 | */ | |
50 | ||
51 | /* | |
52 | =for apidoc is_ascii_string | |
53 | ||
a1433954 | 54 | Returns true if the first C<len> bytes of the string C<s> are the same whether |
970ea3cb KW |
55 | or not the string is encoded in UTF-8 (or UTF-EBCDIC on EBCDIC machines). That |
56 | is, if they are invariant. On ASCII-ish machines, only ASCII characters | |
57 | fit this definition, hence the function's name. | |
eaf7a4d2 | 58 | |
9f7e3d64 MH |
59 | If C<len> is 0, it will be calculated using C<strlen(s)>. |
60 | ||
a1433954 | 61 | See also L</is_utf8_string>(), L</is_utf8_string_loclen>(), and L</is_utf8_string_loc>(). |
eaf7a4d2 CS |
62 | |
63 | =cut | |
64 | */ | |
65 | ||
66 | bool | |
668b6d8d | 67 | Perl_is_ascii_string(const U8 *s, STRLEN len) |
eaf7a4d2 CS |
68 | { |
69 | const U8* const send = s + (len ? len : strlen((const char *)s)); | |
70 | const U8* x = s; | |
71 | ||
72 | PERL_ARGS_ASSERT_IS_ASCII_STRING; | |
eaf7a4d2 CS |
73 | |
74 | for (; x < send; ++x) { | |
75 | if (!UTF8_IS_INVARIANT(*x)) | |
76 | break; | |
77 | } | |
78 | ||
79 | return x == send; | |
80 | } | |
81 | ||
82 | /* | |
378516de | 83 | =for apidoc uvoffuni_to_utf8_flags |
eebe1485 | 84 | |
a27992cc | 85 | THIS FUNCTION SHOULD BE USED IN ONLY VERY SPECIALIZED CIRCUMSTANCES. |
de69f3af KW |
86 | Instead, B<Almost all code should use L</uvchr_to_utf8> or |
87 | L</uvchr_to_utf8_flags>>. | |
a27992cc | 88 | |
de69f3af KW |
89 | This function is like them, but the input is a strict Unicode |
90 | (as opposed to native) code point. Only in very rare circumstances should code | |
91 | not be using the native code point. | |
949cf498 | 92 | |
de69f3af | 93 | For details, see the description for L</uvchr_to_utf8_flags>>. |
949cf498 | 94 | |
eebe1485 SC |
95 | =cut |
96 | */ | |
97 | ||
dfe13c55 | 98 | U8 * |
378516de | 99 | Perl_uvoffuni_to_utf8_flags(pTHX_ U8 *d, UV uv, UV flags) |
a0ed51b3 | 100 | { |
378516de | 101 | PERL_ARGS_ASSERT_UVOFFUNI_TO_UTF8_FLAGS; |
7918f24d | 102 | |
d9432125 KW |
103 | if (UNI_IS_INVARIANT(uv)) { |
104 | *d++ = (U8) LATIN1_TO_NATIVE(uv); | |
105 | return d; | |
106 | } | |
107 | ||
979f77b6 KW |
108 | /* The first problematic code point is the first surrogate */ |
109 | if (uv >= UNICODE_SURROGATE_FIRST | |
110 | && ckWARN4_d(WARN_UTF8, WARN_SURROGATE, WARN_NON_UNICODE, WARN_NONCHAR)) | |
111 | { | |
949cf498 KW |
112 | if (UNICODE_IS_SURROGATE(uv)) { |
113 | if (flags & UNICODE_WARN_SURROGATE) { | |
8457b38f | 114 | Perl_ck_warner_d(aTHX_ packWARN(WARN_SURROGATE), |
949cf498 KW |
115 | "UTF-16 surrogate U+%04"UVXf, uv); |
116 | } | |
117 | if (flags & UNICODE_DISALLOW_SURROGATE) { | |
118 | return NULL; | |
119 | } | |
120 | } | |
121 | else if (UNICODE_IS_SUPER(uv)) { | |
122 | if (flags & UNICODE_WARN_SUPER | |
123 | || (UNICODE_IS_FE_FF(uv) && (flags & UNICODE_WARN_FE_FF))) | |
124 | { | |
8457b38f | 125 | Perl_ck_warner_d(aTHX_ packWARN(WARN_NON_UNICODE), |
949cf498 KW |
126 | "Code point 0x%04"UVXf" is not Unicode, may not be portable", uv); |
127 | } | |
128 | if (flags & UNICODE_DISALLOW_SUPER | |
129 | || (UNICODE_IS_FE_FF(uv) && (flags & UNICODE_DISALLOW_FE_FF))) | |
130 | { | |
131 | return NULL; | |
132 | } | |
133 | } | |
134 | else if (UNICODE_IS_NONCHAR(uv)) { | |
135 | if (flags & UNICODE_WARN_NONCHAR) { | |
8457b38f | 136 | Perl_ck_warner_d(aTHX_ packWARN(WARN_NONCHAR), |
949cf498 KW |
137 | "Unicode non-character U+%04"UVXf" is illegal for open interchange", |
138 | uv); | |
139 | } | |
140 | if (flags & UNICODE_DISALLOW_NONCHAR) { | |
141 | return NULL; | |
142 | } | |
143 | } | |
507b9800 | 144 | } |
d9432125 | 145 | |
2d331972 | 146 | #if defined(EBCDIC) |
d9432125 | 147 | { |
5aaebcb3 | 148 | STRLEN len = OFFUNISKIP(uv); |
1d72bdf6 NIS |
149 | U8 *p = d+len-1; |
150 | while (p > d) { | |
bc3632a8 | 151 | *p-- = (U8) I8_TO_NATIVE_UTF8((uv & UTF_CONTINUATION_MASK) | UTF_CONTINUATION_MARK); |
1d72bdf6 NIS |
152 | uv >>= UTF_ACCUMULATION_SHIFT; |
153 | } | |
bc3632a8 | 154 | *p = (U8) I8_TO_NATIVE_UTF8((uv & UTF_START_MASK(len)) | UTF_START_MARK(len)); |
1d72bdf6 NIS |
155 | return d+len; |
156 | } | |
157 | #else /* Non loop style */ | |
a0ed51b3 | 158 | if (uv < 0x800) { |
eb160463 GS |
159 | *d++ = (U8)(( uv >> 6) | 0xc0); |
160 | *d++ = (U8)(( uv & 0x3f) | 0x80); | |
a0ed51b3 LW |
161 | return d; |
162 | } | |
163 | if (uv < 0x10000) { | |
eb160463 GS |
164 | *d++ = (U8)(( uv >> 12) | 0xe0); |
165 | *d++ = (U8)(((uv >> 6) & 0x3f) | 0x80); | |
166 | *d++ = (U8)(( uv & 0x3f) | 0x80); | |
a0ed51b3 LW |
167 | return d; |
168 | } | |
169 | if (uv < 0x200000) { | |
eb160463 GS |
170 | *d++ = (U8)(( uv >> 18) | 0xf0); |
171 | *d++ = (U8)(((uv >> 12) & 0x3f) | 0x80); | |
172 | *d++ = (U8)(((uv >> 6) & 0x3f) | 0x80); | |
173 | *d++ = (U8)(( uv & 0x3f) | 0x80); | |
a0ed51b3 LW |
174 | return d; |
175 | } | |
176 | if (uv < 0x4000000) { | |
eb160463 GS |
177 | *d++ = (U8)(( uv >> 24) | 0xf8); |
178 | *d++ = (U8)(((uv >> 18) & 0x3f) | 0x80); | |
179 | *d++ = (U8)(((uv >> 12) & 0x3f) | 0x80); | |
180 | *d++ = (U8)(((uv >> 6) & 0x3f) | 0x80); | |
181 | *d++ = (U8)(( uv & 0x3f) | 0x80); | |
a0ed51b3 LW |
182 | return d; |
183 | } | |
184 | if (uv < 0x80000000) { | |
eb160463 GS |
185 | *d++ = (U8)(( uv >> 30) | 0xfc); |
186 | *d++ = (U8)(((uv >> 24) & 0x3f) | 0x80); | |
187 | *d++ = (U8)(((uv >> 18) & 0x3f) | 0x80); | |
188 | *d++ = (U8)(((uv >> 12) & 0x3f) | 0x80); | |
189 | *d++ = (U8)(((uv >> 6) & 0x3f) | 0x80); | |
190 | *d++ = (U8)(( uv & 0x3f) | 0x80); | |
a0ed51b3 LW |
191 | return d; |
192 | } | |
6588300d | 193 | #ifdef UTF8_QUAD_MAX |
d7578b48 | 194 | if (uv < UTF8_QUAD_MAX) |
a0ed51b3 LW |
195 | #endif |
196 | { | |
eb160463 GS |
197 | *d++ = 0xfe; /* Can't match U+FEFF! */ |
198 | *d++ = (U8)(((uv >> 30) & 0x3f) | 0x80); | |
199 | *d++ = (U8)(((uv >> 24) & 0x3f) | 0x80); | |
200 | *d++ = (U8)(((uv >> 18) & 0x3f) | 0x80); | |
201 | *d++ = (U8)(((uv >> 12) & 0x3f) | 0x80); | |
202 | *d++ = (U8)(((uv >> 6) & 0x3f) | 0x80); | |
203 | *d++ = (U8)(( uv & 0x3f) | 0x80); | |
a0ed51b3 LW |
204 | return d; |
205 | } | |
6588300d | 206 | #ifdef UTF8_QUAD_MAX |
a0ed51b3 | 207 | { |
eb160463 GS |
208 | *d++ = 0xff; /* Can't match U+FFFE! */ |
209 | *d++ = 0x80; /* 6 Reserved bits */ | |
210 | *d++ = (U8)(((uv >> 60) & 0x0f) | 0x80); /* 2 Reserved bits */ | |
211 | *d++ = (U8)(((uv >> 54) & 0x3f) | 0x80); | |
212 | *d++ = (U8)(((uv >> 48) & 0x3f) | 0x80); | |
213 | *d++ = (U8)(((uv >> 42) & 0x3f) | 0x80); | |
214 | *d++ = (U8)(((uv >> 36) & 0x3f) | 0x80); | |
215 | *d++ = (U8)(((uv >> 30) & 0x3f) | 0x80); | |
216 | *d++ = (U8)(((uv >> 24) & 0x3f) | 0x80); | |
217 | *d++ = (U8)(((uv >> 18) & 0x3f) | 0x80); | |
218 | *d++ = (U8)(((uv >> 12) & 0x3f) | 0x80); | |
219 | *d++ = (U8)(((uv >> 6) & 0x3f) | 0x80); | |
220 | *d++ = (U8)(( uv & 0x3f) | 0x80); | |
a0ed51b3 LW |
221 | return d; |
222 | } | |
223 | #endif | |
537124e4 | 224 | #endif /* Non loop style */ |
a0ed51b3 | 225 | } |
646ca15d | 226 | /* |
07693fe6 KW |
227 | =for apidoc uvchr_to_utf8 |
228 | ||
bcb1a2d4 | 229 | Adds the UTF-8 representation of the native code point C<uv> to the end |
07693fe6 KW |
230 | of the string C<d>; C<d> should have at least C<UTF8_MAXBYTES+1> free |
231 | bytes available. The return value is the pointer to the byte after the | |
232 | end of the new character. In other words, | |
233 | ||
234 | d = uvchr_to_utf8(d, uv); | |
235 | ||
236 | is the recommended wide native character-aware way of saying | |
237 | ||
238 | *(d++) = uv; | |
239 | ||
de69f3af KW |
240 | This function accepts any UV as input. To forbid or warn on non-Unicode code |
241 | points, or those that may be problematic, see L</uvchr_to_utf8_flags>. | |
242 | ||
07693fe6 KW |
243 | =cut |
244 | */ | |
245 | ||
de69f3af KW |
246 | /* This is also a macro */ |
247 | PERL_CALLCONV U8* Perl_uvchr_to_utf8(pTHX_ U8 *d, UV uv); | |
248 | ||
07693fe6 KW |
249 | U8 * |
250 | Perl_uvchr_to_utf8(pTHX_ U8 *d, UV uv) | |
251 | { | |
de69f3af | 252 | return uvchr_to_utf8(d, uv); |
07693fe6 KW |
253 | } |
254 | ||
de69f3af KW |
255 | /* |
256 | =for apidoc uvchr_to_utf8_flags | |
257 | ||
258 | Adds the UTF-8 representation of the native code point C<uv> to the end | |
259 | of the string C<d>; C<d> should have at least C<UTF8_MAXBYTES+1> free | |
260 | bytes available. The return value is the pointer to the byte after the | |
261 | end of the new character. In other words, | |
262 | ||
263 | d = uvchr_to_utf8_flags(d, uv, flags); | |
264 | ||
265 | or, in most cases, | |
266 | ||
267 | d = uvchr_to_utf8_flags(d, uv, 0); | |
268 | ||
269 | This is the Unicode-aware way of saying | |
270 | ||
271 | *(d++) = uv; | |
272 | ||
273 | This function will convert to UTF-8 (and not warn) even code points that aren't | |
274 | legal Unicode or are problematic, unless C<flags> contains one or more of the | |
275 | following flags: | |
276 | ||
277 | If C<uv> is a Unicode surrogate code point and UNICODE_WARN_SURROGATE is set, | |
278 | the function will raise a warning, provided UTF8 warnings are enabled. If instead | |
279 | UNICODE_DISALLOW_SURROGATE is set, the function will fail and return NULL. | |
280 | If both flags are set, the function will both warn and return NULL. | |
281 | ||
282 | The UNICODE_WARN_NONCHAR and UNICODE_DISALLOW_NONCHAR flags correspondingly | |
283 | affect how the function handles a Unicode non-character. And likewise, the | |
284 | UNICODE_WARN_SUPER and UNICODE_DISALLOW_SUPER flags, affect the handling of | |
285 | code points that are | |
286 | above the Unicode maximum of 0x10FFFF. Code points above 0x7FFF_FFFF (which are | |
287 | even less portable) can be warned and/or disallowed even if other above-Unicode | |
288 | code points are accepted, by the UNICODE_WARN_FE_FF and UNICODE_DISALLOW_FE_FF | |
289 | flags. | |
290 | ||
291 | And finally, the flag UNICODE_WARN_ILLEGAL_INTERCHANGE selects all four of the | |
292 | above WARN flags; and UNICODE_DISALLOW_ILLEGAL_INTERCHANGE selects all four | |
293 | DISALLOW flags. | |
294 | ||
295 | =cut | |
296 | */ | |
297 | ||
298 | /* This is also a macro */ | |
299 | PERL_CALLCONV U8* Perl_uvchr_to_utf8_flags(pTHX_ U8 *d, UV uv, UV flags); | |
300 | ||
07693fe6 KW |
301 | U8 * |
302 | Perl_uvchr_to_utf8_flags(pTHX_ U8 *d, UV uv, UV flags) | |
303 | { | |
de69f3af | 304 | return uvchr_to_utf8_flags(d, uv, flags); |
07693fe6 KW |
305 | } |
306 | ||
307 | /* | |
646ca15d | 308 | |
f7d739d1 | 309 | Tests if the first C<len> bytes of string C<s> form a valid UTF-8 |
bcb1a2d4 KW |
310 | character. Note that an INVARIANT (i.e. ASCII on non-EBCDIC) character is a |
311 | valid UTF-8 character. The number of bytes in the UTF-8 character | |
646ca15d JH |
312 | will be returned if it is valid, otherwise 0. |
313 | ||
314 | This is the "slow" version as opposed to the "fast" version which is | |
315 | the "unrolled" IS_UTF8_CHAR(). E.g. for t/uni/class.t the speed | |
316 | difference is a factor of 2 to 3. For lengths (UTF8SKIP(s)) of four | |
317 | or less you should use the IS_UTF8_CHAR(), for lengths of five or more | |
318 | you should use the _slow(). In practice this means that the _slow() | |
319 | will be used very rarely, since the maximum Unicode code point (as of | |
320 | Unicode 4.1) is U+10FFFF, which encodes in UTF-8 to four bytes. Only | |
537124e4 | 321 | the "Perl extended UTF-8" (e.g, the infamous 'v-strings') will encode into |
646ca15d JH |
322 | five bytes or more. |
323 | ||
324 | =cut */ | |
7af276bc | 325 | PERL_STATIC_INLINE STRLEN |
5f66b61c | 326 | S_is_utf8_char_slow(const U8 *s, const STRLEN len) |
646ca15d | 327 | { |
cd7e6c88 | 328 | dTHX; /* The function called below requires thread context */ |
646ca15d | 329 | |
cd7e6c88 | 330 | STRLEN actual_len; |
646ca15d | 331 | |
cd7e6c88 | 332 | PERL_ARGS_ASSERT_IS_UTF8_CHAR_SLOW; |
646ca15d | 333 | |
18712bce | 334 | utf8n_to_uvchr(s, len, &actual_len, UTF8_CHECK_ONLY); |
646ca15d | 335 | |
cd7e6c88 | 336 | return (actual_len == (STRLEN) -1) ? 0 : actual_len; |
646ca15d | 337 | } |
9041c2e3 NIS |
338 | |
339 | /* | |
492a624f KW |
340 | =for apidoc is_utf8_char_buf |
341 | ||
342 | Returns the number of bytes that comprise the first UTF-8 encoded character in | |
343 | buffer C<buf>. C<buf_end> should point to one position beyond the end of the | |
344 | buffer. 0 is returned if C<buf> does not point to a complete, valid UTF-8 | |
345 | encoded character. | |
346 | ||
347 | Note that an INVARIANT character (i.e. ASCII on non-EBCDIC | |
348 | machines) is a valid UTF-8 character. | |
349 | ||
350 | =cut */ | |
351 | ||
352 | STRLEN | |
353 | Perl_is_utf8_char_buf(const U8 *buf, const U8* buf_end) | |
354 | { | |
355 | ||
356 | STRLEN len; | |
357 | ||
358 | PERL_ARGS_ASSERT_IS_UTF8_CHAR_BUF; | |
359 | ||
360 | if (buf_end <= buf) { | |
361 | return 0; | |
362 | } | |
363 | ||
364 | len = buf_end - buf; | |
365 | if (len > UTF8SKIP(buf)) { | |
366 | len = UTF8SKIP(buf); | |
367 | } | |
368 | ||
492a624f KW |
369 | if (IS_UTF8_CHAR_FAST(len)) |
370 | return IS_UTF8_CHAR(buf, len) ? len : 0; | |
492a624f KW |
371 | return is_utf8_char_slow(buf, len); |
372 | } | |
373 | ||
374 | /* | |
87cea99e | 375 | =for apidoc is_utf8_char |
eebe1485 | 376 | |
5da9da9e | 377 | Tests if some arbitrary number of bytes begins in a valid UTF-8 |
2bbc8d55 SP |
378 | character. Note that an INVARIANT (i.e. ASCII on non-EBCDIC machines) |
379 | character is a valid UTF-8 character. The actual number of bytes in the UTF-8 | |
380 | character will be returned if it is valid, otherwise 0. | |
9041c2e3 | 381 | |
76848387 | 382 | This function is deprecated due to the possibility that malformed input could |
a1433954 | 383 | cause reading beyond the end of the input buffer. Use L</is_utf8_char_buf> |
76848387 | 384 | instead. |
e0328548 | 385 | |
82686b01 | 386 | =cut */ |
76848387 | 387 | |
067a85ef | 388 | STRLEN |
668b6d8d | 389 | Perl_is_utf8_char(const U8 *s) |
386d01d6 | 390 | { |
7918f24d | 391 | PERL_ARGS_ASSERT_IS_UTF8_CHAR; |
492a624f | 392 | |
76848387 | 393 | /* Assumes we have enough space, which is why this is deprecated */ |
492a624f | 394 | return is_utf8_char_buf(s, s + UTF8SKIP(s)); |
386d01d6 GS |
395 | } |
396 | ||
eaf7a4d2 | 397 | |
6662521e | 398 | /* |
87cea99e | 399 | =for apidoc is_utf8_string |
6662521e | 400 | |
a1433954 | 401 | Returns true if the first C<len> bytes of string C<s> form a valid |
9f7e3d64 | 402 | UTF-8 string, false otherwise. If C<len> is 0, it will be calculated |
e0328548 KW |
403 | using C<strlen(s)> (which means if you use this option, that C<s> has to have a |
404 | terminating NUL byte). Note that all characters being ASCII constitute 'a | |
405 | valid UTF-8 string'. | |
6662521e | 406 | |
a1433954 | 407 | See also L</is_ascii_string>(), L</is_utf8_string_loclen>(), and L</is_utf8_string_loc>(). |
768c67ee | 408 | |
6662521e GS |
409 | =cut |
410 | */ | |
411 | ||
8e84507e | 412 | bool |
668b6d8d | 413 | Perl_is_utf8_string(const U8 *s, STRLEN len) |
6662521e | 414 | { |
35da51f7 | 415 | const U8* const send = s + (len ? len : strlen((const char *)s)); |
7fc63493 | 416 | const U8* x = s; |
067a85ef | 417 | |
7918f24d | 418 | PERL_ARGS_ASSERT_IS_UTF8_STRING; |
1aa99e6b | 419 | |
6662521e | 420 | while (x < send) { |
1acdb0da | 421 | /* Inline the easy bits of is_utf8_char() here for speed... */ |
e0328548 KW |
422 | if (UTF8_IS_INVARIANT(*x)) { |
423 | x++; | |
424 | } | |
1acdb0da JH |
425 | else { |
426 | /* ... and call is_utf8_char() only if really needed. */ | |
e0328548 KW |
427 | const STRLEN c = UTF8SKIP(x); |
428 | const U8* const next_char_ptr = x + c; | |
429 | ||
430 | if (next_char_ptr > send) { | |
431 | return FALSE; | |
432 | } | |
433 | ||
768c67ee JH |
434 | if (IS_UTF8_CHAR_FAST(c)) { |
435 | if (!IS_UTF8_CHAR(x, c)) | |
e0328548 | 436 | return FALSE; |
3c614e38 | 437 | } |
e0328548 KW |
438 | else if (! is_utf8_char_slow(x, c)) { |
439 | return FALSE; | |
440 | } | |
441 | x = next_char_ptr; | |
1acdb0da | 442 | } |
6662521e | 443 | } |
768c67ee | 444 | |
067a85ef | 445 | return TRUE; |
6662521e GS |
446 | } |
447 | ||
67e989fb | 448 | /* |
814fafa7 NC |
449 | Implemented as a macro in utf8.h |
450 | ||
87cea99e | 451 | =for apidoc is_utf8_string_loc |
814fafa7 | 452 | |
a1433954 KW |
453 | Like L</is_utf8_string> but stores the location of the failure (in the |
454 | case of "utf8ness failure") or the location C<s>+C<len> (in the case of | |
814fafa7 NC |
455 | "utf8ness success") in the C<ep>. |
456 | ||
a1433954 | 457 | See also L</is_utf8_string_loclen>() and L</is_utf8_string>(). |
814fafa7 | 458 | |
87cea99e | 459 | =for apidoc is_utf8_string_loclen |
81cd54e3 | 460 | |
a1433954 KW |
461 | Like L</is_utf8_string>() but stores the location of the failure (in the |
462 | case of "utf8ness failure") or the location C<s>+C<len> (in the case of | |
768c67ee JH |
463 | "utf8ness success") in the C<ep>, and the number of UTF-8 |
464 | encoded characters in the C<el>. | |
465 | ||
a1433954 | 466 | See also L</is_utf8_string_loc>() and L</is_utf8_string>(). |
81cd54e3 JH |
467 | |
468 | =cut | |
469 | */ | |
470 | ||
471 | bool | |
668b6d8d | 472 | Perl_is_utf8_string_loclen(const U8 *s, STRLEN len, const U8 **ep, STRLEN *el) |
81cd54e3 | 473 | { |
35da51f7 | 474 | const U8* const send = s + (len ? len : strlen((const char *)s)); |
7fc63493 | 475 | const U8* x = s; |
81cd54e3 | 476 | STRLEN c; |
3ebfea28 | 477 | STRLEN outlen = 0; |
7918f24d NC |
478 | |
479 | PERL_ARGS_ASSERT_IS_UTF8_STRING_LOCLEN; | |
81cd54e3 | 480 | |
81cd54e3 | 481 | while (x < send) { |
e0328548 KW |
482 | const U8* next_char_ptr; |
483 | ||
81cd54e3 JH |
484 | /* Inline the easy bits of is_utf8_char() here for speed... */ |
485 | if (UTF8_IS_INVARIANT(*x)) | |
e0328548 | 486 | next_char_ptr = x + 1; |
81cd54e3 | 487 | else { |
768c67ee | 488 | /* ... and call is_utf8_char() only if really needed. */ |
768c67ee | 489 | c = UTF8SKIP(x); |
e0328548 KW |
490 | next_char_ptr = c + x; |
491 | if (next_char_ptr > send) { | |
492 | goto out; | |
493 | } | |
768c67ee JH |
494 | if (IS_UTF8_CHAR_FAST(c)) { |
495 | if (!IS_UTF8_CHAR(x, c)) | |
496 | c = 0; | |
497 | } else | |
498 | c = is_utf8_char_slow(x, c); | |
768c67ee JH |
499 | if (!c) |
500 | goto out; | |
81cd54e3 | 501 | } |
e0328548 | 502 | x = next_char_ptr; |
3ebfea28 | 503 | outlen++; |
81cd54e3 | 504 | } |
768c67ee JH |
505 | |
506 | out: | |
3ebfea28 AL |
507 | if (el) |
508 | *el = outlen; | |
509 | ||
768c67ee JH |
510 | if (ep) |
511 | *ep = x; | |
3ebfea28 | 512 | return (x == send); |
81cd54e3 JH |
513 | } |
514 | ||
515 | /* | |
768c67ee | 516 | |
de69f3af | 517 | =for apidoc utf8n_to_uvchr |
378516de KW |
518 | |
519 | THIS FUNCTION SHOULD BE USED IN ONLY VERY SPECIALIZED CIRCUMSTANCES. | |
de69f3af | 520 | Most code should use L</utf8_to_uvchr_buf>() rather than call this directly. |
67e989fb | 521 | |
9041c2e3 | 522 | Bottom level UTF-8 decode routine. |
de69f3af | 523 | Returns the native code point value of the first character in the string C<s>, |
746afd53 KW |
524 | which is assumed to be in UTF-8 (or UTF-EBCDIC) encoding, and no longer than |
525 | C<curlen> bytes; C<*retlen> (if C<retlen> isn't NULL) will be set to | |
526 | the length, in bytes, of that character. | |
949cf498 KW |
527 | |
528 | The value of C<flags> determines the behavior when C<s> does not point to a | |
529 | well-formed UTF-8 character. If C<flags> is 0, when a malformation is found, | |
524080c4 KW |
530 | zero is returned and C<*retlen> is set so that (S<C<s> + C<*retlen>>) is the |
531 | next possible position in C<s> that could begin a non-malformed character. | |
532 | Also, if UTF-8 warnings haven't been lexically disabled, a warning is raised. | |
949cf498 KW |
533 | |
534 | Various ALLOW flags can be set in C<flags> to allow (and not warn on) | |
535 | individual types of malformations, such as the sequence being overlong (that | |
536 | is, when there is a shorter sequence that can express the same code point; | |
537 | overlong sequences are expressly forbidden in the UTF-8 standard due to | |
538 | potential security issues). Another malformation example is the first byte of | |
539 | a character not being a legal first byte. See F<utf8.h> for the list of such | |
524080c4 KW |
540 | flags. For allowed 0 length strings, this function returns 0; for allowed |
541 | overlong sequences, the computed code point is returned; for all other allowed | |
542 | malformations, the Unicode REPLACEMENT CHARACTER is returned, as these have no | |
543 | determinable reasonable value. | |
949cf498 KW |
544 | |
545 | The UTF8_CHECK_ONLY flag overrides the behavior when a non-allowed (by other | |
546 | flags) malformation is found. If this flag is set, the routine assumes that | |
547 | the caller will raise a warning, and this function will silently just set | |
d088425d KW |
548 | C<retlen> to C<-1> (cast to C<STRLEN>) and return zero. |
549 | ||
550 | Note that this API requires disambiguation between successful decoding a NUL | |
551 | character, and an error return (unless the UTF8_CHECK_ONLY flag is set), as | |
552 | in both cases, 0 is returned. To disambiguate, upon a zero return, see if the | |
553 | first byte of C<s> is 0 as well. If so, the input was a NUL; if not, the input | |
554 | had an error. | |
949cf498 KW |
555 | |
556 | Certain code points are considered problematic. These are Unicode surrogates, | |
746afd53 | 557 | Unicode non-characters, and code points above the Unicode maximum of 0x10FFFF. |
949cf498 | 558 | By default these are considered regular code points, but certain situations |
5eafe189 | 559 | warrant special handling for them. If C<flags> contains |
949cf498 KW |
560 | UTF8_DISALLOW_ILLEGAL_INTERCHANGE, all three classes are treated as |
561 | malformations and handled as such. The flags UTF8_DISALLOW_SURROGATE, | |
562 | UTF8_DISALLOW_NONCHAR, and UTF8_DISALLOW_SUPER (meaning above the legal Unicode | |
563 | maximum) can be set to disallow these categories individually. | |
564 | ||
565 | The flags UTF8_WARN_ILLEGAL_INTERCHANGE, UTF8_WARN_SURROGATE, | |
566 | UTF8_WARN_NONCHAR, and UTF8_WARN_SUPER will cause warning messages to be raised | |
567 | for their respective categories, but otherwise the code points are considered | |
568 | valid (not malformations). To get a category to both be treated as a | |
569 | malformation and raise a warning, specify both the WARN and DISALLOW flags. | |
570 | (But note that warnings are not raised if lexically disabled nor if | |
571 | UTF8_CHECK_ONLY is also specified.) | |
572 | ||
573 | Very large code points (above 0x7FFF_FFFF) are considered more problematic than | |
574 | the others that are above the Unicode legal maximum. There are several | |
eb83ed87 KW |
575 | reasons: they requre at least 32 bits to represent them on ASCII platforms, are |
576 | not representable at all on EBCDIC platforms, and the original UTF-8 | |
577 | specification never went above this number (the current 0x10FFFF limit was | |
578 | imposed later). (The smaller ones, those that fit into 32 bits, are | |
579 | representable by a UV on ASCII platforms, but not by an IV, which means that | |
580 | the number of operations that can be performed on them is quite restricted.) | |
581 | The UTF-8 encoding on ASCII platforms for these large code points begins with a | |
582 | byte containing 0xFE or 0xFF. The UTF8_DISALLOW_FE_FF flag will cause them to | |
583 | be treated as malformations, while allowing smaller above-Unicode code points. | |
584 | (Of course UTF8_DISALLOW_SUPER will treat all above-Unicode code points, | |
585 | including these, as malformations.) Similarly, UTF8_WARN_FE_FF acts just like | |
586 | the other WARN flags, but applies just to these code points. | |
949cf498 KW |
587 | |
588 | All other code points corresponding to Unicode characters, including private | |
589 | use and those yet to be assigned, are never considered malformed and never | |
590 | warn. | |
67e989fb | 591 | |
37607a96 PK |
592 | =cut |
593 | */ | |
67e989fb | 594 | |
a0ed51b3 | 595 | UV |
de69f3af | 596 | Perl_utf8n_to_uvchr(pTHX_ const U8 *s, STRLEN curlen, STRLEN *retlen, U32 flags) |
a0ed51b3 | 597 | { |
97aff369 | 598 | dVAR; |
d4c19fe8 | 599 | const U8 * const s0 = s; |
eb83ed87 | 600 | U8 overflow_byte = '\0'; /* Save byte in case of overflow */ |
0b8d30e8 | 601 | U8 * send; |
eb83ed87 KW |
602 | UV uv = *s; |
603 | STRLEN expectlen; | |
949cf498 | 604 | SV* sv = NULL; |
eb83ed87 KW |
605 | UV outlier_ret = 0; /* return value when input is in error or problematic |
606 | */ | |
607 | UV pack_warn = 0; /* Save result of packWARN() for later */ | |
608 | bool unexpected_non_continuation = FALSE; | |
609 | bool overflowed = FALSE; | |
2f8f112e | 610 | bool do_overlong_test = TRUE; /* May have to skip this test */ |
a0dbb045 | 611 | |
eb83ed87 | 612 | const char* const malformed_text = "Malformed UTF-8 character"; |
7918f24d | 613 | |
de69f3af | 614 | PERL_ARGS_ASSERT_UTF8N_TO_UVCHR; |
a0dbb045 | 615 | |
eb83ed87 KW |
616 | /* The order of malformation tests here is important. We should consume as |
617 | * few bytes as possible in order to not skip any valid character. This is | |
618 | * required by the Unicode Standard (section 3.9 of Unicode 6.0); see also | |
619 | * http://unicode.org/reports/tr36 for more discussion as to why. For | |
620 | * example, once we've done a UTF8SKIP, we can tell the expected number of | |
621 | * bytes, and could fail right off the bat if the input parameters indicate | |
622 | * that there are too few available. But it could be that just that first | |
623 | * byte is garbled, and the intended character occupies fewer bytes. If we | |
624 | * blindly assumed that the first byte is correct, and skipped based on | |
625 | * that number, we could skip over a valid input character. So instead, we | |
626 | * always examine the sequence byte-by-byte. | |
627 | * | |
628 | * We also should not consume too few bytes, otherwise someone could inject | |
629 | * things. For example, an input could be deliberately designed to | |
630 | * overflow, and if this code bailed out immediately upon discovering that, | |
e2660c54 | 631 | * returning to the caller C<*retlen> pointing to the very next byte (one |
eb83ed87 KW |
632 | * which is actually part of of the overflowing sequence), that could look |
633 | * legitimate to the caller, which could discard the initial partial | |
634 | * sequence and process the rest, inappropriately */ | |
635 | ||
636 | /* Zero length strings, if allowed, of necessity are zero */ | |
b5b9af04 | 637 | if (UNLIKELY(curlen == 0)) { |
eb83ed87 KW |
638 | if (retlen) { |
639 | *retlen = 0; | |
640 | } | |
a0dbb045 | 641 | |
eb83ed87 KW |
642 | if (flags & UTF8_ALLOW_EMPTY) { |
643 | return 0; | |
644 | } | |
645 | if (! (flags & UTF8_CHECK_ONLY)) { | |
646 | sv = sv_2mortal(Perl_newSVpvf(aTHX_ "%s (empty string)", malformed_text)); | |
647 | } | |
0c443dc2 JH |
648 | goto malformed; |
649 | } | |
650 | ||
eb83ed87 KW |
651 | expectlen = UTF8SKIP(s); |
652 | ||
653 | /* A well-formed UTF-8 character, as the vast majority of calls to this | |
654 | * function will be for, has this expected length. For efficiency, set | |
655 | * things up here to return it. It will be overriden only in those rare | |
656 | * cases where a malformation is found */ | |
657 | if (retlen) { | |
658 | *retlen = expectlen; | |
659 | } | |
660 | ||
661 | /* An invariant is trivially well-formed */ | |
1d72bdf6 | 662 | if (UTF8_IS_INVARIANT(uv)) { |
de69f3af | 663 | return uv; |
a0ed51b3 | 664 | } |
67e989fb | 665 | |
eb83ed87 | 666 | /* A continuation character can't start a valid sequence */ |
b5b9af04 | 667 | if (UNLIKELY(UTF8_IS_CONTINUATION(uv))) { |
eb83ed87 KW |
668 | if (flags & UTF8_ALLOW_CONTINUATION) { |
669 | if (retlen) { | |
670 | *retlen = 1; | |
671 | } | |
672 | return UNICODE_REPLACEMENT; | |
673 | } | |
ba210ebe | 674 | |
eb83ed87 KW |
675 | if (! (flags & UTF8_CHECK_ONLY)) { |
676 | sv = sv_2mortal(Perl_newSVpvf(aTHX_ "%s (unexpected continuation byte 0x%02x, with no preceding start byte)", malformed_text, *s0)); | |
677 | } | |
678 | curlen = 1; | |
ba210ebe JH |
679 | goto malformed; |
680 | } | |
9041c2e3 | 681 | |
dcd27b3c KW |
682 | /* Here is not a continuation byte, nor an invariant. The only thing left |
683 | * is a start byte (possibly for an overlong) */ | |
684 | ||
1d72bdf6 | 685 | #ifdef EBCDIC |
bc3632a8 | 686 | uv = NATIVE_UTF8_TO_I8(uv); |
1d72bdf6 NIS |
687 | #endif |
688 | ||
eb83ed87 KW |
689 | /* Remove the leading bits that indicate the number of bytes in the |
690 | * character's whole UTF-8 sequence, leaving just the bits that are part of | |
691 | * the value */ | |
692 | uv &= UTF_START_MASK(expectlen); | |
ba210ebe | 693 | |
eb83ed87 KW |
694 | /* Now, loop through the remaining bytes in the character's sequence, |
695 | * accumulating each into the working value as we go. Be sure to not look | |
696 | * past the end of the input string */ | |
0b8d30e8 KW |
697 | send = (U8*) s0 + ((expectlen <= curlen) ? expectlen : curlen); |
698 | ||
eb83ed87 | 699 | for (s = s0 + 1; s < send; s++) { |
b5b9af04 | 700 | if (LIKELY(UTF8_IS_CONTINUATION(*s))) { |
eb83ed87 KW |
701 | #ifndef EBCDIC /* Can't overflow in EBCDIC */ |
702 | if (uv & UTF_ACCUMULATION_OVERFLOW_MASK) { | |
703 | ||
704 | /* The original implementors viewed this malformation as more | |
705 | * serious than the others (though I, khw, don't understand | |
706 | * why, since other malformations also give very very wrong | |
707 | * results), so there is no way to turn off checking for it. | |
708 | * Set a flag, but keep going in the loop, so that we absorb | |
709 | * the rest of the bytes that comprise the character. */ | |
710 | overflowed = TRUE; | |
711 | overflow_byte = *s; /* Save for warning message's use */ | |
712 | } | |
713 | #endif | |
8850bf83 | 714 | uv = UTF8_ACCUMULATE(uv, *s); |
eb83ed87 KW |
715 | } |
716 | else { | |
717 | /* Here, found a non-continuation before processing all expected | |
718 | * bytes. This byte begins a new character, so quit, even if | |
719 | * allowing this malformation. */ | |
720 | unexpected_non_continuation = TRUE; | |
721 | break; | |
722 | } | |
723 | } /* End of loop through the character's bytes */ | |
724 | ||
725 | /* Save how many bytes were actually in the character */ | |
726 | curlen = s - s0; | |
727 | ||
728 | /* The loop above finds two types of malformations: non-continuation and/or | |
729 | * overflow. The non-continuation malformation is really a too-short | |
730 | * malformation, as it means that the current character ended before it was | |
731 | * expected to (being terminated prematurely by the beginning of the next | |
732 | * character, whereas in the too-short malformation there just are too few | |
733 | * bytes available to hold the character. In both cases, the check below | |
734 | * that we have found the expected number of bytes would fail if executed.) | |
735 | * Thus the non-continuation malformation is really unnecessary, being a | |
736 | * subset of the too-short malformation. But there may be existing | |
737 | * applications that are expecting the non-continuation type, so we retain | |
738 | * it, and return it in preference to the too-short malformation. (If this | |
739 | * code were being written from scratch, the two types might be collapsed | |
740 | * into one.) I, khw, am also giving priority to returning the | |
741 | * non-continuation and too-short malformations over overflow when multiple | |
742 | * ones are present. I don't know of any real reason to prefer one over | |
743 | * the other, except that it seems to me that multiple-byte errors trumps | |
744 | * errors from a single byte */ | |
b5b9af04 | 745 | if (UNLIKELY(unexpected_non_continuation)) { |
eb83ed87 KW |
746 | if (!(flags & UTF8_ALLOW_NON_CONTINUATION)) { |
747 | if (! (flags & UTF8_CHECK_ONLY)) { | |
748 | if (curlen == 1) { | |
749 | sv = sv_2mortal(Perl_newSVpvf(aTHX_ "%s (unexpected non-continuation byte 0x%02x, immediately after start byte 0x%02x)", malformed_text, *s, *s0)); | |
750 | } | |
751 | else { | |
752 | sv = sv_2mortal(Perl_newSVpvf(aTHX_ "%s (unexpected non-continuation byte 0x%02x, %d bytes after start byte 0x%02x, expected %d bytes)", malformed_text, *s, (int) curlen, *s0, (int)expectlen)); | |
a0dbb045 JH |
753 | } |
754 | } | |
eb83ed87 KW |
755 | goto malformed; |
756 | } | |
757 | uv = UNICODE_REPLACEMENT; | |
2f8f112e KW |
758 | |
759 | /* Skip testing for overlongs, as the REPLACEMENT may not be the same | |
760 | * as what the original expectations were. */ | |
761 | do_overlong_test = FALSE; | |
eb83ed87 KW |
762 | if (retlen) { |
763 | *retlen = curlen; | |
764 | } | |
765 | } | |
b5b9af04 | 766 | else if (UNLIKELY(curlen < expectlen)) { |
eb83ed87 KW |
767 | if (! (flags & UTF8_ALLOW_SHORT)) { |
768 | if (! (flags & UTF8_CHECK_ONLY)) { | |
769 | sv = sv_2mortal(Perl_newSVpvf(aTHX_ "%s (%d byte%s, need %d, after start byte 0x%02x)", malformed_text, (int)curlen, curlen == 1 ? "" : "s", (int)expectlen, *s0)); | |
a0dbb045 | 770 | } |
eb83ed87 KW |
771 | goto malformed; |
772 | } | |
773 | uv = UNICODE_REPLACEMENT; | |
2f8f112e | 774 | do_overlong_test = FALSE; |
eb83ed87 KW |
775 | if (retlen) { |
776 | *retlen = curlen; | |
777 | } | |
778 | } | |
779 | ||
780 | #ifndef EBCDIC /* EBCDIC allows FE, FF, can't overflow */ | |
2f8f112e | 781 | if ((*s0 & 0xFE) == 0xFE /* matches both FE, FF */ |
eb83ed87 KW |
782 | && (flags & (UTF8_WARN_FE_FF|UTF8_DISALLOW_FE_FF))) |
783 | { | |
784 | /* By adding UTF8_CHECK_ONLY to the test, we avoid unnecessary | |
785 | * generation of the sv, since no warnings are raised under CHECK */ | |
786 | if ((flags & (UTF8_WARN_FE_FF|UTF8_CHECK_ONLY)) == UTF8_WARN_FE_FF | |
787 | && ckWARN_d(WARN_UTF8)) | |
788 | { | |
42303544 KW |
789 | /* This message is deliberately not of the same syntax as the other |
790 | * messages for malformations, for backwards compatibility in the | |
791 | * unlikely event that code is relying on its precise earlier text | |
792 | */ | |
eb83ed87 KW |
793 | sv = sv_2mortal(Perl_newSVpvf(aTHX_ "%s Code point beginning with byte 0x%02X is not Unicode, and not portable", malformed_text, *s0)); |
794 | pack_warn = packWARN(WARN_UTF8); | |
795 | } | |
796 | if (flags & UTF8_DISALLOW_FE_FF) { | |
797 | goto malformed; | |
ba210ebe | 798 | } |
ba210ebe | 799 | } |
b5b9af04 | 800 | if (UNLIKELY(overflowed)) { |
ba210ebe | 801 | |
eb83ed87 KW |
802 | /* If the first byte is FF, it will overflow a 32-bit word. If the |
803 | * first byte is FE, it will overflow a signed 32-bit word. The | |
804 | * above preserves backward compatibility, since its message was used | |
805 | * in earlier versions of this code in preference to overflow */ | |
806 | sv = sv_2mortal(Perl_newSVpvf(aTHX_ "%s (overflow at byte 0x%02x, after start byte 0x%02x)", malformed_text, overflow_byte, *s0)); | |
ba210ebe | 807 | goto malformed; |
eb83ed87 KW |
808 | } |
809 | #endif | |
810 | ||
2f8f112e | 811 | if (do_overlong_test |
5aaebcb3 | 812 | && expectlen > (STRLEN) OFFUNISKIP(uv) |
2f8f112e KW |
813 | && ! (flags & UTF8_ALLOW_LONG)) |
814 | { | |
eb83ed87 KW |
815 | /* The overlong malformation has lower precedence than the others. |
816 | * Note that if this malformation is allowed, we return the actual | |
817 | * value, instead of the replacement character. This is because this | |
818 | * value is actually well-defined. */ | |
819 | if (! (flags & UTF8_CHECK_ONLY)) { | |
5aaebcb3 | 820 | sv = sv_2mortal(Perl_newSVpvf(aTHX_ "%s (%d byte%s, need %d, after start byte 0x%02x)", malformed_text, (int)expectlen, expectlen == 1 ? "": "s", OFFUNISKIP(uv), *s0)); |
eb83ed87 KW |
821 | } |
822 | goto malformed; | |
823 | } | |
824 | ||
825 | /* Here, the input is considered to be well-formed , but could be a | |
826 | * problematic code point that is not allowed by the input parameters. */ | |
827 | if (uv >= UNICODE_SURROGATE_FIRST /* isn't problematic if < this */ | |
828 | && (flags & (UTF8_DISALLOW_ILLEGAL_INTERCHANGE | |
829 | |UTF8_WARN_ILLEGAL_INTERCHANGE))) | |
830 | { | |
949cf498 | 831 | if (UNICODE_IS_SURROGATE(uv)) { |
eb83ed87 KW |
832 | if ((flags & (UTF8_WARN_SURROGATE|UTF8_CHECK_ONLY)) == UTF8_WARN_SURROGATE |
833 | && ckWARN2_d(WARN_UTF8, WARN_SURROGATE)) | |
834 | { | |
111d382d | 835 | sv = sv_2mortal(Perl_newSVpvf(aTHX_ "UTF-16 surrogate U+%04"UVXf"", uv)); |
eb83ed87 | 836 | pack_warn = packWARN2(WARN_UTF8, WARN_SURROGATE); |
949cf498 KW |
837 | } |
838 | if (flags & UTF8_DISALLOW_SURROGATE) { | |
839 | goto disallowed; | |
840 | } | |
841 | } | |
949cf498 | 842 | else if ((uv > PERL_UNICODE_MAX)) { |
eb83ed87 KW |
843 | if ((flags & (UTF8_WARN_SUPER|UTF8_CHECK_ONLY)) == UTF8_WARN_SUPER |
844 | && ckWARN2_d(WARN_UTF8, WARN_NON_UNICODE)) | |
845 | { | |
111d382d | 846 | sv = sv_2mortal(Perl_newSVpvf(aTHX_ "Code point 0x%04"UVXf" is not Unicode, may not be portable", uv)); |
eb83ed87 | 847 | pack_warn = packWARN2(WARN_UTF8, WARN_NON_UNICODE); |
949cf498 KW |
848 | } |
849 | if (flags & UTF8_DISALLOW_SUPER) { | |
850 | goto disallowed; | |
851 | } | |
852 | } | |
4190d317 KW |
853 | else if (UNICODE_IS_NONCHAR(uv)) { |
854 | if ((flags & (UTF8_WARN_NONCHAR|UTF8_CHECK_ONLY)) == UTF8_WARN_NONCHAR | |
855 | && ckWARN2_d(WARN_UTF8, WARN_NONCHAR)) | |
856 | { | |
857 | sv = sv_2mortal(Perl_newSVpvf(aTHX_ "Unicode non-character U+%04"UVXf" is illegal for open interchange", uv)); | |
858 | pack_warn = packWARN2(WARN_UTF8, WARN_NONCHAR); | |
859 | } | |
860 | if (flags & UTF8_DISALLOW_NONCHAR) { | |
861 | goto disallowed; | |
862 | } | |
863 | } | |
949cf498 | 864 | |
eb83ed87 | 865 | if (sv) { |
de69f3af KW |
866 | outlier_ret = uv; /* Note we don't bother to convert to native, |
867 | as all the outlier code points are the same | |
868 | in both ASCII and EBCDIC */ | |
eb83ed87 KW |
869 | goto do_warn; |
870 | } | |
871 | ||
949cf498 KW |
872 | /* Here, this is not considered a malformed character, so drop through |
873 | * to return it */ | |
a0ed51b3 | 874 | } |
ba210ebe | 875 | |
de69f3af | 876 | return UNI_TO_NATIVE(uv); |
ba210ebe | 877 | |
eb83ed87 KW |
878 | /* There are three cases which get to beyond this point. In all 3 cases: |
879 | * <sv> if not null points to a string to print as a warning. | |
880 | * <curlen> is what <*retlen> should be set to if UTF8_CHECK_ONLY isn't | |
881 | * set. | |
882 | * <outlier_ret> is what return value to use if UTF8_CHECK_ONLY isn't set. | |
883 | * This is done by initializing it to 0, and changing it only | |
884 | * for case 1). | |
885 | * The 3 cases are: | |
886 | * 1) The input is valid but problematic, and to be warned about. The | |
887 | * return value is the resultant code point; <*retlen> is set to | |
888 | * <curlen>, the number of bytes that comprise the code point. | |
889 | * <pack_warn> contains the result of packWARN() for the warning | |
890 | * types. The entry point for this case is the label <do_warn>; | |
891 | * 2) The input is a valid code point but disallowed by the parameters to | |
892 | * this function. The return value is 0. If UTF8_CHECK_ONLY is set, | |
893 | * <*relen> is -1; otherwise it is <curlen>, the number of bytes that | |
894 | * comprise the code point. <pack_warn> contains the result of | |
895 | * packWARN() for the warning types. The entry point for this case is | |
896 | * the label <disallowed>. | |
897 | * 3) The input is malformed. The return value is 0. If UTF8_CHECK_ONLY | |
898 | * is set, <*relen> is -1; otherwise it is <curlen>, the number of | |
899 | * bytes that comprise the malformation. All such malformations are | |
900 | * assumed to be warning type <utf8>. The entry point for this case | |
901 | * is the label <malformed>. | |
902 | */ | |
949cf498 | 903 | |
ba210ebe JH |
904 | malformed: |
905 | ||
eb83ed87 KW |
906 | if (sv && ckWARN_d(WARN_UTF8)) { |
907 | pack_warn = packWARN(WARN_UTF8); | |
908 | } | |
909 | ||
910 | disallowed: | |
911 | ||
fcc8fcf6 | 912 | if (flags & UTF8_CHECK_ONLY) { |
ba210ebe | 913 | if (retlen) |
10edeb5d | 914 | *retlen = ((STRLEN) -1); |
ba210ebe JH |
915 | return 0; |
916 | } | |
917 | ||
eb83ed87 | 918 | do_warn: |
5b311467 | 919 | |
eb83ed87 KW |
920 | if (pack_warn) { /* <pack_warn> was initialized to 0, and changed only |
921 | if warnings are to be raised. */ | |
f555bc63 | 922 | const char * const string = SvPVX_const(sv); |
a0dbb045 | 923 | |
f555bc63 KW |
924 | if (PL_op) |
925 | Perl_warner(aTHX_ pack_warn, "%s in %s", string, OP_DESC(PL_op)); | |
926 | else | |
927 | Perl_warner(aTHX_ pack_warn, "%s", string); | |
a0dbb045 JH |
928 | } |
929 | ||
eb83ed87 KW |
930 | if (retlen) { |
931 | *retlen = curlen; | |
932 | } | |
ba210ebe | 933 | |
eb83ed87 | 934 | return outlier_ret; |
a0ed51b3 LW |
935 | } |
936 | ||
8e84507e | 937 | /* |
ec5f19d0 KW |
938 | =for apidoc utf8_to_uvchr_buf |
939 | ||
940 | Returns the native code point of the first character in the string C<s> which | |
941 | is assumed to be in UTF-8 encoding; C<send> points to 1 beyond the end of C<s>. | |
524080c4 | 942 | C<*retlen> will be set to the length, in bytes, of that character. |
ec5f19d0 | 943 | |
524080c4 KW |
944 | If C<s> does not point to a well-formed UTF-8 character and UTF8 warnings are |
945 | enabled, zero is returned and C<*retlen> is set (if C<retlen> isn't | |
173db420 KW |
946 | NULL) to -1. If those warnings are off, the computed value, if well-defined |
947 | (or the Unicode REPLACEMENT CHARACTER if not), is silently returned, and | |
948 | C<*retlen> is set (if C<retlen> isn't NULL) so that (S<C<s> + C<*retlen>>) is | |
949 | the next possible position in C<s> that could begin a non-malformed character. | |
de69f3af | 950 | See L</utf8n_to_uvchr> for details on when the REPLACEMENT CHARACTER is |
173db420 | 951 | returned. |
ec5f19d0 KW |
952 | |
953 | =cut | |
954 | */ | |
955 | ||
956 | ||
957 | UV | |
958 | Perl_utf8_to_uvchr_buf(pTHX_ const U8 *s, const U8 *send, STRLEN *retlen) | |
959 | { | |
ec5f19d0 KW |
960 | assert(s < send); |
961 | ||
962 | return utf8n_to_uvchr(s, send - s, retlen, | |
963 | ckWARN_d(WARN_UTF8) ? 0 : UTF8_ALLOW_ANY); | |
964 | } | |
965 | ||
27d6c58a | 966 | /* Like L</utf8_to_uvchr_buf>(), but should only be called when it is known that |
3986bb7c | 967 | * there are no malformations in the input UTF-8 string C<s>. surrogates, |
57b0056d | 968 | * non-character code points, and non-Unicode code points are allowed. */ |
27d6c58a KW |
969 | |
970 | UV | |
971 | Perl_valid_utf8_to_uvchr(pTHX_ const U8 *s, STRLEN *retlen) | |
972 | { | |
010ab96b KW |
973 | UV expectlen = UTF8SKIP(s); |
974 | const U8* send = s + expectlen; | |
9ff2f0f7 | 975 | UV uv = *s; |
3986bb7c | 976 | |
27d6c58a KW |
977 | PERL_ARGS_ASSERT_VALID_UTF8_TO_UVCHR; |
978 | ||
010ab96b KW |
979 | if (retlen) { |
980 | *retlen = expectlen; | |
981 | } | |
982 | ||
983 | /* An invariant is trivially returned */ | |
984 | if (expectlen == 1) { | |
9ff2f0f7 | 985 | return uv; |
010ab96b KW |
986 | } |
987 | ||
9ff2f0f7 KW |
988 | #ifdef EBCDIC |
989 | uv = NATIVE_UTF8_TO_I8(uv); | |
990 | #endif | |
991 | ||
010ab96b KW |
992 | /* Remove the leading bits that indicate the number of bytes, leaving just |
993 | * the bits that are part of the value */ | |
994 | uv &= UTF_START_MASK(expectlen); | |
995 | ||
996 | /* Now, loop through the remaining bytes, accumulating each into the | |
997 | * working total as we go. (I khw tried unrolling the loop for up to 4 | |
998 | * bytes, but there was no performance improvement) */ | |
999 | for (++s; s < send; s++) { | |
1000 | uv = UTF8_ACCUMULATE(uv, *s); | |
1001 | } | |
1002 | ||
3986bb7c | 1003 | return UNI_TO_NATIVE(uv); |
010ab96b | 1004 | |
27d6c58a KW |
1005 | } |
1006 | ||
ec5f19d0 | 1007 | /* |
87cea99e | 1008 | =for apidoc utf8_to_uvchr |
9041c2e3 | 1009 | |
6ee84de2 | 1010 | Returns the native code point of the first character in the string C<s> |
1e54db1a | 1011 | which is assumed to be in UTF-8 encoding; C<retlen> will be set to the |
9041c2e3 NIS |
1012 | length, in bytes, of that character. |
1013 | ||
4b88fb76 | 1014 | Some, but not all, UTF-8 malformations are detected, and in fact, some |
977c1d31 KW |
1015 | malformed input could cause reading beyond the end of the input buffer, which |
1016 | is why this function is deprecated. Use L</utf8_to_uvchr_buf> instead. | |
4b88fb76 | 1017 | |
524080c4 KW |
1018 | If C<s> points to one of the detected malformations, and UTF8 warnings are |
1019 | enabled, zero is returned and C<*retlen> is set (if C<retlen> isn't | |
1020 | NULL) to -1. If those warnings are off, the computed value if well-defined (or | |
1021 | the Unicode REPLACEMENT CHARACTER, if not) is silently returned, and C<*retlen> | |
1022 | is set (if C<retlen> isn't NULL) so that (S<C<s> + C<*retlen>>) is the | |
1023 | next possible position in C<s> that could begin a non-malformed character. | |
de69f3af | 1024 | See L</utf8n_to_uvchr> for details on when the REPLACEMENT CHARACTER is returned. |
9041c2e3 NIS |
1025 | |
1026 | =cut | |
1027 | */ | |
1028 | ||
1029 | UV | |
7fc63493 | 1030 | Perl_utf8_to_uvchr(pTHX_ const U8 *s, STRLEN *retlen) |
9041c2e3 | 1031 | { |
7918f24d NC |
1032 | PERL_ARGS_ASSERT_UTF8_TO_UVCHR; |
1033 | ||
2ff6c191 | 1034 | return utf8_to_uvchr_buf(s, s + UTF8_MAXBYTES, retlen); |
9041c2e3 NIS |
1035 | } |
1036 | ||
1037 | /* | |
ec5f19d0 KW |
1038 | =for apidoc utf8_to_uvuni_buf |
1039 | ||
de69f3af KW |
1040 | Only in very rare circumstances should code need to be dealing in Unicode |
1041 | (as opposed to native) code points. In those few cases, use | |
1042 | C<L<NATIVE_TO_UNI(utf8_to_uvchr_buf(...))|/utf8_to_uvchr_buf>> instead. | |
4f83cdcd KW |
1043 | |
1044 | Returns the Unicode (not-native) code point of the first character in the | |
1045 | string C<s> which | |
ec5f19d0 KW |
1046 | is assumed to be in UTF-8 encoding; C<send> points to 1 beyond the end of C<s>. |
1047 | C<retlen> will be set to the length, in bytes, of that character. | |
1048 | ||
524080c4 KW |
1049 | If C<s> does not point to a well-formed UTF-8 character and UTF8 warnings are |
1050 | enabled, zero is returned and C<*retlen> is set (if C<retlen> isn't | |
1051 | NULL) to -1. If those warnings are off, the computed value if well-defined (or | |
1052 | the Unicode REPLACEMENT CHARACTER, if not) is silently returned, and C<*retlen> | |
1053 | is set (if C<retlen> isn't NULL) so that (S<C<s> + C<*retlen>>) is the | |
1054 | next possible position in C<s> that could begin a non-malformed character. | |
de69f3af | 1055 | See L</utf8n_to_uvchr> for details on when the REPLACEMENT CHARACTER is returned. |
ec5f19d0 KW |
1056 | |
1057 | =cut | |
1058 | */ | |
1059 | ||
1060 | UV | |
1061 | Perl_utf8_to_uvuni_buf(pTHX_ const U8 *s, const U8 *send, STRLEN *retlen) | |
1062 | { | |
1063 | PERL_ARGS_ASSERT_UTF8_TO_UVUNI_BUF; | |
1064 | ||
1065 | assert(send > s); | |
1066 | ||
1067 | /* Call the low level routine asking for checks */ | |
de69f3af KW |
1068 | return NATIVE_TO_UNI(Perl_utf8n_to_uvchr(aTHX_ s, send -s, retlen, |
1069 | ckWARN_d(WARN_UTF8) ? 0 : UTF8_ALLOW_ANY)); | |
ec5f19d0 KW |
1070 | } |
1071 | ||
5495102a KW |
1072 | /* DEPRECATED! |
1073 | * Like L</utf8_to_uvuni_buf>(), but should only be called when it is known that | |
2114036c | 1074 | * there are no malformations in the input UTF-8 string C<s>. Surrogates, |
3986bb7c | 1075 | * non-character code points, and non-Unicode code points are allowed */ |
27d6c58a KW |
1076 | |
1077 | UV | |
1078 | Perl_valid_utf8_to_uvuni(pTHX_ const U8 *s, STRLEN *retlen) | |
1079 | { | |
1080 | PERL_ARGS_ASSERT_VALID_UTF8_TO_UVUNI; | |
1081 | ||
010ab96b | 1082 | return NATIVE_TO_UNI(valid_utf8_to_uvchr(s, retlen)); |
27d6c58a KW |
1083 | } |
1084 | ||
ec5f19d0 | 1085 | /* |
87cea99e | 1086 | =for apidoc utf8_to_uvuni |
9041c2e3 NIS |
1087 | |
1088 | Returns the Unicode code point of the first character in the string C<s> | |
1e54db1a | 1089 | which is assumed to be in UTF-8 encoding; C<retlen> will be set to the |
9041c2e3 NIS |
1090 | length, in bytes, of that character. |
1091 | ||
4b88fb76 | 1092 | Some, but not all, UTF-8 malformations are detected, and in fact, some |
977c1d31 | 1093 | malformed input could cause reading beyond the end of the input buffer, which |
4f83cdcd KW |
1094 | is one reason why this function is deprecated. The other is that only in |
1095 | extremely limited circumstances should the Unicode versus native code point be | |
de69f3af | 1096 | of any interest to you. See L</utf8_to_uvuni_buf> for alternatives. |
9041c2e3 | 1097 | |
524080c4 KW |
1098 | If C<s> points to one of the detected malformations, and UTF8 warnings are |
1099 | enabled, zero is returned and C<*retlen> is set (if C<retlen> doesn't point to | |
1100 | NULL) to -1. If those warnings are off, the computed value if well-defined (or | |
1101 | the Unicode REPLACEMENT CHARACTER, if not) is silently returned, and C<*retlen> | |
1102 | is set (if C<retlen> isn't NULL) so that (S<C<s> + C<*retlen>>) is the | |
1103 | next possible position in C<s> that could begin a non-malformed character. | |
de69f3af | 1104 | See L</utf8n_to_uvchr> for details on when the REPLACEMENT CHARACTER is returned. |
8e84507e NIS |
1105 | |
1106 | =cut | |
1107 | */ | |
1108 | ||
1109 | UV | |
7fc63493 | 1110 | Perl_utf8_to_uvuni(pTHX_ const U8 *s, STRLEN *retlen) |
8e84507e | 1111 | { |
7918f24d NC |
1112 | PERL_ARGS_ASSERT_UTF8_TO_UVUNI; |
1113 | ||
5495102a | 1114 | return NATIVE_TO_UNI(valid_utf8_to_uvchr(s, retlen)); |
8e84507e NIS |
1115 | } |
1116 | ||
b76347f2 | 1117 | /* |
87cea99e | 1118 | =for apidoc utf8_length |
b76347f2 JH |
1119 | |
1120 | Return the length of the UTF-8 char encoded string C<s> in characters. | |
02eb7b47 JH |
1121 | Stops at C<e> (inclusive). If C<e E<lt> s> or if the scan would end |
1122 | up past C<e>, croaks. | |
b76347f2 JH |
1123 | |
1124 | =cut | |
1125 | */ | |
1126 | ||
1127 | STRLEN | |
35a4481c | 1128 | Perl_utf8_length(pTHX_ const U8 *s, const U8 *e) |
b76347f2 | 1129 | { |
97aff369 | 1130 | dVAR; |
b76347f2 JH |
1131 | STRLEN len = 0; |
1132 | ||
7918f24d NC |
1133 | PERL_ARGS_ASSERT_UTF8_LENGTH; |
1134 | ||
8850bf83 JH |
1135 | /* Note: cannot use UTF8_IS_...() too eagerly here since e.g. |
1136 | * the bitops (especially ~) can create illegal UTF-8. | |
1137 | * In other words: in Perl UTF-8 is not just for Unicode. */ | |
1138 | ||
a3b680e6 AL |
1139 | if (e < s) |
1140 | goto warn_and_return; | |
b76347f2 | 1141 | while (s < e) { |
4cbf4130 | 1142 | s += UTF8SKIP(s); |
8e91ec7f AV |
1143 | len++; |
1144 | } | |
1145 | ||
1146 | if (e != s) { | |
1147 | len--; | |
1148 | warn_and_return: | |
9b387841 NC |
1149 | if (PL_op) |
1150 | Perl_ck_warner_d(aTHX_ packWARN(WARN_UTF8), | |
1151 | "%s in %s", unees, OP_DESC(PL_op)); | |
1152 | else | |
61a12c31 | 1153 | Perl_ck_warner_d(aTHX_ packWARN(WARN_UTF8), "%s", unees); |
b76347f2 JH |
1154 | } |
1155 | ||
1156 | return len; | |
1157 | } | |
1158 | ||
b06226ff | 1159 | /* |
87cea99e | 1160 | =for apidoc utf8_distance |
b06226ff | 1161 | |
1e54db1a | 1162 | Returns the number of UTF-8 characters between the UTF-8 pointers C<a> |
b06226ff JH |
1163 | and C<b>. |
1164 | ||
1165 | WARNING: use only if you *know* that the pointers point inside the | |
1166 | same UTF-8 buffer. | |
1167 | ||
37607a96 PK |
1168 | =cut |
1169 | */ | |
a0ed51b3 | 1170 | |
02eb7b47 | 1171 | IV |
35a4481c | 1172 | Perl_utf8_distance(pTHX_ const U8 *a, const U8 *b) |
a0ed51b3 | 1173 | { |
7918f24d NC |
1174 | PERL_ARGS_ASSERT_UTF8_DISTANCE; |
1175 | ||
bf1665bc | 1176 | return (a < b) ? -1 * (IV) utf8_length(a, b) : (IV) utf8_length(b, a); |
a0ed51b3 LW |
1177 | } |
1178 | ||
b06226ff | 1179 | /* |
87cea99e | 1180 | =for apidoc utf8_hop |
b06226ff | 1181 | |
8850bf83 JH |
1182 | Return the UTF-8 pointer C<s> displaced by C<off> characters, either |
1183 | forward or backward. | |
b06226ff JH |
1184 | |
1185 | WARNING: do not use the following unless you *know* C<off> is within | |
8850bf83 JH |
1186 | the UTF-8 data pointed to by C<s> *and* that on entry C<s> is aligned |
1187 | on the first byte of character or just after the last byte of a character. | |
b06226ff | 1188 | |
37607a96 PK |
1189 | =cut |
1190 | */ | |
a0ed51b3 LW |
1191 | |
1192 | U8 * | |
4373e329 | 1193 | Perl_utf8_hop(pTHX_ const U8 *s, I32 off) |
a0ed51b3 | 1194 | { |
7918f24d NC |
1195 | PERL_ARGS_ASSERT_UTF8_HOP; |
1196 | ||
96a5add6 | 1197 | PERL_UNUSED_CONTEXT; |
8850bf83 JH |
1198 | /* Note: cannot use UTF8_IS_...() too eagerly here since e.g |
1199 | * the bitops (especially ~) can create illegal UTF-8. | |
1200 | * In other words: in Perl UTF-8 is not just for Unicode. */ | |
1201 | ||
a0ed51b3 LW |
1202 | if (off >= 0) { |
1203 | while (off--) | |
1204 | s += UTF8SKIP(s); | |
1205 | } | |
1206 | else { | |
1207 | while (off++) { | |
1208 | s--; | |
8850bf83 JH |
1209 | while (UTF8_IS_CONTINUATION(*s)) |
1210 | s--; | |
a0ed51b3 LW |
1211 | } |
1212 | } | |
4373e329 | 1213 | return (U8 *)s; |
a0ed51b3 LW |
1214 | } |
1215 | ||
6940069f | 1216 | /* |
fed3ba5d NC |
1217 | =for apidoc bytes_cmp_utf8 |
1218 | ||
a1433954 KW |
1219 | Compares the sequence of characters (stored as octets) in C<b>, C<blen> with the |
1220 | sequence of characters (stored as UTF-8) in C<u>, C<ulen>. Returns 0 if they are | |
fed3ba5d NC |
1221 | equal, -1 or -2 if the first string is less than the second string, +1 or +2 |
1222 | if the first string is greater than the second string. | |
1223 | ||
1224 | -1 or +1 is returned if the shorter string was identical to the start of the | |
1225 | longer string. -2 or +2 is returned if the was a difference between characters | |
1226 | within the strings. | |
1227 | ||
1228 | =cut | |
1229 | */ | |
1230 | ||
1231 | int | |
1232 | Perl_bytes_cmp_utf8(pTHX_ const U8 *b, STRLEN blen, const U8 *u, STRLEN ulen) | |
1233 | { | |
1234 | const U8 *const bend = b + blen; | |
1235 | const U8 *const uend = u + ulen; | |
1236 | ||
1237 | PERL_ARGS_ASSERT_BYTES_CMP_UTF8; | |
1238 | ||
1239 | PERL_UNUSED_CONTEXT; | |
1240 | ||
1241 | while (b < bend && u < uend) { | |
1242 | U8 c = *u++; | |
1243 | if (!UTF8_IS_INVARIANT(c)) { | |
1244 | if (UTF8_IS_DOWNGRADEABLE_START(c)) { | |
1245 | if (u < uend) { | |
1246 | U8 c1 = *u++; | |
1247 | if (UTF8_IS_CONTINUATION(c1)) { | |
94bb8c36 | 1248 | c = TWO_BYTE_UTF8_TO_NATIVE(c, c1); |
fed3ba5d NC |
1249 | } else { |
1250 | Perl_ck_warner_d(aTHX_ packWARN(WARN_UTF8), | |
1251 | "Malformed UTF-8 character " | |
1252 | "(unexpected non-continuation byte 0x%02x" | |
1253 | ", immediately after start byte 0x%02x)" | |
1254 | /* Dear diag.t, it's in the pod. */ | |
1255 | "%s%s", c1, c, | |
1256 | PL_op ? " in " : "", | |
1257 | PL_op ? OP_DESC(PL_op) : ""); | |
1258 | return -2; | |
1259 | } | |
1260 | } else { | |
1261 | if (PL_op) | |
1262 | Perl_ck_warner_d(aTHX_ packWARN(WARN_UTF8), | |
1263 | "%s in %s", unees, OP_DESC(PL_op)); | |
1264 | else | |
61a12c31 | 1265 | Perl_ck_warner_d(aTHX_ packWARN(WARN_UTF8), "%s", unees); |
fed3ba5d NC |
1266 | return -2; /* Really want to return undef :-) */ |
1267 | } | |
1268 | } else { | |
1269 | return -2; | |
1270 | } | |
1271 | } | |
1272 | if (*b != c) { | |
1273 | return *b < c ? -2 : +2; | |
1274 | } | |
1275 | ++b; | |
1276 | } | |
1277 | ||
1278 | if (b == bend && u == uend) | |
1279 | return 0; | |
1280 | ||
1281 | return b < bend ? +1 : -1; | |
1282 | } | |
1283 | ||
1284 | /* | |
87cea99e | 1285 | =for apidoc utf8_to_bytes |
6940069f | 1286 | |
2bbc8d55 | 1287 | Converts a string C<s> of length C<len> from UTF-8 into native byte encoding. |
a1433954 KW |
1288 | Unlike L</bytes_to_utf8>, this over-writes the original string, and |
1289 | updates C<len> to contain the new length. | |
67e989fb | 1290 | Returns zero on failure, setting C<len> to -1. |
6940069f | 1291 | |
a1433954 | 1292 | If you need a copy of the string, see L</bytes_from_utf8>. |
95be277c | 1293 | |
6940069f GS |
1294 | =cut |
1295 | */ | |
1296 | ||
1297 | U8 * | |
37607a96 | 1298 | Perl_utf8_to_bytes(pTHX_ U8 *s, STRLEN *len) |
6940069f | 1299 | { |
d4c19fe8 AL |
1300 | U8 * const save = s; |
1301 | U8 * const send = s + *len; | |
6940069f | 1302 | U8 *d; |
246fae53 | 1303 | |
7918f24d NC |
1304 | PERL_ARGS_ASSERT_UTF8_TO_BYTES; |
1305 | ||
1e54db1a | 1306 | /* ensure valid UTF-8 and chars < 256 before updating string */ |
d4c19fe8 | 1307 | while (s < send) { |
d59937ca KW |
1308 | if (! UTF8_IS_INVARIANT(*s)) { |
1309 | if (! UTF8_IS_NEXT_CHAR_DOWNGRADEABLE(s, send)) { | |
1310 | *len = ((STRLEN) -1); | |
1311 | return 0; | |
1312 | } | |
1313 | s++; | |
dcad2880 | 1314 | } |
d59937ca | 1315 | s++; |
246fae53 | 1316 | } |
dcad2880 JH |
1317 | |
1318 | d = s = save; | |
6940069f | 1319 | while (s < send) { |
80e0b38f KW |
1320 | U8 c = *s++; |
1321 | if (! UTF8_IS_INVARIANT(c)) { | |
1322 | /* Then it is two-byte encoded */ | |
1323 | c = TWO_BYTE_UTF8_TO_NATIVE(c, *s); | |
1324 | s++; | |
1325 | } | |
1326 | *d++ = c; | |
6940069f GS |
1327 | } |
1328 | *d = '\0'; | |
246fae53 | 1329 | *len = d - save; |
6940069f GS |
1330 | return save; |
1331 | } | |
1332 | ||
1333 | /* | |
87cea99e | 1334 | =for apidoc bytes_from_utf8 |
f9a63242 | 1335 | |
2bbc8d55 | 1336 | Converts a string C<s> of length C<len> from UTF-8 into native byte encoding. |
a1433954 | 1337 | Unlike L</utf8_to_bytes> but like L</bytes_to_utf8>, returns a pointer to |
ef9edfd0 JH |
1338 | the newly-created string, and updates C<len> to contain the new |
1339 | length. Returns the original string if no conversion occurs, C<len> | |
1340 | is unchanged. Do nothing if C<is_utf8> points to 0. Sets C<is_utf8> to | |
2bbc8d55 SP |
1341 | 0 if C<s> is converted or consisted entirely of characters that are invariant |
1342 | in utf8 (i.e., US-ASCII on non-EBCDIC machines). | |
f9a63242 | 1343 | |
37607a96 PK |
1344 | =cut |
1345 | */ | |
f9a63242 JH |
1346 | |
1347 | U8 * | |
e1ec3a88 | 1348 | Perl_bytes_from_utf8(pTHX_ const U8 *s, STRLEN *len, bool *is_utf8) |
f9a63242 | 1349 | { |
f9a63242 | 1350 | U8 *d; |
e1ec3a88 AL |
1351 | const U8 *start = s; |
1352 | const U8 *send; | |
f9a63242 JH |
1353 | I32 count = 0; |
1354 | ||
7918f24d NC |
1355 | PERL_ARGS_ASSERT_BYTES_FROM_UTF8; |
1356 | ||
96a5add6 | 1357 | PERL_UNUSED_CONTEXT; |
f9a63242 | 1358 | if (!*is_utf8) |
73d840c0 | 1359 | return (U8 *)start; |
f9a63242 | 1360 | |
1e54db1a | 1361 | /* ensure valid UTF-8 and chars < 256 before converting string */ |
f9a63242 | 1362 | for (send = s + *len; s < send;) { |
d59937ca KW |
1363 | if (! UTF8_IS_INVARIANT(*s)) { |
1364 | if (! UTF8_IS_NEXT_CHAR_DOWNGRADEABLE(s, send)) { | |
73d840c0 | 1365 | return (U8 *)start; |
d59937ca KW |
1366 | } |
1367 | count++; | |
1368 | s++; | |
db42d148 | 1369 | } |
d59937ca | 1370 | s++; |
f9a63242 JH |
1371 | } |
1372 | ||
35da51f7 | 1373 | *is_utf8 = FALSE; |
f9a63242 | 1374 | |
212542aa | 1375 | Newx(d, (*len) - count + 1, U8); |
ef9edfd0 | 1376 | s = start; start = d; |
f9a63242 JH |
1377 | while (s < send) { |
1378 | U8 c = *s++; | |
1a91c45d | 1379 | if (! UTF8_IS_INVARIANT(c)) { |
c4d5f83a | 1380 | /* Then it is two-byte encoded */ |
1a91c45d KW |
1381 | c = TWO_BYTE_UTF8_TO_NATIVE(c, *s); |
1382 | s++; | |
c4d5f83a NIS |
1383 | } |
1384 | *d++ = c; | |
f9a63242 JH |
1385 | } |
1386 | *d = '\0'; | |
1387 | *len = d - start; | |
73d840c0 | 1388 | return (U8 *)start; |
f9a63242 JH |
1389 | } |
1390 | ||
1391 | /* | |
87cea99e | 1392 | =for apidoc bytes_to_utf8 |
6940069f | 1393 | |
ff97e5cf KW |
1394 | Converts a string C<s> of length C<len> bytes from the native encoding into |
1395 | UTF-8. | |
6662521e | 1396 | Returns a pointer to the newly-created string, and sets C<len> to |
ff97e5cf | 1397 | reflect the new length in bytes. |
6940069f | 1398 | |
2bbc8d55 SP |
1399 | A NUL character will be written after the end of the string. |
1400 | ||
1401 | If you want to convert to UTF-8 from encodings other than | |
1402 | the native (Latin1 or EBCDIC), | |
a1433954 | 1403 | see L</sv_recode_to_utf8>(). |
c9ada85f | 1404 | |
497711e7 | 1405 | =cut |
6940069f GS |
1406 | */ |
1407 | ||
c682ebef FC |
1408 | /* This logic is duplicated in sv_catpvn_flags, so any bug fixes will |
1409 | likewise need duplication. */ | |
1410 | ||
6940069f | 1411 | U8* |
35a4481c | 1412 | Perl_bytes_to_utf8(pTHX_ const U8 *s, STRLEN *len) |
6940069f | 1413 | { |
35a4481c | 1414 | const U8 * const send = s + (*len); |
6940069f GS |
1415 | U8 *d; |
1416 | U8 *dst; | |
7918f24d NC |
1417 | |
1418 | PERL_ARGS_ASSERT_BYTES_TO_UTF8; | |
96a5add6 | 1419 | PERL_UNUSED_CONTEXT; |
6940069f | 1420 | |
212542aa | 1421 | Newx(d, (*len) * 2 + 1, U8); |
6940069f GS |
1422 | dst = d; |
1423 | ||
1424 | while (s < send) { | |
55d09dc8 KW |
1425 | append_utf8_from_native_byte(*s, &d); |
1426 | s++; | |
6940069f GS |
1427 | } |
1428 | *d = '\0'; | |
6662521e | 1429 | *len = d-dst; |
6940069f GS |
1430 | return dst; |
1431 | } | |
1432 | ||
a0ed51b3 | 1433 | /* |
dea0fc0b | 1434 | * Convert native (big-endian) or reversed (little-endian) UTF-16 to UTF-8. |
a0ed51b3 LW |
1435 | * |
1436 | * Destination must be pre-extended to 3/2 source. Do not use in-place. | |
1437 | * We optimize for native, for obvious reasons. */ | |
1438 | ||
1439 | U8* | |
dea0fc0b | 1440 | Perl_utf16_to_utf8(pTHX_ U8* p, U8* d, I32 bytelen, I32 *newlen) |
a0ed51b3 | 1441 | { |
dea0fc0b JH |
1442 | U8* pend; |
1443 | U8* dstart = d; | |
1444 | ||
7918f24d NC |
1445 | PERL_ARGS_ASSERT_UTF16_TO_UTF8; |
1446 | ||
dea0fc0b | 1447 | if (bytelen & 1) |
f5992bc4 | 1448 | Perl_croak(aTHX_ "panic: utf16_to_utf8: odd bytelen %"UVuf, (UV)bytelen); |
dea0fc0b JH |
1449 | |
1450 | pend = p + bytelen; | |
1451 | ||
a0ed51b3 | 1452 | while (p < pend) { |
dea0fc0b JH |
1453 | UV uv = (p[0] << 8) + p[1]; /* UTF-16BE */ |
1454 | p += 2; | |
56d37426 KW |
1455 | if (UNI_IS_INVARIANT(uv)) { |
1456 | *d++ = LATIN1_TO_NATIVE((U8) uv); | |
a0ed51b3 LW |
1457 | continue; |
1458 | } | |
56d37426 KW |
1459 | if (uv <= MAX_UTF8_TWO_BYTE) { |
1460 | *d++ = UTF8_TWO_BYTE_HI(UNI_TO_NATIVE(uv)); | |
1461 | *d++ = UTF8_TWO_BYTE_LO(UNI_TO_NATIVE(uv)); | |
a0ed51b3 LW |
1462 | continue; |
1463 | } | |
46956fad KW |
1464 | #define FIRST_HIGH_SURROGATE UNICODE_SURROGATE_FIRST |
1465 | #define LAST_HIGH_SURROGATE 0xDBFF | |
1466 | #define FIRST_LOW_SURROGATE 0xDC00 | |
1467 | #define LAST_LOW_SURROGATE UNICODE_SURROGATE_LAST | |
1468 | if (uv >= FIRST_HIGH_SURROGATE && uv <= LAST_HIGH_SURROGATE) { | |
01ea242b | 1469 | if (p >= pend) { |
dea0fc0b | 1470 | Perl_croak(aTHX_ "Malformed UTF-16 surrogate"); |
01ea242b NC |
1471 | } else { |
1472 | UV low = (p[0] << 8) + p[1]; | |
1473 | p += 2; | |
46956fad | 1474 | if (low < FIRST_LOW_SURROGATE || low > LAST_LOW_SURROGATE) |
01ea242b | 1475 | Perl_croak(aTHX_ "Malformed UTF-16 surrogate"); |
46956fad KW |
1476 | uv = ((uv - FIRST_HIGH_SURROGATE) << 10) |
1477 | + (low - FIRST_LOW_SURROGATE) + 0x10000; | |
01ea242b | 1478 | } |
46956fad | 1479 | } else if (uv >= FIRST_LOW_SURROGATE && uv <= LAST_LOW_SURROGATE) { |
dbde1951 | 1480 | Perl_croak(aTHX_ "Malformed UTF-16 surrogate"); |
a0ed51b3 | 1481 | } |
56d37426 KW |
1482 | #ifdef EBCDIC |
1483 | d = uvoffuni_to_utf8_flags(d, uv, 0); | |
1484 | #else | |
a0ed51b3 | 1485 | if (uv < 0x10000) { |
eb160463 GS |
1486 | *d++ = (U8)(( uv >> 12) | 0xe0); |
1487 | *d++ = (U8)(((uv >> 6) & 0x3f) | 0x80); | |
1488 | *d++ = (U8)(( uv & 0x3f) | 0x80); | |
a0ed51b3 LW |
1489 | continue; |
1490 | } | |
1491 | else { | |
eb160463 GS |
1492 | *d++ = (U8)(( uv >> 18) | 0xf0); |
1493 | *d++ = (U8)(((uv >> 12) & 0x3f) | 0x80); | |
1494 | *d++ = (U8)(((uv >> 6) & 0x3f) | 0x80); | |
1495 | *d++ = (U8)(( uv & 0x3f) | 0x80); | |
a0ed51b3 LW |
1496 | continue; |
1497 | } | |
56d37426 | 1498 | #endif |
a0ed51b3 | 1499 | } |
dea0fc0b | 1500 | *newlen = d - dstart; |
a0ed51b3 LW |
1501 | return d; |
1502 | } | |
1503 | ||
1504 | /* Note: this one is slightly destructive of the source. */ | |
1505 | ||
1506 | U8* | |
dea0fc0b | 1507 | Perl_utf16_to_utf8_reversed(pTHX_ U8* p, U8* d, I32 bytelen, I32 *newlen) |
a0ed51b3 LW |
1508 | { |
1509 | U8* s = (U8*)p; | |
d4c19fe8 | 1510 | U8* const send = s + bytelen; |
7918f24d NC |
1511 | |
1512 | PERL_ARGS_ASSERT_UTF16_TO_UTF8_REVERSED; | |
1513 | ||
e0ea5e2d NC |
1514 | if (bytelen & 1) |
1515 | Perl_croak(aTHX_ "panic: utf16_to_utf8_reversed: odd bytelen %"UVuf, | |
1516 | (UV)bytelen); | |
1517 | ||
a0ed51b3 | 1518 | while (s < send) { |
d4c19fe8 | 1519 | const U8 tmp = s[0]; |
a0ed51b3 LW |
1520 | s[0] = s[1]; |
1521 | s[1] = tmp; | |
1522 | s += 2; | |
1523 | } | |
dea0fc0b | 1524 | return utf16_to_utf8(p, d, bytelen, newlen); |
a0ed51b3 LW |
1525 | } |
1526 | ||
922e8cb4 KW |
1527 | bool |
1528 | Perl__is_uni_FOO(pTHX_ const U8 classnum, const UV c) | |
1529 | { | |
1530 | U8 tmpbuf[UTF8_MAXBYTES+1]; | |
1531 | uvchr_to_utf8(tmpbuf, c); | |
1532 | return _is_utf8_FOO(classnum, tmpbuf); | |
1533 | } | |
1534 | ||
c3fd2246 KW |
1535 | /* for now these are all defined (inefficiently) in terms of the utf8 versions. |
1536 | * Note that the macros in handy.h that call these short-circuit calling them | |
1537 | * for Latin-1 range inputs */ | |
a0ed51b3 LW |
1538 | |
1539 | bool | |
84afefe6 | 1540 | Perl_is_uni_alnum(pTHX_ UV c) |
a0ed51b3 | 1541 | { |
89ebb4a3 | 1542 | U8 tmpbuf[UTF8_MAXBYTES+1]; |
230880c1 | 1543 | uvchr_to_utf8(tmpbuf, c); |
922e8cb4 | 1544 | return _is_utf8_FOO(_CC_WORDCHAR, tmpbuf); |
a0ed51b3 LW |
1545 | } |
1546 | ||
1547 | bool | |
5092f92a KW |
1548 | Perl_is_uni_alnumc(pTHX_ UV c) |
1549 | { | |
1550 | U8 tmpbuf[UTF8_MAXBYTES+1]; | |
1551 | uvchr_to_utf8(tmpbuf, c); | |
922e8cb4 | 1552 | return _is_utf8_FOO(_CC_ALPHANUMERIC, tmpbuf); |
5092f92a KW |
1553 | } |
1554 | ||
f9ae8fb6 JD |
1555 | /* Internal function so we can deprecate the external one, and call |
1556 | this one from other deprecated functions in this file */ | |
1557 | ||
1558 | PERL_STATIC_INLINE bool | |
61b19385 KW |
1559 | S_is_utf8_idfirst(pTHX_ const U8 *p) |
1560 | { | |
1561 | dVAR; | |
1562 | ||
1563 | if (*p == '_') | |
1564 | return TRUE; | |
1565 | /* is_utf8_idstart would be more logical. */ | |
1566 | return is_utf8_common(p, &PL_utf8_idstart, "IdStart"); | |
1567 | } | |
1568 | ||
5092f92a | 1569 | bool |
84afefe6 | 1570 | Perl_is_uni_idfirst(pTHX_ UV c) |
a0ed51b3 | 1571 | { |
89ebb4a3 | 1572 | U8 tmpbuf[UTF8_MAXBYTES+1]; |
230880c1 | 1573 | uvchr_to_utf8(tmpbuf, c); |
61b19385 | 1574 | return S_is_utf8_idfirst(aTHX_ tmpbuf); |
a0ed51b3 LW |
1575 | } |
1576 | ||
1577 | bool | |
eba68aa0 KW |
1578 | Perl__is_uni_perl_idcont(pTHX_ UV c) |
1579 | { | |
1580 | U8 tmpbuf[UTF8_MAXBYTES+1]; | |
1581 | uvchr_to_utf8(tmpbuf, c); | |
1582 | return _is_utf8_perl_idcont(tmpbuf); | |
1583 | } | |
1584 | ||
1585 | bool | |
f91dcd13 KW |
1586 | Perl__is_uni_perl_idstart(pTHX_ UV c) |
1587 | { | |
1588 | U8 tmpbuf[UTF8_MAXBYTES+1]; | |
1589 | uvchr_to_utf8(tmpbuf, c); | |
1590 | return _is_utf8_perl_idstart(tmpbuf); | |
1591 | } | |
1592 | ||
1593 | bool | |
84afefe6 | 1594 | Perl_is_uni_alpha(pTHX_ UV c) |
a0ed51b3 | 1595 | { |
89ebb4a3 | 1596 | U8 tmpbuf[UTF8_MAXBYTES+1]; |
230880c1 | 1597 | uvchr_to_utf8(tmpbuf, c); |
922e8cb4 | 1598 | return _is_utf8_FOO(_CC_ALPHA, tmpbuf); |
a0ed51b3 LW |
1599 | } |
1600 | ||
1601 | bool | |
84afefe6 | 1602 | Perl_is_uni_ascii(pTHX_ UV c) |
4d61ec05 | 1603 | { |
bc39fe24 | 1604 | return isASCII(c); |
4d61ec05 GS |
1605 | } |
1606 | ||
1607 | bool | |
bdd8600f KW |
1608 | Perl_is_uni_blank(pTHX_ UV c) |
1609 | { | |
2cafb56b | 1610 | return isBLANK_uni(c); |
bdd8600f KW |
1611 | } |
1612 | ||
1613 | bool | |
84afefe6 | 1614 | Perl_is_uni_space(pTHX_ UV c) |
a0ed51b3 | 1615 | { |
add4123a | 1616 | return isSPACE_uni(c); |
a0ed51b3 LW |
1617 | } |
1618 | ||
1619 | bool | |
84afefe6 | 1620 | Perl_is_uni_digit(pTHX_ UV c) |
a0ed51b3 | 1621 | { |
89ebb4a3 | 1622 | U8 tmpbuf[UTF8_MAXBYTES+1]; |
230880c1 | 1623 | uvchr_to_utf8(tmpbuf, c); |
922e8cb4 | 1624 | return _is_utf8_FOO(_CC_DIGIT, tmpbuf); |
a0ed51b3 LW |
1625 | } |
1626 | ||
1627 | bool | |
84afefe6 | 1628 | Perl_is_uni_upper(pTHX_ UV c) |
a0ed51b3 | 1629 | { |
89ebb4a3 | 1630 | U8 tmpbuf[UTF8_MAXBYTES+1]; |
230880c1 | 1631 | uvchr_to_utf8(tmpbuf, c); |
922e8cb4 | 1632 | return _is_utf8_FOO(_CC_UPPER, tmpbuf); |
a0ed51b3 LW |
1633 | } |
1634 | ||
1635 | bool | |
84afefe6 | 1636 | Perl_is_uni_lower(pTHX_ UV c) |
a0ed51b3 | 1637 | { |
89ebb4a3 | 1638 | U8 tmpbuf[UTF8_MAXBYTES+1]; |
230880c1 | 1639 | uvchr_to_utf8(tmpbuf, c); |
922e8cb4 | 1640 | return _is_utf8_FOO(_CC_LOWER, tmpbuf); |
a0ed51b3 LW |
1641 | } |
1642 | ||
1643 | bool | |
84afefe6 | 1644 | Perl_is_uni_cntrl(pTHX_ UV c) |
b8c5462f | 1645 | { |
7b952154 | 1646 | return isCNTRL_L1(c); |
b8c5462f JH |
1647 | } |
1648 | ||
1649 | bool | |
84afefe6 | 1650 | Perl_is_uni_graph(pTHX_ UV c) |
b8c5462f | 1651 | { |
89ebb4a3 | 1652 | U8 tmpbuf[UTF8_MAXBYTES+1]; |
230880c1 | 1653 | uvchr_to_utf8(tmpbuf, c); |
922e8cb4 | 1654 | return _is_utf8_FOO(_CC_GRAPH, tmpbuf); |
b8c5462f JH |
1655 | } |
1656 | ||
1657 | bool | |
84afefe6 | 1658 | Perl_is_uni_print(pTHX_ UV c) |
a0ed51b3 | 1659 | { |
89ebb4a3 | 1660 | U8 tmpbuf[UTF8_MAXBYTES+1]; |
230880c1 | 1661 | uvchr_to_utf8(tmpbuf, c); |
922e8cb4 | 1662 | return _is_utf8_FOO(_CC_PRINT, tmpbuf); |
a0ed51b3 LW |
1663 | } |
1664 | ||
b8c5462f | 1665 | bool |
84afefe6 | 1666 | Perl_is_uni_punct(pTHX_ UV c) |
b8c5462f | 1667 | { |
89ebb4a3 | 1668 | U8 tmpbuf[UTF8_MAXBYTES+1]; |
230880c1 | 1669 | uvchr_to_utf8(tmpbuf, c); |
922e8cb4 | 1670 | return _is_utf8_FOO(_CC_PUNCT, tmpbuf); |
b8c5462f JH |
1671 | } |
1672 | ||
4d61ec05 | 1673 | bool |
84afefe6 | 1674 | Perl_is_uni_xdigit(pTHX_ UV c) |
4d61ec05 | 1675 | { |
4ac6419d | 1676 | return isXDIGIT_uni(c); |
4d61ec05 GS |
1677 | } |
1678 | ||
3a4c58c9 KW |
1679 | UV |
1680 | Perl__to_upper_title_latin1(pTHX_ const U8 c, U8* p, STRLEN *lenp, const char S_or_s) | |
1681 | { | |
1682 | /* We have the latin1-range values compiled into the core, so just use | |
1683 | * those, converting the result to utf8. The only difference between upper | |
1684 | * and title case in this range is that LATIN_SMALL_LETTER_SHARP_S is | |
1685 | * either "SS" or "Ss". Which one to use is passed into the routine in | |
1686 | * 'S_or_s' to avoid a test */ | |
1687 | ||
1688 | UV converted = toUPPER_LATIN1_MOD(c); | |
1689 | ||
1690 | PERL_ARGS_ASSERT__TO_UPPER_TITLE_LATIN1; | |
1691 | ||
1692 | assert(S_or_s == 'S' || S_or_s == 's'); | |
1693 | ||
6f2d5cbc | 1694 | if (UVCHR_IS_INVARIANT(converted)) { /* No difference between the two for |
f4cd282c | 1695 | characters in this range */ |
3a4c58c9 KW |
1696 | *p = (U8) converted; |
1697 | *lenp = 1; | |
1698 | return converted; | |
1699 | } | |
1700 | ||
1701 | /* toUPPER_LATIN1_MOD gives the correct results except for three outliers, | |
1702 | * which it maps to one of them, so as to only have to have one check for | |
1703 | * it in the main case */ | |
1704 | if (UNLIKELY(converted == LATIN_SMALL_LETTER_Y_WITH_DIAERESIS)) { | |
1705 | switch (c) { | |
1706 | case LATIN_SMALL_LETTER_Y_WITH_DIAERESIS: | |
1707 | converted = LATIN_CAPITAL_LETTER_Y_WITH_DIAERESIS; | |
1708 | break; | |
1709 | case MICRO_SIGN: | |
1710 | converted = GREEK_CAPITAL_LETTER_MU; | |
1711 | break; | |
1712 | case LATIN_SMALL_LETTER_SHARP_S: | |
1713 | *(p)++ = 'S'; | |
1714 | *p = S_or_s; | |
1715 | *lenp = 2; | |
1716 | return 'S'; | |
1717 | default: | |
1718 | Perl_croak(aTHX_ "panic: to_upper_title_latin1 did not expect '%c' to map to '%c'", c, LATIN_SMALL_LETTER_Y_WITH_DIAERESIS); | |
118e2215 | 1719 | assert(0); /* NOTREACHED */ |
3a4c58c9 KW |
1720 | } |
1721 | } | |
1722 | ||
1723 | *(p)++ = UTF8_TWO_BYTE_HI(converted); | |
1724 | *p = UTF8_TWO_BYTE_LO(converted); | |
1725 | *lenp = 2; | |
1726 | ||
1727 | return converted; | |
1728 | } | |
1729 | ||
50bda2c3 KW |
1730 | /* Call the function to convert a UTF-8 encoded character to the specified case. |
1731 | * Note that there may be more than one character in the result. | |
1732 | * INP is a pointer to the first byte of the input character | |
1733 | * OUTP will be set to the first byte of the string of changed characters. It | |
1734 | * needs to have space for UTF8_MAXBYTES_CASE+1 bytes | |
1735 | * LENP will be set to the length in bytes of the string of changed characters | |
1736 | * | |
1737 | * The functions return the ordinal of the first character in the string of OUTP */ | |
f90a9a02 KW |
1738 | #define CALL_UPPER_CASE(INP, OUTP, LENP) Perl_to_utf8_case(aTHX_ INP, OUTP, LENP, &PL_utf8_toupper, "ToUc", "utf8::ToSpecUc") |
1739 | #define CALL_TITLE_CASE(INP, OUTP, LENP) Perl_to_utf8_case(aTHX_ INP, OUTP, LENP, &PL_utf8_totitle, "ToTc", "utf8::ToSpecTc") | |
1740 | #define CALL_LOWER_CASE(INP, OUTP, LENP) Perl_to_utf8_case(aTHX_ INP, OUTP, LENP, &PL_utf8_tolower, "ToLc", "utf8::ToSpecLc") | |
50bda2c3 KW |
1741 | |
1742 | /* This additionally has the input parameter SPECIALS, which if non-zero will | |
1743 | * cause this to use the SPECIALS hash for folding (meaning get full case | |
1744 | * folding); otherwise, when zero, this implies a simple case fold */ | |
f90a9a02 | 1745 | #define CALL_FOLD_CASE(INP, OUTP, LENP, SPECIALS) Perl_to_utf8_case(aTHX_ INP, OUTP, LENP, &PL_utf8_tofold, "ToCf", (SPECIALS) ? "utf8::ToSpecCf" : NULL) |
c3fd2246 | 1746 | |
84afefe6 JH |
1747 | UV |
1748 | Perl_to_uni_upper(pTHX_ UV c, U8* p, STRLEN *lenp) | |
a0ed51b3 | 1749 | { |
3a4c58c9 KW |
1750 | dVAR; |
1751 | ||
a1433954 KW |
1752 | /* Convert the Unicode character whose ordinal is <c> to its uppercase |
1753 | * version and store that in UTF-8 in <p> and its length in bytes in <lenp>. | |
1754 | * Note that the <p> needs to be at least UTF8_MAXBYTES_CASE+1 bytes since | |
c3fd2246 KW |
1755 | * the changed version may be longer than the original character. |
1756 | * | |
1757 | * The ordinal of the first character of the changed version is returned | |
1758 | * (but note, as explained above, that there may be more.) */ | |
1759 | ||
7918f24d NC |
1760 | PERL_ARGS_ASSERT_TO_UNI_UPPER; |
1761 | ||
3a4c58c9 KW |
1762 | if (c < 256) { |
1763 | return _to_upper_title_latin1((U8) c, p, lenp, 'S'); | |
1764 | } | |
1765 | ||
0ebc6274 | 1766 | uvchr_to_utf8(p, c); |
3a4c58c9 | 1767 | return CALL_UPPER_CASE(p, p, lenp); |
a0ed51b3 LW |
1768 | } |
1769 | ||
84afefe6 JH |
1770 | UV |
1771 | Perl_to_uni_title(pTHX_ UV c, U8* p, STRLEN *lenp) | |
a0ed51b3 | 1772 | { |
3a4c58c9 KW |
1773 | dVAR; |
1774 | ||
7918f24d NC |
1775 | PERL_ARGS_ASSERT_TO_UNI_TITLE; |
1776 | ||
3a4c58c9 KW |
1777 | if (c < 256) { |
1778 | return _to_upper_title_latin1((U8) c, p, lenp, 's'); | |
1779 | } | |
1780 | ||
0ebc6274 | 1781 | uvchr_to_utf8(p, c); |
3a4c58c9 | 1782 | return CALL_TITLE_CASE(p, p, lenp); |
a0ed51b3 LW |
1783 | } |
1784 | ||
afc16117 KW |
1785 | STATIC U8 |
1786 | S_to_lower_latin1(pTHX_ const U8 c, U8* p, STRLEN *lenp) | |
1787 | { | |
1788 | /* We have the latin1-range values compiled into the core, so just use | |
1789 | * those, converting the result to utf8. Since the result is always just | |
a1433954 | 1790 | * one character, we allow <p> to be NULL */ |
afc16117 KW |
1791 | |
1792 | U8 converted = toLOWER_LATIN1(c); | |
1793 | ||
1794 | if (p != NULL) { | |
6f2d5cbc | 1795 | if (NATIVE_BYTE_IS_INVARIANT(converted)) { |
afc16117 KW |
1796 | *p = converted; |
1797 | *lenp = 1; | |
1798 | } | |
1799 | else { | |
1800 | *p = UTF8_TWO_BYTE_HI(converted); | |
1801 | *(p+1) = UTF8_TWO_BYTE_LO(converted); | |
1802 | *lenp = 2; | |
1803 | } | |
1804 | } | |
1805 | return converted; | |
1806 | } | |
1807 | ||
84afefe6 JH |
1808 | UV |
1809 | Perl_to_uni_lower(pTHX_ UV c, U8* p, STRLEN *lenp) | |
a0ed51b3 | 1810 | { |
968c5e6a KW |
1811 | dVAR; |
1812 | ||
7918f24d NC |
1813 | PERL_ARGS_ASSERT_TO_UNI_LOWER; |
1814 | ||
afc16117 KW |
1815 | if (c < 256) { |
1816 | return to_lower_latin1((U8) c, p, lenp); | |
bca00c02 KW |
1817 | } |
1818 | ||
afc16117 | 1819 | uvchr_to_utf8(p, c); |
968c5e6a | 1820 | return CALL_LOWER_CASE(p, p, lenp); |
a0ed51b3 LW |
1821 | } |
1822 | ||
84afefe6 | 1823 | UV |
51910141 | 1824 | Perl__to_fold_latin1(pTHX_ const U8 c, U8* p, STRLEN *lenp, const unsigned int flags) |
a1dde8de | 1825 | { |
51910141 | 1826 | /* Corresponds to to_lower_latin1(); <flags> bits meanings: |
1ca267a5 | 1827 | * FOLD_FLAGS_NOMIX_ASCII iff non-ASCII to ASCII folds are prohibited |
51910141 | 1828 | * FOLD_FLAGS_FULL iff full folding is to be used; |
1ca267a5 KW |
1829 | * |
1830 | * Not to be used for locale folds | |
51910141 | 1831 | */ |
f673fad4 | 1832 | |
a1dde8de KW |
1833 | UV converted; |
1834 | ||
1835 | PERL_ARGS_ASSERT__TO_FOLD_LATIN1; | |
1836 | ||
1ca267a5 KW |
1837 | assert (! (flags & FOLD_FLAGS_LOCALE)); |
1838 | ||
a1dde8de KW |
1839 | if (c == MICRO_SIGN) { |
1840 | converted = GREEK_SMALL_LETTER_MU; | |
1841 | } | |
51910141 | 1842 | else if ((flags & FOLD_FLAGS_FULL) && c == LATIN_SMALL_LETTER_SHARP_S) { |
1ca267a5 KW |
1843 | |
1844 | /* If can't cross 127/128 boundary, can't return "ss"; instead return | |
1845 | * two U+017F characters, as fc("\df") should eq fc("\x{17f}\x{17f}") | |
1846 | * under those circumstances. */ | |
1847 | if (flags & FOLD_FLAGS_NOMIX_ASCII) { | |
1848 | *lenp = 2 * sizeof(LATIN_SMALL_LETTER_LONG_S_UTF8) - 2; | |
1849 | Copy(LATIN_SMALL_LETTER_LONG_S_UTF8 LATIN_SMALL_LETTER_LONG_S_UTF8, | |
1850 | p, *lenp, U8); | |
1851 | return LATIN_SMALL_LETTER_LONG_S; | |
1852 | } | |
1853 | else { | |
4f489194 KW |
1854 | *(p)++ = 's'; |
1855 | *p = 's'; | |
1856 | *lenp = 2; | |
1857 | return 's'; | |
1ca267a5 | 1858 | } |
a1dde8de KW |
1859 | } |
1860 | else { /* In this range the fold of all other characters is their lower | |
1861 | case */ | |
1862 | converted = toLOWER_LATIN1(c); | |
1863 | } | |
1864 | ||
6f2d5cbc | 1865 | if (UVCHR_IS_INVARIANT(converted)) { |
a1dde8de KW |
1866 | *p = (U8) converted; |
1867 | *lenp = 1; | |
1868 | } | |
1869 | else { | |
1870 | *(p)++ = UTF8_TWO_BYTE_HI(converted); | |
1871 | *p = UTF8_TWO_BYTE_LO(converted); | |
1872 | *lenp = 2; | |
1873 | } | |
1874 | ||
1875 | return converted; | |
1876 | } | |
1877 | ||
1878 | UV | |
a0270393 | 1879 | Perl__to_uni_fold_flags(pTHX_ UV c, U8* p, STRLEN *lenp, const U8 flags) |
84afefe6 | 1880 | { |
4b593389 | 1881 | |
a0270393 KW |
1882 | /* Not currently externally documented, and subject to change |
1883 | * <flags> bits meanings: | |
1884 | * FOLD_FLAGS_FULL iff full folding is to be used; | |
1885 | * FOLD_FLAGS_LOCALE iff in locale | |
1886 | * FOLD_FLAGS_NOMIX_ASCII iff non-ASCII to ASCII folds are prohibited | |
1887 | */ | |
4b593389 | 1888 | |
36bb2ab6 | 1889 | PERL_ARGS_ASSERT__TO_UNI_FOLD_FLAGS; |
7918f24d | 1890 | |
a1dde8de | 1891 | if (c < 256) { |
a0270393 | 1892 | UV result = _to_fold_latin1((U8) c, p, lenp, |
1ca267a5 | 1893 | flags & (FOLD_FLAGS_FULL | FOLD_FLAGS_NOMIX_ASCII)); |
a0270393 KW |
1894 | /* It is illegal for the fold to cross the 255/256 boundary under |
1895 | * locale; in this case return the original */ | |
1896 | return (result > 256 && flags & FOLD_FLAGS_LOCALE) | |
1897 | ? c | |
1898 | : result; | |
a1dde8de KW |
1899 | } |
1900 | ||
a0270393 KW |
1901 | /* If no special needs, just use the macro */ |
1902 | if ( ! (flags & (FOLD_FLAGS_LOCALE|FOLD_FLAGS_NOMIX_ASCII))) { | |
1903 | uvchr_to_utf8(p, c); | |
1904 | return CALL_FOLD_CASE(p, p, lenp, flags & FOLD_FLAGS_FULL); | |
1905 | } | |
1906 | else { /* Otherwise, _to_utf8_fold_flags has the intelligence to deal with | |
1907 | the special flags. */ | |
1908 | U8 utf8_c[UTF8_MAXBYTES + 1]; | |
1909 | uvchr_to_utf8(utf8_c, c); | |
1910 | return _to_utf8_fold_flags(utf8_c, p, lenp, flags, NULL); | |
1911 | } | |
84afefe6 JH |
1912 | } |
1913 | ||
a0ed51b3 | 1914 | bool |
84afefe6 | 1915 | Perl_is_uni_alnum_lc(pTHX_ UV c) |
a0ed51b3 | 1916 | { |
edfb3318 | 1917 | if (c < 256) { |
f4cd282c | 1918 | return isALNUM_LC(c); |
edfb3318 | 1919 | } |
922e8cb4 | 1920 | return _is_uni_FOO(_CC_WORDCHAR, c); |
a0ed51b3 LW |
1921 | } |
1922 | ||
1923 | bool | |
5092f92a KW |
1924 | Perl_is_uni_alnumc_lc(pTHX_ UV c) |
1925 | { | |
1926 | if (c < 256) { | |
f4cd282c | 1927 | return isALPHANUMERIC_LC(c); |
5092f92a | 1928 | } |
922e8cb4 | 1929 | return _is_uni_FOO(_CC_ALPHANUMERIC, c); |
5092f92a KW |
1930 | } |
1931 | ||
1932 | bool | |
84afefe6 | 1933 | Perl_is_uni_idfirst_lc(pTHX_ UV c) |
a0ed51b3 | 1934 | { |
edfb3318 | 1935 | if (c < 256) { |
f4cd282c | 1936 | return isIDFIRST_LC(c); |
edfb3318 KW |
1937 | } |
1938 | return _is_uni_perl_idstart(c); | |
a0ed51b3 LW |
1939 | } |
1940 | ||
1941 | bool | |
84afefe6 | 1942 | Perl_is_uni_alpha_lc(pTHX_ UV c) |
a0ed51b3 | 1943 | { |
edfb3318 | 1944 | if (c < 256) { |
f4cd282c | 1945 | return isALPHA_LC(c); |
edfb3318 | 1946 | } |
922e8cb4 | 1947 | return _is_uni_FOO(_CC_ALPHA, c); |
a0ed51b3 LW |
1948 | } |
1949 | ||
1950 | bool | |
84afefe6 | 1951 | Perl_is_uni_ascii_lc(pTHX_ UV c) |
4d61ec05 | 1952 | { |
edfb3318 | 1953 | if (c < 256) { |
f4cd282c | 1954 | return isASCII_LC(c); |
edfb3318 KW |
1955 | } |
1956 | return 0; | |
4d61ec05 GS |
1957 | } |
1958 | ||
1959 | bool | |
bdd8600f KW |
1960 | Perl_is_uni_blank_lc(pTHX_ UV c) |
1961 | { | |
edfb3318 | 1962 | if (c < 256) { |
f4cd282c | 1963 | return isBLANK_LC(c); |
edfb3318 | 1964 | } |
61b19385 | 1965 | return isBLANK_uni(c); |
bdd8600f KW |
1966 | } |
1967 | ||
1968 | bool | |
84afefe6 | 1969 | Perl_is_uni_space_lc(pTHX_ UV c) |
a0ed51b3 | 1970 | { |
edfb3318 | 1971 | if (c < 256) { |
f4cd282c | 1972 | return isSPACE_LC(c); |
edfb3318 | 1973 | } |
61b19385 | 1974 | return isSPACE_uni(c); |
a0ed51b3 LW |
1975 | } |
1976 | ||
1977 | bool | |
84afefe6 | 1978 | Perl_is_uni_digit_lc(pTHX_ UV c) |
a0ed51b3 | 1979 | { |
edfb3318 | 1980 | if (c < 256) { |
f4cd282c | 1981 | return isDIGIT_LC(c); |
edfb3318 | 1982 | } |
922e8cb4 | 1983 | return _is_uni_FOO(_CC_DIGIT, c); |
a0ed51b3 LW |
1984 | } |
1985 | ||
1986 | bool | |
84afefe6 | 1987 | Perl_is_uni_upper_lc(pTHX_ UV c) |
a0ed51b3 | 1988 | { |
edfb3318 | 1989 | if (c < 256) { |
f4cd282c | 1990 | return isUPPER_LC(c); |
edfb3318 | 1991 | } |
922e8cb4 | 1992 | return _is_uni_FOO(_CC_UPPER, c); |
a0ed51b3 LW |
1993 | } |
1994 | ||
1995 | bool | |
84afefe6 | 1996 | Perl_is_uni_lower_lc(pTHX_ UV c) |
a0ed51b3 | 1997 | { |
edfb3318 | 1998 | if (c < 256) { |
f4cd282c | 1999 | return isLOWER_LC(c); |
edfb3318 | 2000 | } |
922e8cb4 | 2001 | return _is_uni_FOO(_CC_LOWER, c); |
a0ed51b3 LW |
2002 | } |
2003 | ||
2004 | bool | |
84afefe6 | 2005 | Perl_is_uni_cntrl_lc(pTHX_ UV c) |
b8c5462f | 2006 | { |
edfb3318 | 2007 | if (c < 256) { |
f4cd282c | 2008 | return isCNTRL_LC(c); |
edfb3318 | 2009 | } |
61b19385 | 2010 | return 0; |
b8c5462f JH |
2011 | } |
2012 | ||
2013 | bool | |
84afefe6 | 2014 | Perl_is_uni_graph_lc(pTHX_ UV c) |
b8c5462f | 2015 | { |
edfb3318 | 2016 | if (c < 256) { |
f4cd282c | 2017 | return isGRAPH_LC(c); |
edfb3318 | 2018 | } |
922e8cb4 | 2019 | return _is_uni_FOO(_CC_GRAPH, c); |
b8c5462f JH |
2020 | } |
2021 | ||
2022 | bool | |
84afefe6 | 2023 | Perl_is_uni_print_lc(pTHX_ UV c) |
a0ed51b3 | 2024 | { |
edfb3318 | 2025 | if (c < 256) { |
f4cd282c | 2026 | return isPRINT_LC(c); |
edfb3318 | 2027 | } |
922e8cb4 | 2028 | return _is_uni_FOO(_CC_PRINT, c); |
a0ed51b3 LW |
2029 | } |
2030 | ||
b8c5462f | 2031 | bool |
84afefe6 | 2032 | Perl_is_uni_punct_lc(pTHX_ UV c) |
b8c5462f | 2033 | { |
edfb3318 | 2034 | if (c < 256) { |
f4cd282c | 2035 | return isPUNCT_LC(c); |
edfb3318 | 2036 | } |
922e8cb4 | 2037 | return _is_uni_FOO(_CC_PUNCT, c); |
b8c5462f JH |
2038 | } |
2039 | ||
4d61ec05 | 2040 | bool |
84afefe6 | 2041 | Perl_is_uni_xdigit_lc(pTHX_ UV c) |
4d61ec05 | 2042 | { |
edfb3318 | 2043 | if (c < 256) { |
f4cd282c | 2044 | return isXDIGIT_LC(c); |
edfb3318 | 2045 | } |
61b19385 | 2046 | return isXDIGIT_uni(c); |
4d61ec05 GS |
2047 | } |
2048 | ||
b7ac61fa JH |
2049 | U32 |
2050 | Perl_to_uni_upper_lc(pTHX_ U32 c) | |
2051 | { | |
ee099d14 JH |
2052 | /* XXX returns only the first character -- do not use XXX */ |
2053 | /* XXX no locale support yet */ | |
2054 | STRLEN len; | |
89ebb4a3 | 2055 | U8 tmpbuf[UTF8_MAXBYTES_CASE+1]; |
ee099d14 | 2056 | return (U32)to_uni_upper(c, tmpbuf, &len); |
b7ac61fa JH |
2057 | } |
2058 | ||
2059 | U32 | |
2060 | Perl_to_uni_title_lc(pTHX_ U32 c) | |
2061 | { | |
ee099d14 JH |
2062 | /* XXX returns only the first character XXX -- do not use XXX */ |
2063 | /* XXX no locale support yet */ | |
2064 | STRLEN len; | |
89ebb4a3 | 2065 | U8 tmpbuf[UTF8_MAXBYTES_CASE+1]; |
ee099d14 | 2066 | return (U32)to_uni_title(c, tmpbuf, &len); |
b7ac61fa JH |
2067 | } |
2068 | ||
2069 | U32 | |
2070 | Perl_to_uni_lower_lc(pTHX_ U32 c) | |
2071 | { | |
ee099d14 JH |
2072 | /* XXX returns only the first character -- do not use XXX */ |
2073 | /* XXX no locale support yet */ | |
2074 | STRLEN len; | |
89ebb4a3 | 2075 | U8 tmpbuf[UTF8_MAXBYTES_CASE+1]; |
ee099d14 | 2076 | return (U32)to_uni_lower(c, tmpbuf, &len); |
b7ac61fa JH |
2077 | } |
2078 | ||
26483009 | 2079 | PERL_STATIC_INLINE bool |
5141f98e | 2080 | S_is_utf8_common(pTHX_ const U8 *const p, SV **swash, |
bde6a22d NC |
2081 | const char *const swashname) |
2082 | { | |
ea317ccb KW |
2083 | /* returns a boolean giving whether or not the UTF8-encoded character that |
2084 | * starts at <p> is in the swash indicated by <swashname>. <swash> | |
2085 | * contains a pointer to where the swash indicated by <swashname> | |
2086 | * is to be stored; which this routine will do, so that future calls will | |
2087 | * look at <*swash> and only generate a swash if it is not null | |
2088 | * | |
2089 | * Note that it is assumed that the buffer length of <p> is enough to | |
2090 | * contain all the bytes that comprise the character. Thus, <*p> should | |
2091 | * have been checked before this call for mal-formedness enough to assure | |
2092 | * that. */ | |
2093 | ||
97aff369 | 2094 | dVAR; |
7918f24d NC |
2095 | |
2096 | PERL_ARGS_ASSERT_IS_UTF8_COMMON; | |
2097 | ||
492a624f | 2098 | /* The API should have included a length for the UTF-8 character in <p>, |
28123549 | 2099 | * but it doesn't. We therefore assume that p has been validated at least |
492a624f KW |
2100 | * as far as there being enough bytes available in it to accommodate the |
2101 | * character without reading beyond the end, and pass that number on to the | |
2102 | * validating routine */ | |
28123549 KW |
2103 | if (! is_utf8_char_buf(p, p + UTF8SKIP(p))) { |
2104 | if (ckWARN_d(WARN_UTF8)) { | |
2105 | Perl_warner(aTHX_ packWARN2(WARN_DEPRECATED,WARN_UTF8), | |
9816f121 | 2106 | "Passing malformed UTF-8 to \"%s\" is deprecated", swashname); |
28123549 KW |
2107 | if (ckWARN(WARN_UTF8)) { /* This will output details as to the |
2108 | what the malformation is */ | |
2109 | utf8_to_uvchr_buf(p, p + UTF8SKIP(p), NULL); | |
2110 | } | |
2111 | } | |
2112 | return FALSE; | |
2113 | } | |
87367d5f KW |
2114 | if (!*swash) { |
2115 | U8 flags = _CORE_SWASH_INIT_ACCEPT_INVLIST; | |
2116 | *swash = _core_swash_init("utf8", swashname, &PL_sv_undef, 1, 0, NULL, &flags); | |
2117 | } | |
28123549 | 2118 | |
bde6a22d NC |
2119 | return swash_fetch(*swash, p, TRUE) != 0; |
2120 | } | |
2121 | ||
2122 | bool | |
922e8cb4 KW |
2123 | Perl__is_utf8_FOO(pTHX_ const U8 classnum, const U8 *p) |
2124 | { | |
2125 | dVAR; | |
2126 | ||
2127 | PERL_ARGS_ASSERT__IS_UTF8_FOO; | |
2128 | ||
2129 | assert(classnum < _FIRST_NON_SWASH_CC); | |
2130 | ||
2131 | return is_utf8_common(p, &PL_utf8_swash_ptrs[classnum], swash_property_names[classnum]); | |
2132 | } | |
2133 | ||
2134 | bool | |
7fc63493 | 2135 | Perl_is_utf8_alnum(pTHX_ const U8 *p) |
a0ed51b3 | 2136 | { |
97aff369 | 2137 | dVAR; |
7918f24d NC |
2138 | |
2139 | PERL_ARGS_ASSERT_IS_UTF8_ALNUM; | |
2140 | ||
671c33bf NC |
2141 | /* NOTE: "IsWord", not "IsAlnum", since Alnum is a true |
2142 | * descendant of isalnum(3), in other words, it doesn't | |
2143 | * contain the '_'. --jhi */ | |
03940dc2 | 2144 | return is_utf8_common(p, &PL_utf8_swash_ptrs[_CC_WORDCHAR], "IsWord"); |
a0ed51b3 LW |
2145 | } |
2146 | ||
2147 | bool | |
5092f92a KW |
2148 | Perl_is_utf8_alnumc(pTHX_ const U8 *p) |
2149 | { | |
2150 | dVAR; | |
2151 | ||
2152 | PERL_ARGS_ASSERT_IS_UTF8_ALNUMC; | |
2153 | ||
03940dc2 | 2154 | return is_utf8_common(p, &PL_utf8_swash_ptrs[_CC_ALPHANUMERIC], "IsAlnum"); |
5092f92a KW |
2155 | } |
2156 | ||
2157 | bool | |
7fc63493 | 2158 | Perl_is_utf8_idfirst(pTHX_ const U8 *p) /* The naming is historical. */ |
a0ed51b3 | 2159 | { |
97aff369 | 2160 | dVAR; |
7918f24d NC |
2161 | |
2162 | PERL_ARGS_ASSERT_IS_UTF8_IDFIRST; | |
2163 | ||
61b19385 | 2164 | return S_is_utf8_idfirst(aTHX_ p); |
82686b01 JH |
2165 | } |
2166 | ||
2167 | bool | |
c11ff943 KW |
2168 | Perl_is_utf8_xidfirst(pTHX_ const U8 *p) /* The naming is historical. */ |
2169 | { | |
2170 | dVAR; | |
2171 | ||
2172 | PERL_ARGS_ASSERT_IS_UTF8_XIDFIRST; | |
2173 | ||
2174 | if (*p == '_') | |
2175 | return TRUE; | |
2176 | /* is_utf8_idstart would be more logical. */ | |
2177 | return is_utf8_common(p, &PL_utf8_xidstart, "XIdStart"); | |
2178 | } | |
2179 | ||
2180 | bool | |
d65654cb | 2181 | Perl__is_utf8_perl_idstart(pTHX_ const U8 *p) |
b6912c02 KW |
2182 | { |
2183 | dVAR; | |
2184 | ||
d65654cb | 2185 | PERL_ARGS_ASSERT__IS_UTF8_PERL_IDSTART; |
b6912c02 KW |
2186 | |
2187 | return is_utf8_common(p, &PL_utf8_perl_idstart, "_Perl_IDStart"); | |
2188 | } | |
2189 | ||
2190 | bool | |
eba68aa0 KW |
2191 | Perl__is_utf8_perl_idcont(pTHX_ const U8 *p) |
2192 | { | |
2193 | dVAR; | |
2194 | ||
2195 | PERL_ARGS_ASSERT__IS_UTF8_PERL_IDCONT; | |
2196 | ||
2197 | return is_utf8_common(p, &PL_utf8_perl_idcont, "_Perl_IDCont"); | |
2198 | } | |
2199 | ||
2200 | ||
2201 | bool | |
7fc63493 | 2202 | Perl_is_utf8_idcont(pTHX_ const U8 *p) |
82686b01 | 2203 | { |
97aff369 | 2204 | dVAR; |
7918f24d NC |
2205 | |
2206 | PERL_ARGS_ASSERT_IS_UTF8_IDCONT; | |
2207 | ||
d4c19fe8 | 2208 | return is_utf8_common(p, &PL_utf8_idcont, "IdContinue"); |
a0ed51b3 LW |
2209 | } |
2210 | ||
2211 | bool | |
c11ff943 KW |
2212 | Perl_is_utf8_xidcont(pTHX_ const U8 *p) |
2213 | { | |
2214 | dVAR; | |
2215 | ||
2216 | PERL_ARGS_ASSERT_IS_UTF8_XIDCONT; | |
2217 | ||
c11ff943 KW |
2218 | return is_utf8_common(p, &PL_utf8_idcont, "XIdContinue"); |
2219 | } | |
2220 | ||
2221 | bool | |
7fc63493 | 2222 | Perl_is_utf8_alpha(pTHX_ const U8 *p) |
a0ed51b3 | 2223 | { |
97aff369 | 2224 | dVAR; |
7918f24d NC |
2225 | |
2226 | PERL_ARGS_ASSERT_IS_UTF8_ALPHA; | |
2227 | ||
03940dc2 | 2228 | return is_utf8_common(p, &PL_utf8_swash_ptrs[_CC_ALPHA], "IsAlpha"); |
a0ed51b3 LW |
2229 | } |
2230 | ||
2231 | bool | |
7fc63493 | 2232 | Perl_is_utf8_ascii(pTHX_ const U8 *p) |
b8c5462f | 2233 | { |
97aff369 | 2234 | dVAR; |
7918f24d NC |
2235 | |
2236 | PERL_ARGS_ASSERT_IS_UTF8_ASCII; | |
2237 | ||
bc39fe24 KW |
2238 | /* ASCII characters are the same whether in utf8 or not. So the macro |
2239 | * works on both utf8 and non-utf8 representations. */ | |
2240 | return isASCII(*p); | |
b8c5462f JH |
2241 | } |
2242 | ||
2243 | bool | |
bdd8600f KW |
2244 | Perl_is_utf8_blank(pTHX_ const U8 *p) |
2245 | { | |
2246 | dVAR; | |
2247 | ||
2248 | PERL_ARGS_ASSERT_IS_UTF8_BLANK; | |
2249 | ||
2cafb56b | 2250 | return isBLANK_utf8(p); |
bdd8600f KW |
2251 | } |
2252 | ||
2253 | bool | |
7fc63493 | 2254 | Perl_is_utf8_space(pTHX_ const U8 *p) |
a0ed51b3 | 2255 | { |
97aff369 | 2256 | dVAR; |
7918f24d NC |
2257 | |
2258 | PERL_ARGS_ASSERT_IS_UTF8_SPACE; | |
2259 | ||
add4123a | 2260 | return isSPACE_utf8(p); |
a0ed51b3 LW |
2261 | } |
2262 | ||
2263 | bool | |
d1eb3177 YO |
2264 | Perl_is_utf8_perl_space(pTHX_ const U8 *p) |
2265 | { | |
2266 | dVAR; | |
2267 | ||
2268 | PERL_ARGS_ASSERT_IS_UTF8_PERL_SPACE; | |
2269 | ||
c4428693 KW |
2270 | /* Only true if is an ASCII space-like character, and ASCII is invariant |
2271 | * under utf8, so can just use the macro */ | |
2272 | return isSPACE_A(*p); | |
d1eb3177 YO |
2273 | } |
2274 | ||
2275 | bool | |
2276 | Perl_is_utf8_perl_word(pTHX_ const U8 *p) | |
2277 | { | |
2278 | dVAR; | |
2279 | ||
2280 | PERL_ARGS_ASSERT_IS_UTF8_PERL_WORD; | |
2281 | ||
c4428693 KW |
2282 | /* Only true if is an ASCII word character, and ASCII is invariant |
2283 | * under utf8, so can just use the macro */ | |
2284 | return isWORDCHAR_A(*p); | |
d1eb3177 YO |
2285 | } |
2286 | ||
2287 | bool | |
7fc63493 | 2288 | Perl_is_utf8_digit(pTHX_ const U8 *p) |
a0ed51b3 | 2289 | { |
97aff369 | 2290 | dVAR; |
7918f24d NC |
2291 | |
2292 | PERL_ARGS_ASSERT_IS_UTF8_DIGIT; | |
2293 | ||
03940dc2 | 2294 | return is_utf8_common(p, &PL_utf8_swash_ptrs[_CC_DIGIT], "IsDigit"); |
a0ed51b3 LW |
2295 | } |
2296 | ||
2297 | bool | |
d1eb3177 YO |
2298 | Perl_is_utf8_posix_digit(pTHX_ const U8 *p) |
2299 | { | |
2300 | dVAR; | |
2301 | ||
2302 | PERL_ARGS_ASSERT_IS_UTF8_POSIX_DIGIT; | |
2303 | ||
c4428693 KW |
2304 | /* Only true if is an ASCII digit character, and ASCII is invariant |
2305 | * under utf8, so can just use the macro */ | |
2306 | return isDIGIT_A(*p); | |
d1eb3177 YO |
2307 | } |
2308 | ||
2309 | bool | |
7fc63493 | 2310 | Perl_is_utf8_upper(pTHX_ const U8 *p) |
a0ed51b3 | 2311 | { |
97aff369 | 2312 | dVAR; |
7918f24d NC |
2313 | |
2314 | PERL_ARGS_ASSERT_IS_UTF8_UPPER; | |
2315 | ||
03940dc2 | 2316 | return is_utf8_common(p, &PL_utf8_swash_ptrs[_CC_UPPER], "IsUppercase"); |
a0ed51b3 LW |
2317 | } |
2318 | ||
2319 | bool | |
7fc63493 | 2320 | Perl_is_utf8_lower(pTHX_ const U8 *p) |
a0ed51b3 | 2321 | { |
97aff369 | 2322 | dVAR; |
7918f24d NC |
2323 | |
2324 | PERL_ARGS_ASSERT_IS_UTF8_LOWER; | |
2325 | ||
03940dc2 | 2326 | return is_utf8_common(p, &PL_utf8_swash_ptrs[_CC_LOWER], "IsLowercase"); |
a0ed51b3 LW |
2327 | } |
2328 | ||
2329 | bool | |
7fc63493 | 2330 | Perl_is_utf8_cntrl(pTHX_ const U8 *p) |
b8c5462f | 2331 | { |
97aff369 | 2332 | dVAR; |
7918f24d NC |
2333 | |
2334 | PERL_ARGS_ASSERT_IS_UTF8_CNTRL; | |
2335 | ||
a35d759a | 2336 | return isCNTRL_utf8(p); |
b8c5462f JH |
2337 | } |
2338 | ||
2339 | bool | |
7fc63493 | 2340 | Perl_is_utf8_graph(pTHX_ const U8 *p) |
b8c5462f | 2341 | { |
97aff369 | 2342 | dVAR; |
7918f24d NC |
2343 | |
2344 | PERL_ARGS_ASSERT_IS_UTF8_GRAPH; | |
2345 | ||
03940dc2 | 2346 | return is_utf8_common(p, &PL_utf8_swash_ptrs[_CC_GRAPH], "IsGraph"); |
b8c5462f JH |
2347 | } |
2348 | ||
2349 | bool | |
7fc63493 | 2350 | Perl_is_utf8_print(pTHX_ const U8 *p) |
a0ed51b3 | 2351 | { |
97aff369 | 2352 | dVAR; |
7918f24d NC |
2353 | |
2354 | PERL_ARGS_ASSERT_IS_UTF8_PRINT; | |
2355 | ||
03940dc2 | 2356 | return is_utf8_common(p, &PL_utf8_swash_ptrs[_CC_PRINT], "IsPrint"); |
a0ed51b3 LW |
2357 | } |
2358 | ||
2359 | bool | |
7fc63493 | 2360 | Perl_is_utf8_punct(pTHX_ const U8 *p) |
b8c5462f | 2361 | { |
97aff369 | 2362 | dVAR; |
7918f24d NC |
2363 | |
2364 | PERL_ARGS_ASSERT_IS_UTF8_PUNCT; | |
2365 | ||
03940dc2 | 2366 | return is_utf8_common(p, &PL_utf8_swash_ptrs[_CC_PUNCT], "IsPunct"); |
b8c5462f JH |
2367 | } |
2368 | ||
2369 | bool | |
7fc63493 | 2370 | Perl_is_utf8_xdigit(pTHX_ const U8 *p) |
b8c5462f | 2371 | { |
97aff369 | 2372 | dVAR; |
7918f24d NC |
2373 | |
2374 | PERL_ARGS_ASSERT_IS_UTF8_XDIGIT; | |
2375 | ||
4ac6419d | 2376 | return is_XDIGIT_utf8(p); |
b8c5462f JH |
2377 | } |
2378 | ||
2379 | bool | |
7dbf68d2 KW |
2380 | Perl__is_utf8_mark(pTHX_ const U8 *p) |
2381 | { | |
2382 | dVAR; | |
2383 | ||
2384 | PERL_ARGS_ASSERT__IS_UTF8_MARK; | |
2385 | ||
2386 | return is_utf8_common(p, &PL_utf8_mark, "IsM"); | |
2387 | } | |
2388 | ||
2389 | ||
2390 | bool | |
7fc63493 | 2391 | Perl_is_utf8_mark(pTHX_ const U8 *p) |
a0ed51b3 | 2392 | { |
97aff369 | 2393 | dVAR; |
7918f24d NC |
2394 | |
2395 | PERL_ARGS_ASSERT_IS_UTF8_MARK; | |
2396 | ||
7dbf68d2 | 2397 | return _is_utf8_mark(p); |
a0ed51b3 LW |
2398 | } |
2399 | ||
6b5c0936 | 2400 | /* |
87cea99e | 2401 | =for apidoc to_utf8_case |
6b5c0936 | 2402 | |
a1433954 KW |
2403 | The C<p> contains the pointer to the UTF-8 string encoding |
2404 | the character that is being converted. This routine assumes that the character | |
2405 | at C<p> is well-formed. | |
6b5c0936 | 2406 | |
a1433954 KW |
2407 | The C<ustrp> is a pointer to the character buffer to put the |
2408 | conversion result to. The C<lenp> is a pointer to the length | |
6b5c0936 JH |
2409 | of the result. |
2410 | ||
a1433954 | 2411 | The C<swashp> is a pointer to the swash to use. |
6b5c0936 | 2412 | |
a1433954 KW |
2413 | Both the special and normal mappings are stored in F<lib/unicore/To/Foo.pl>, |
2414 | and loaded by SWASHNEW, using F<lib/utf8_heavy.pl>. The C<special> (usually, | |
0134edef | 2415 | but not always, a multicharacter mapping), is tried first. |
6b5c0936 | 2416 | |
a1433954 | 2417 | The C<special> is a string like "utf8::ToSpecLower", which means the |
0134edef JH |
2418 | hash %utf8::ToSpecLower. The access to the hash is through |
2419 | Perl_to_utf8_case(). | |
6b5c0936 | 2420 | |
a1433954 | 2421 | The C<normal> is a string like "ToLower" which means the swash |
0134edef JH |
2422 | %utf8::ToLower. |
2423 | ||
2424 | =cut */ | |
6b5c0936 | 2425 | |
2104c8d9 | 2426 | UV |
9a957fbc AL |
2427 | Perl_to_utf8_case(pTHX_ const U8 *p, U8* ustrp, STRLEN *lenp, |
2428 | SV **swashp, const char *normal, const char *special) | |
a0ed51b3 | 2429 | { |
97aff369 | 2430 | dVAR; |
0134edef | 2431 | STRLEN len = 0; |
f4cd282c | 2432 | const UV uv1 = valid_utf8_to_uvchr(p, NULL); |
7918f24d NC |
2433 | |
2434 | PERL_ARGS_ASSERT_TO_UTF8_CASE; | |
2435 | ||
9ae3ac1a KW |
2436 | /* Note that swash_fetch() doesn't output warnings for these because it |
2437 | * assumes we will */ | |
8457b38f | 2438 | if (uv1 >= UNICODE_SURROGATE_FIRST) { |
9ae3ac1a | 2439 | if (uv1 <= UNICODE_SURROGATE_LAST) { |
8457b38f KW |
2440 | if (ckWARN_d(WARN_SURROGATE)) { |
2441 | const char* desc = (PL_op) ? OP_DESC(PL_op) : normal; | |
2442 | Perl_warner(aTHX_ packWARN(WARN_SURROGATE), | |
2443 | "Operation \"%s\" returns its argument for UTF-16 surrogate U+%04"UVXf"", desc, uv1); | |
2444 | } | |
9ae3ac1a KW |
2445 | } |
2446 | else if (UNICODE_IS_SUPER(uv1)) { | |
8457b38f KW |
2447 | if (ckWARN_d(WARN_NON_UNICODE)) { |
2448 | const char* desc = (PL_op) ? OP_DESC(PL_op) : normal; | |
2449 | Perl_warner(aTHX_ packWARN(WARN_NON_UNICODE), | |
2450 | "Operation \"%s\" returns its argument for non-Unicode code point 0x%04"UVXf"", desc, uv1); | |
2451 | } | |
9ae3ac1a KW |
2452 | } |
2453 | ||
2454 | /* Note that non-characters are perfectly legal, so no warning should | |
2455 | * be given */ | |
2456 | } | |
2457 | ||
0134edef | 2458 | if (!*swashp) /* load on-demand */ |
5ab9d2ef | 2459 | *swashp = _core_swash_init("utf8", normal, &PL_sv_undef, 4, 0, NULL, NULL); |
0134edef | 2460 | |
a6f87d8c | 2461 | if (special) { |
0134edef | 2462 | /* It might be "special" (sometimes, but not always, |
2a37f04d | 2463 | * a multicharacter mapping) */ |
6673a63c | 2464 | HV * const hv = get_hv(special, 0); |
b08cf34e JH |
2465 | SV **svp; |
2466 | ||
35da51f7 | 2467 | if (hv && |
f4cd282c | 2468 | (svp = hv_fetch(hv, (const char*)p, UNISKIP(uv1), FALSE)) && |
b08cf34e | 2469 | (*svp)) { |
cfd0369c | 2470 | const char *s; |
47654450 | 2471 | |
cfd0369c | 2472 | s = SvPV_const(*svp, len); |
47654450 | 2473 | if (len == 1) |
f4cd282c | 2474 | /* EIGHTBIT */ |
c80e42f3 | 2475 | len = uvchr_to_utf8(ustrp, *(U8*)s) - ustrp; |
2a37f04d | 2476 | else { |
d2dcd0fb | 2477 | Copy(s, ustrp, len, U8); |
29e98929 | 2478 | } |
983ffd37 | 2479 | } |
0134edef JH |
2480 | } |
2481 | ||
2482 | if (!len && *swashp) { | |
f4cd282c | 2483 | const UV uv2 = swash_fetch(*swashp, p, TRUE /* => is utf8 */); |
d4c19fe8 | 2484 | |
0134edef JH |
2485 | if (uv2) { |
2486 | /* It was "normal" (a single character mapping). */ | |
f4cd282c | 2487 | len = uvchr_to_utf8(ustrp, uv2) - ustrp; |
2a37f04d JH |
2488 | } |
2489 | } | |
1feea2c7 | 2490 | |
cbe07460 KW |
2491 | if (len) { |
2492 | if (lenp) { | |
2493 | *lenp = len; | |
2494 | } | |
2495 | return valid_utf8_to_uvchr(ustrp, 0); | |
2496 | } | |
2497 | ||
2498 | /* Here, there was no mapping defined, which means that the code point maps | |
2499 | * to itself. Return the inputs */ | |
bfdf22ec | 2500 | len = UTF8SKIP(p); |
ca9fab46 KW |
2501 | if (p != ustrp) { /* Don't copy onto itself */ |
2502 | Copy(p, ustrp, len, U8); | |
2503 | } | |
0134edef | 2504 | |
2a37f04d JH |
2505 | if (lenp) |
2506 | *lenp = len; | |
2507 | ||
f4cd282c | 2508 | return uv1; |
cbe07460 | 2509 | |
a0ed51b3 LW |
2510 | } |
2511 | ||
051a06d4 KW |
2512 | STATIC UV |
2513 | S_check_locale_boundary_crossing(pTHX_ const U8* const p, const UV result, U8* const ustrp, STRLEN *lenp) | |
2514 | { | |
2515 | /* This is called when changing the case of a utf8-encoded character above | |
2516 | * the Latin1 range, and the operation is in locale. If the result | |
2517 | * contains a character that crosses the 255/256 boundary, disallow the | |
2518 | * change, and return the original code point. See L<perlfunc/lc> for why; | |
2519 | * | |
a1433954 KW |
2520 | * p points to the original string whose case was changed; assumed |
2521 | * by this routine to be well-formed | |
051a06d4 KW |
2522 | * result the code point of the first character in the changed-case string |
2523 | * ustrp points to the changed-case string (<result> represents its first char) | |
2524 | * lenp points to the length of <ustrp> */ | |
2525 | ||
2526 | UV original; /* To store the first code point of <p> */ | |
2527 | ||
2528 | PERL_ARGS_ASSERT_CHECK_LOCALE_BOUNDARY_CROSSING; | |
2529 | ||
a4f12ed7 | 2530 | assert(UTF8_IS_ABOVE_LATIN1(*p)); |
051a06d4 KW |
2531 | |
2532 | /* We know immediately if the first character in the string crosses the | |
2533 | * boundary, so can skip */ | |
2534 | if (result > 255) { | |
2535 | ||
2536 | /* Look at every character in the result; if any cross the | |
2537 | * boundary, the whole thing is disallowed */ | |
2538 | U8* s = ustrp + UTF8SKIP(ustrp); | |
2539 | U8* e = ustrp + *lenp; | |
2540 | while (s < e) { | |
a4f12ed7 | 2541 | if (! UTF8_IS_ABOVE_LATIN1(*s)) { |
051a06d4 KW |
2542 | goto bad_crossing; |
2543 | } | |
2544 | s += UTF8SKIP(s); | |
2545 | } | |
2546 | ||
2547 | /* Here, no characters crossed, result is ok as-is */ | |
2548 | return result; | |
2549 | } | |
2550 | ||
2551 | bad_crossing: | |
2552 | ||
2553 | /* Failed, have to return the original */ | |
4b88fb76 | 2554 | original = valid_utf8_to_uvchr(p, lenp); |
051a06d4 KW |
2555 | Copy(p, ustrp, *lenp, char); |
2556 | return original; | |
2557 | } | |
2558 | ||
d3e79532 | 2559 | /* |
87cea99e | 2560 | =for apidoc to_utf8_upper |
d3e79532 | 2561 | |
1f607577 | 2562 | Instead use L</toUPPER_utf8>. |
a1433954 | 2563 | |
d3e79532 JH |
2564 | =cut */ |
2565 | ||
051a06d4 KW |
2566 | /* Not currently externally documented, and subject to change: |
2567 | * <flags> is set iff locale semantics are to be used for code points < 256 | |
2568 | * <tainted_ptr> if non-null, *tainted_ptr will be set TRUE iff locale rules | |
2569 | * were used in the calculation; otherwise unchanged. */ | |
2570 | ||
2104c8d9 | 2571 | UV |
051a06d4 | 2572 | Perl__to_utf8_upper_flags(pTHX_ const U8 *p, U8* ustrp, STRLEN *lenp, const bool flags, bool* tainted_ptr) |
a0ed51b3 | 2573 | { |
97aff369 | 2574 | dVAR; |
7918f24d | 2575 | |
051a06d4 KW |
2576 | UV result; |
2577 | ||
2578 | PERL_ARGS_ASSERT__TO_UTF8_UPPER_FLAGS; | |
7918f24d | 2579 | |
3a4c58c9 | 2580 | if (UTF8_IS_INVARIANT(*p)) { |
051a06d4 KW |
2581 | if (flags) { |
2582 | result = toUPPER_LC(*p); | |
2583 | } | |
2584 | else { | |
81c6c7ce | 2585 | return _to_upper_title_latin1(*p, ustrp, lenp, 'S'); |
051a06d4 | 2586 | } |
3a4c58c9 KW |
2587 | } |
2588 | else if UTF8_IS_DOWNGRADEABLE_START(*p) { | |
051a06d4 | 2589 | if (flags) { |
94bb8c36 | 2590 | result = toUPPER_LC(TWO_BYTE_UTF8_TO_NATIVE(*p, *(p+1))); |
051a06d4 KW |
2591 | } |
2592 | else { | |
94bb8c36 | 2593 | return _to_upper_title_latin1(TWO_BYTE_UTF8_TO_NATIVE(*p, *(p+1)), |
81c6c7ce | 2594 | ustrp, lenp, 'S'); |
051a06d4 KW |
2595 | } |
2596 | } | |
2597 | else { /* utf8, ord above 255 */ | |
2598 | result = CALL_UPPER_CASE(p, ustrp, lenp); | |
2599 | ||
2600 | if (flags) { | |
2601 | result = check_locale_boundary_crossing(p, result, ustrp, lenp); | |
2602 | } | |
2603 | return result; | |
2604 | } | |
2605 | ||
2606 | /* Here, used locale rules. Convert back to utf8 */ | |
2607 | if (UTF8_IS_INVARIANT(result)) { | |
2608 | *ustrp = (U8) result; | |
2609 | *lenp = 1; | |
2610 | } | |
2611 | else { | |
62cb07ea KW |
2612 | *ustrp = UTF8_EIGHT_BIT_HI((U8) result); |
2613 | *(ustrp + 1) = UTF8_EIGHT_BIT_LO((U8) result); | |
051a06d4 | 2614 | *lenp = 2; |
3a4c58c9 KW |
2615 | } |
2616 | ||
051a06d4 KW |
2617 | if (tainted_ptr) { |
2618 | *tainted_ptr = TRUE; | |
2619 | } | |
2620 | return result; | |
983ffd37 | 2621 | } |
a0ed51b3 | 2622 | |
d3e79532 | 2623 | /* |
87cea99e | 2624 | =for apidoc to_utf8_title |
d3e79532 | 2625 | |
1f607577 | 2626 | Instead use L</toTITLE_utf8>. |
a1433954 | 2627 | |
d3e79532 JH |
2628 | =cut */ |
2629 | ||
051a06d4 KW |
2630 | /* Not currently externally documented, and subject to change: |
2631 | * <flags> is set iff locale semantics are to be used for code points < 256 | |
2632 | * Since titlecase is not defined in POSIX, uppercase is used instead | |
2633 | * for these/ | |
2634 | * <tainted_ptr> if non-null, *tainted_ptr will be set TRUE iff locale rules | |
2635 | * were used in the calculation; otherwise unchanged. */ | |
2636 | ||
983ffd37 | 2637 | UV |
051a06d4 | 2638 | Perl__to_utf8_title_flags(pTHX_ const U8 *p, U8* ustrp, STRLEN *lenp, const bool flags, bool* tainted_ptr) |
983ffd37 | 2639 | { |
97aff369 | 2640 | dVAR; |
7918f24d | 2641 | |
051a06d4 KW |
2642 | UV result; |
2643 | ||
2644 | PERL_ARGS_ASSERT__TO_UTF8_TITLE_FLAGS; | |
7918f24d | 2645 | |
3a4c58c9 | 2646 | if (UTF8_IS_INVARIANT(*p)) { |
051a06d4 KW |
2647 | if (flags) { |
2648 | result = toUPPER_LC(*p); | |
2649 | } | |
2650 | else { | |
81c6c7ce | 2651 | return _to_upper_title_latin1(*p, ustrp, lenp, 's'); |
051a06d4 | 2652 | } |
3a4c58c9 KW |
2653 | } |
2654 | else if UTF8_IS_DOWNGRADEABLE_START(*p) { | |
051a06d4 | 2655 | if (flags) { |
94bb8c36 | 2656 | result = toUPPER_LC(TWO_BYTE_UTF8_TO_NATIVE(*p, *(p+1))); |
051a06d4 KW |
2657 | } |
2658 | else { | |
94bb8c36 | 2659 | return _to_upper_title_latin1(TWO_BYTE_UTF8_TO_NATIVE(*p, *(p+1)), |
81c6c7ce | 2660 | ustrp, lenp, 's'); |
051a06d4 KW |
2661 | } |
2662 | } | |
2663 | else { /* utf8, ord above 255 */ | |
2664 | result = CALL_TITLE_CASE(p, ustrp, lenp); | |
2665 | ||
2666 | if (flags) { | |
2667 | result = check_locale_boundary_crossing(p, result, ustrp, lenp); | |
2668 | } | |
2669 | return result; | |
2670 | } | |
2671 | ||
2672 | /* Here, used locale rules. Convert back to utf8 */ | |
2673 | if (UTF8_IS_INVARIANT(result)) { | |
2674 | *ustrp = (U8) result; | |
2675 | *lenp = 1; | |
2676 | } | |
2677 | else { | |
62cb07ea KW |
2678 | *ustrp = UTF8_EIGHT_BIT_HI((U8) result); |
2679 | *(ustrp + 1) = UTF8_EIGHT_BIT_LO((U8) result); | |
051a06d4 | 2680 | *lenp = 2; |
3a4c58c9 KW |
2681 | } |
2682 | ||
051a06d4 KW |
2683 | if (tainted_ptr) { |
2684 | *tainted_ptr = TRUE; | |
2685 | } | |
2686 | return result; | |
a0ed51b3 LW |
2687 | } |
2688 | ||
d3e79532 | 2689 | /* |
87cea99e | 2690 | =for apidoc to_utf8_lower |
d3e79532 | 2691 | |
1f607577 | 2692 | Instead use L</toLOWER_utf8>. |
a1433954 | 2693 | |
d3e79532 JH |
2694 | =cut */ |
2695 | ||
051a06d4 KW |
2696 | /* Not currently externally documented, and subject to change: |
2697 | * <flags> is set iff locale semantics are to be used for code points < 256 | |
2698 | * <tainted_ptr> if non-null, *tainted_ptr will be set TRUE iff locale rules | |
2699 | * were used in the calculation; otherwise unchanged. */ | |
2700 | ||
2104c8d9 | 2701 | UV |
051a06d4 | 2702 | Perl__to_utf8_lower_flags(pTHX_ const U8 *p, U8* ustrp, STRLEN *lenp, const bool flags, bool* tainted_ptr) |
a0ed51b3 | 2703 | { |
051a06d4 KW |
2704 | UV result; |
2705 | ||
97aff369 | 2706 | dVAR; |
7918f24d | 2707 | |
051a06d4 | 2708 | PERL_ARGS_ASSERT__TO_UTF8_LOWER_FLAGS; |
7918f24d | 2709 | |
968c5e6a | 2710 | if (UTF8_IS_INVARIANT(*p)) { |
051a06d4 KW |
2711 | if (flags) { |
2712 | result = toLOWER_LC(*p); | |
2713 | } | |
2714 | else { | |
81c6c7ce | 2715 | return to_lower_latin1(*p, ustrp, lenp); |
051a06d4 | 2716 | } |
968c5e6a KW |
2717 | } |
2718 | else if UTF8_IS_DOWNGRADEABLE_START(*p) { | |
051a06d4 | 2719 | if (flags) { |
94bb8c36 | 2720 | result = toLOWER_LC(TWO_BYTE_UTF8_TO_NATIVE(*p, *(p+1))); |
051a06d4 KW |
2721 | } |
2722 | else { | |
94bb8c36 | 2723 | return to_lower_latin1(TWO_BYTE_UTF8_TO_NATIVE(*p, *(p+1)), |
81c6c7ce | 2724 | ustrp, lenp); |
051a06d4 | 2725 | } |
968c5e6a | 2726 | } |
051a06d4 KW |
2727 | else { /* utf8, ord above 255 */ |
2728 | result = CALL_LOWER_CASE(p, ustrp, lenp); | |
2729 | ||
2730 | if (flags) { | |
2731 | result = check_locale_boundary_crossing(p, result, ustrp, lenp); | |
2732 | } | |
968c5e6a | 2733 | |
051a06d4 KW |
2734 | return result; |
2735 | } | |
2736 | ||
2737 | /* Here, used locale rules. Convert back to utf8 */ | |
2738 | if (UTF8_IS_INVARIANT(result)) { | |
2739 | *ustrp = (U8) result; | |
2740 | *lenp = 1; | |
2741 | } | |
2742 | else { | |
62cb07ea KW |
2743 | *ustrp = UTF8_EIGHT_BIT_HI((U8) result); |
2744 | *(ustrp + 1) = UTF8_EIGHT_BIT_LO((U8) result); | |
051a06d4 KW |
2745 | *lenp = 2; |
2746 | } | |
2747 | ||
2748 | if (tainted_ptr) { | |
2749 | *tainted_ptr = TRUE; | |
2750 | } | |
2751 | return result; | |
b4e400f9 JH |
2752 | } |
2753 | ||
d3e79532 | 2754 | /* |
87cea99e | 2755 | =for apidoc to_utf8_fold |
d3e79532 | 2756 | |
1f607577 | 2757 | Instead use L</toFOLD_utf8>. |
a1433954 | 2758 | |
d3e79532 JH |
2759 | =cut */ |
2760 | ||
051a06d4 KW |
2761 | /* Not currently externally documented, and subject to change, |
2762 | * in <flags> | |
2763 | * bit FOLD_FLAGS_LOCALE is set iff locale semantics are to be used for code | |
2764 | * points < 256. Since foldcase is not defined in | |
2765 | * POSIX, lowercase is used instead | |
2766 | * bit FOLD_FLAGS_FULL is set iff full case folds are to be used; | |
2767 | * otherwise simple folds | |
a0270393 KW |
2768 | * bit FOLD_FLAGS_NOMIX_ASCII is set iff folds of non-ASCII to ASCII are |
2769 | * prohibited | |
051a06d4 KW |
2770 | * <tainted_ptr> if non-null, *tainted_ptr will be set TRUE iff locale rules |
2771 | * were used in the calculation; otherwise unchanged. */ | |
36bb2ab6 | 2772 | |
b4e400f9 | 2773 | UV |
051a06d4 | 2774 | Perl__to_utf8_fold_flags(pTHX_ const U8 *p, U8* ustrp, STRLEN *lenp, U8 flags, bool* tainted_ptr) |
b4e400f9 | 2775 | { |
97aff369 | 2776 | dVAR; |
7918f24d | 2777 | |
051a06d4 KW |
2778 | UV result; |
2779 | ||
36bb2ab6 | 2780 | PERL_ARGS_ASSERT__TO_UTF8_FOLD_FLAGS; |
7918f24d | 2781 | |
a0270393 KW |
2782 | /* These are mutually exclusive */ |
2783 | assert (! ((flags & FOLD_FLAGS_LOCALE) && (flags & FOLD_FLAGS_NOMIX_ASCII))); | |
2784 | ||
50ba90ff KW |
2785 | assert(p != ustrp); /* Otherwise overwrites */ |
2786 | ||
a1dde8de | 2787 | if (UTF8_IS_INVARIANT(*p)) { |
051a06d4 | 2788 | if (flags & FOLD_FLAGS_LOCALE) { |
d22b930b | 2789 | result = toFOLD_LC(*p); |
051a06d4 KW |
2790 | } |
2791 | else { | |
81c6c7ce | 2792 | return _to_fold_latin1(*p, ustrp, lenp, |
1ca267a5 | 2793 | flags & (FOLD_FLAGS_FULL | FOLD_FLAGS_NOMIX_ASCII)); |
051a06d4 | 2794 | } |
a1dde8de KW |
2795 | } |
2796 | else if UTF8_IS_DOWNGRADEABLE_START(*p) { | |
051a06d4 | 2797 | if (flags & FOLD_FLAGS_LOCALE) { |
94bb8c36 | 2798 | result = toFOLD_LC(TWO_BYTE_UTF8_TO_NATIVE(*p, *(p+1))); |
051a06d4 KW |
2799 | } |
2800 | else { | |
94bb8c36 | 2801 | return _to_fold_latin1(TWO_BYTE_UTF8_TO_NATIVE(*p, *(p+1)), |
51910141 | 2802 | ustrp, lenp, |
1ca267a5 | 2803 | flags & (FOLD_FLAGS_FULL | FOLD_FLAGS_NOMIX_ASCII)); |
051a06d4 | 2804 | } |
a1dde8de | 2805 | } |
051a06d4 | 2806 | else { /* utf8, ord above 255 */ |
a0270393 | 2807 | result = CALL_FOLD_CASE(p, ustrp, lenp, flags & FOLD_FLAGS_FULL); |
a1dde8de | 2808 | |
1ca267a5 KW |
2809 | if (flags & FOLD_FLAGS_LOCALE) { |
2810 | ||
9fc2026f | 2811 | /* Special case these characters, as what normally gets returned |
1ca267a5 KW |
2812 | * under locale doesn't work */ |
2813 | if (UTF8SKIP(p) == sizeof(LATIN_CAPITAL_LETTER_SHARP_S_UTF8) - 1 | |
2814 | && memEQ((char *) p, LATIN_CAPITAL_LETTER_SHARP_S_UTF8, | |
2815 | sizeof(LATIN_CAPITAL_LETTER_SHARP_S_UTF8) - 1)) | |
2816 | { | |
2817 | goto return_long_s; | |
2818 | } | |
9fc2026f KW |
2819 | else if (UTF8SKIP(p) == sizeof(LATIN_SMALL_LIGATURE_LONG_S_T) - 1 |
2820 | && memEQ((char *) p, LATIN_SMALL_LIGATURE_LONG_S_T_UTF8, | |
2821 | sizeof(LATIN_SMALL_LIGATURE_LONG_S_T_UTF8) - 1)) | |
2822 | { | |
2823 | goto return_ligature_st; | |
2824 | } | |
a0270393 | 2825 | return check_locale_boundary_crossing(p, result, ustrp, lenp); |
051a06d4 | 2826 | } |
a0270393 KW |
2827 | else if (! (flags & FOLD_FLAGS_NOMIX_ASCII)) { |
2828 | return result; | |
2829 | } | |
2830 | else { | |
2831 | /* This is called when changing the case of a utf8-encoded | |
9fc2026f KW |
2832 | * character above the ASCII range, and the result should not |
2833 | * contain an ASCII character. */ | |
a0270393 KW |
2834 | |
2835 | UV original; /* To store the first code point of <p> */ | |
2836 | ||
2837 | /* Look at every character in the result; if any cross the | |
2838 | * boundary, the whole thing is disallowed */ | |
2839 | U8* s = ustrp; | |
2840 | U8* e = ustrp + *lenp; | |
2841 | while (s < e) { | |
2842 | if (isASCII(*s)) { | |
2843 | /* Crossed, have to return the original */ | |
2844 | original = valid_utf8_to_uvchr(p, lenp); | |
1ca267a5 | 2845 | |
9fc2026f | 2846 | /* But in these instances, there is an alternative we can |
1ca267a5 | 2847 | * return that is valid */ |
9fc2026f KW |
2848 | if (original == LATIN_CAPITAL_LETTER_SHARP_S |
2849 | || original == LATIN_SMALL_LETTER_SHARP_S) | |
2850 | { | |
1ca267a5 KW |
2851 | goto return_long_s; |
2852 | } | |
9fc2026f KW |
2853 | else if (original == LATIN_SMALL_LIGATURE_LONG_S_T) { |
2854 | goto return_ligature_st; | |
2855 | } | |
a0270393 KW |
2856 | Copy(p, ustrp, *lenp, char); |
2857 | return original; | |
2858 | } | |
2859 | s += UTF8SKIP(s); | |
2860 | } | |
051a06d4 | 2861 | |
a0270393 KW |
2862 | /* Here, no characters crossed, result is ok as-is */ |
2863 | return result; | |
2864 | } | |
051a06d4 KW |
2865 | } |
2866 | ||
2867 | /* Here, used locale rules. Convert back to utf8 */ | |
2868 | if (UTF8_IS_INVARIANT(result)) { | |
2869 | *ustrp = (U8) result; | |
2870 | *lenp = 1; | |
2871 | } | |
2872 | else { | |
62cb07ea KW |
2873 | *ustrp = UTF8_EIGHT_BIT_HI((U8) result); |
2874 | *(ustrp + 1) = UTF8_EIGHT_BIT_LO((U8) result); | |
051a06d4 KW |
2875 | *lenp = 2; |
2876 | } | |
2877 | ||
2878 | if (tainted_ptr) { | |
2879 | *tainted_ptr = TRUE; | |
2880 | } | |
2881 | return result; | |
1ca267a5 KW |
2882 | |
2883 | return_long_s: | |
2884 | /* Certain folds to 'ss' are prohibited by the options, but they do allow | |
2885 | * folds to a string of two of these characters. By returning this | |
2886 | * instead, then, e.g., | |
2887 | * fc("\x{1E9E}") eq fc("\x{17F}\x{17F}") | |
2888 | * works. */ | |
2889 | ||
2890 | *lenp = 2 * sizeof(LATIN_SMALL_LETTER_LONG_S_UTF8) - 2; | |
2891 | Copy(LATIN_SMALL_LETTER_LONG_S_UTF8 LATIN_SMALL_LETTER_LONG_S_UTF8, | |
2892 | ustrp, *lenp, U8); | |
2893 | return LATIN_SMALL_LETTER_LONG_S; | |
9fc2026f KW |
2894 | |
2895 | return_ligature_st: | |
2896 | /* Two folds to 'st' are prohibited by the options; instead we pick one and | |
2897 | * have the other one fold to it */ | |
2898 | ||
2899 | *lenp = sizeof(LATIN_SMALL_LIGATURE_ST_UTF8) - 1; | |
2900 | Copy(LATIN_SMALL_LIGATURE_ST_UTF8, ustrp, *lenp, U8); | |
2901 | return LATIN_SMALL_LIGATURE_ST; | |
a0ed51b3 LW |
2902 | } |
2903 | ||
711a919c | 2904 | /* Note: |
f90a9a02 | 2905 | * Returns a "swash" which is a hash described in utf8.c:Perl_swash_fetch(). |
711a919c TS |
2906 | * C<pkg> is a pointer to a package name for SWASHNEW, should be "utf8". |
2907 | * For other parameters, see utf8::SWASHNEW in lib/utf8_heavy.pl. | |
2908 | */ | |
c4a5db0c | 2909 | |
a0ed51b3 | 2910 | SV* |
7fc63493 | 2911 | Perl_swash_init(pTHX_ const char* pkg, const char* name, SV *listsv, I32 minbits, I32 none) |
a0ed51b3 | 2912 | { |
c4a5db0c KW |
2913 | PERL_ARGS_ASSERT_SWASH_INIT; |
2914 | ||
2915 | /* Returns a copy of a swash initiated by the called function. This is the | |
2916 | * public interface, and returning a copy prevents others from doing | |
2917 | * mischief on the original */ | |
2918 | ||
5d3d13d1 | 2919 | return newSVsv(_core_swash_init(pkg, name, listsv, minbits, none, NULL, NULL)); |
c4a5db0c KW |
2920 | } |
2921 | ||
2922 | SV* | |
5d3d13d1 | 2923 | Perl__core_swash_init(pTHX_ const char* pkg, const char* name, SV *listsv, I32 minbits, I32 none, SV* invlist, U8* const flags_p) |
c4a5db0c KW |
2924 | { |
2925 | /* Initialize and return a swash, creating it if necessary. It does this | |
87367d5f KW |
2926 | * by calling utf8_heavy.pl in the general case. The returned value may be |
2927 | * the swash's inversion list instead if the input parameters allow it. | |
2928 | * Which is returned should be immaterial to callers, as the only | |
923b6d4e KW |
2929 | * operations permitted on a swash, swash_fetch(), _get_swash_invlist(), |
2930 | * and swash_to_invlist() handle both these transparently. | |
c4a5db0c KW |
2931 | * |
2932 | * This interface should only be used by functions that won't destroy or | |
2933 | * adversely change the swash, as doing so affects all other uses of the | |
2934 | * swash in the program; the general public should use 'Perl_swash_init' | |
2935 | * instead. | |
2936 | * | |
2937 | * pkg is the name of the package that <name> should be in. | |
2938 | * name is the name of the swash to find. Typically it is a Unicode | |
2939 | * property name, including user-defined ones | |
2940 | * listsv is a string to initialize the swash with. It must be of the form | |
2941 | * documented as the subroutine return value in | |
2942 | * L<perlunicode/User-Defined Character Properties> | |
2943 | * minbits is the number of bits required to represent each data element. | |
2944 | * It is '1' for binary properties. | |
2945 | * none I (khw) do not understand this one, but it is used only in tr///. | |
9a53f6cf | 2946 | * invlist is an inversion list to initialize the swash with (or NULL) |
83199d38 KW |
2947 | * flags_p if non-NULL is the address of various input and output flag bits |
2948 | * to the routine, as follows: ('I' means is input to the routine; | |
2949 | * 'O' means output from the routine. Only flags marked O are | |
2950 | * meaningful on return.) | |
2951 | * _CORE_SWASH_INIT_USER_DEFINED_PROPERTY indicates if the swash | |
2952 | * came from a user-defined property. (I O) | |
5d3d13d1 KW |
2953 | * _CORE_SWASH_INIT_RETURN_IF_UNDEF indicates that instead of croaking |
2954 | * when the swash cannot be located, to simply return NULL. (I) | |
87367d5f KW |
2955 | * _CORE_SWASH_INIT_ACCEPT_INVLIST indicates that the caller will accept a |
2956 | * return of an inversion list instead of a swash hash if this routine | |
2957 | * thinks that would result in faster execution of swash_fetch() later | |
2958 | * on. (I) | |
9a53f6cf KW |
2959 | * |
2960 | * Thus there are three possible inputs to find the swash: <name>, | |
2961 | * <listsv>, and <invlist>. At least one must be specified. The result | |
2962 | * will be the union of the specified ones, although <listsv>'s various | |
2963 | * actions can intersect, etc. what <name> gives. | |
2964 | * | |
2965 | * <invlist> is only valid for binary properties */ | |
c4a5db0c | 2966 | |
27da23d5 | 2967 | dVAR; |
c4a5db0c | 2968 | SV* retval = &PL_sv_undef; |
83199d38 | 2969 | HV* swash_hv = NULL; |
87367d5f KW |
2970 | const int invlist_swash_boundary = |
2971 | (flags_p && *flags_p & _CORE_SWASH_INIT_ACCEPT_INVLIST) | |
2972 | ? 512 /* Based on some benchmarking, but not extensive, see commit | |
2973 | message */ | |
2974 | : -1; /* Never return just an inversion list */ | |
9a53f6cf KW |
2975 | |
2976 | assert(listsv != &PL_sv_undef || strNE(name, "") || invlist); | |
2977 | assert(! invlist || minbits == 1); | |
2978 | ||
2979 | /* If data was passed in to go out to utf8_heavy to find the swash of, do | |
2980 | * so */ | |
2981 | if (listsv != &PL_sv_undef || strNE(name, "")) { | |
69794297 KW |
2982 | dSP; |
2983 | const size_t pkg_len = strlen(pkg); | |
2984 | const size_t name_len = strlen(name); | |
2985 | HV * const stash = gv_stashpvn(pkg, pkg_len, 0); | |
2986 | SV* errsv_save; | |
2987 | GV *method; | |
2988 | ||
2989 | PERL_ARGS_ASSERT__CORE_SWASH_INIT; | |
2990 | ||
2991 | PUSHSTACKi(PERLSI_MAGIC); | |
ce3b816e | 2992 | ENTER; |
69794297 KW |
2993 | SAVEHINTS(); |
2994 | save_re_context(); | |
650f067c JL |
2995 | /* We might get here via a subroutine signature which uses a utf8 |
2996 | * parameter name, at which point PL_subname will have been set | |
2997 | * but not yet used. */ | |
2998 | save_item(PL_subname); | |
69794297 KW |
2999 | if (PL_parser && PL_parser->error_count) |
3000 | SAVEI8(PL_parser->error_count), PL_parser->error_count = 0; | |
3001 | method = gv_fetchmeth(stash, "SWASHNEW", 8, -1); | |
3002 | if (!method) { /* demand load utf8 */ | |
3003 | ENTER; | |
db2c6cb3 FC |
3004 | if ((errsv_save = GvSV(PL_errgv))) SAVEFREESV(errsv_save); |
3005 | GvSV(PL_errgv) = NULL; | |
69794297 KW |
3006 | /* It is assumed that callers of this routine are not passing in |
3007 | * any user derived data. */ | |
3008 | /* Need to do this after save_re_context() as it will set | |
3009 | * PL_tainted to 1 while saving $1 etc (see the code after getrx: | |
3010 | * in Perl_magic_get). Even line to create errsv_save can turn on | |
3011 | * PL_tainted. */ | |
284167a5 S |
3012 | #ifndef NO_TAINT_SUPPORT |
3013 | SAVEBOOL(TAINT_get); | |
3014 | TAINT_NOT; | |
3015 | #endif | |
69794297 KW |
3016 | Perl_load_module(aTHX_ PERL_LOADMOD_NOIMPORT, newSVpvn(pkg,pkg_len), |
3017 | NULL); | |
eed484f9 | 3018 | { |
db2c6cb3 FC |
3019 | /* Not ERRSV, as there is no need to vivify a scalar we are |
3020 | about to discard. */ | |
3021 | SV * const errsv = GvSV(PL_errgv); | |
3022 | if (!SvTRUE(errsv)) { | |
3023 | GvSV(PL_errgv) = SvREFCNT_inc_simple(errsv_save); | |
3024 | SvREFCNT_dec(errsv); | |
3025 | } | |
eed484f9 | 3026 | } |
69794297 KW |
3027 | LEAVE; |
3028 | } | |
3029 | SPAGAIN; | |
3030 | PUSHMARK(SP); | |
3031 | EXTEND(SP,5); | |
3032 | mPUSHp(pkg, pkg_len); | |
3033 | mPUSHp(name, name_len); | |
3034 | PUSHs(listsv); | |
3035 | mPUSHi(minbits); | |
3036 | mPUSHi(none); | |
3037 | PUTBACK; | |
db2c6cb3 FC |
3038 | if ((errsv_save = GvSV(PL_errgv))) SAVEFREESV(errsv_save); |
3039 | GvSV(PL_errgv) = NULL; | |
69794297 KW |
3040 | /* If we already have a pointer to the method, no need to use |
3041 | * call_method() to repeat the lookup. */ | |
c41800a8 KW |
3042 | if (method |
3043 | ? call_sv(MUTABLE_SV(method), G_SCALAR) | |
69794297 KW |
3044 | : call_sv(newSVpvs_flags("SWASHNEW", SVs_TEMP), G_SCALAR | G_METHOD)) |
3045 | { | |
3046 | retval = *PL_stack_sp--; | |
3047 | SvREFCNT_inc(retval); | |
3048 | } | |
eed484f9 | 3049 | { |
db2c6cb3 FC |
3050 | /* Not ERRSV. See above. */ |
3051 | SV * const errsv = GvSV(PL_errgv); | |
3052 | if (!SvTRUE(errsv)) { | |
3053 | GvSV(PL_errgv) = SvREFCNT_inc_simple(errsv_save); | |
3054 | SvREFCNT_dec(errsv); | |
3055 | } | |
eed484f9 | 3056 | } |
ce3b816e | 3057 | LEAVE; |
69794297 KW |
3058 | POPSTACK; |
3059 | if (IN_PERL_COMPILETIME) { | |
3060 | CopHINTS_set(PL_curcop, PL_hints); | |
3061 | } | |
3062 | if (!SvROK(retval) || SvTYPE(SvRV(retval)) != SVt_PVHV) { | |
3063 | if (SvPOK(retval)) | |
3064 | ||
3065 | /* If caller wants to handle missing properties, let them */ | |
5d3d13d1 | 3066 | if (flags_p && *flags_p & _CORE_SWASH_INIT_RETURN_IF_UNDEF) { |
69794297 KW |
3067 | return NULL; |
3068 | } | |
3069 | Perl_croak(aTHX_ | |
3070 | "Can't find Unicode property definition \"%"SVf"\"", | |
3071 | SVfARG(retval)); | |
3072 | Perl_croak(aTHX_ "SWASHNEW didn't return an HV ref"); | |
3073 | } | |
9a53f6cf | 3074 | } /* End of calling the module to find the swash */ |
36eb48b4 | 3075 | |
83199d38 KW |
3076 | /* If this operation fetched a swash, and we will need it later, get it */ |
3077 | if (retval != &PL_sv_undef | |
3078 | && (minbits == 1 || (flags_p | |
3079 | && ! (*flags_p | |
3080 | & _CORE_SWASH_INIT_USER_DEFINED_PROPERTY)))) | |
3081 | { | |
3082 | swash_hv = MUTABLE_HV(SvRV(retval)); | |
3083 | ||
3084 | /* If we don't already know that there is a user-defined component to | |
3085 | * this swash, and the user has indicated they wish to know if there is | |
3086 | * one (by passing <flags_p>), find out */ | |
3087 | if (flags_p && ! (*flags_p & _CORE_SWASH_INIT_USER_DEFINED_PROPERTY)) { | |
3088 | SV** user_defined = hv_fetchs(swash_hv, "USER_DEFINED", FALSE); | |
3089 | if (user_defined && SvUV(*user_defined)) { | |
3090 | *flags_p |= _CORE_SWASH_INIT_USER_DEFINED_PROPERTY; | |
3091 | } | |
3092 | } | |
3093 | } | |
3094 | ||
36eb48b4 KW |
3095 | /* Make sure there is an inversion list for binary properties */ |
3096 | if (minbits == 1) { | |
3097 | SV** swash_invlistsvp = NULL; | |
3098 | SV* swash_invlist = NULL; | |
9a53f6cf | 3099 | bool invlist_in_swash_is_valid = FALSE; |
02c85471 FC |
3100 | bool swash_invlist_unclaimed = FALSE; /* whether swash_invlist has |
3101 | an unclaimed reference count */ | |
36eb48b4 | 3102 | |
9a53f6cf | 3103 | /* If this operation fetched a swash, get its already existing |
83199d38 | 3104 | * inversion list, or create one for it */ |
36eb48b4 | 3105 | |
83199d38 | 3106 | if (swash_hv) { |
5c9f4bd2 | 3107 | swash_invlistsvp = hv_fetchs(swash_hv, "V", FALSE); |
9a53f6cf KW |
3108 | if (swash_invlistsvp) { |
3109 | swash_invlist = *swash_invlistsvp; | |
3110 | invlist_in_swash_is_valid = TRUE; | |
3111 | } | |
3112 | else { | |
36eb48b4 | 3113 | swash_invlist = _swash_to_invlist(retval); |
02c85471 | 3114 | swash_invlist_unclaimed = TRUE; |
9a53f6cf KW |
3115 | } |
3116 | } | |
3117 | ||
3118 | /* If an inversion list was passed in, have to include it */ | |
3119 | if (invlist) { | |
3120 | ||
3121 | /* Any fetched swash will by now have an inversion list in it; | |
3122 | * otherwise <swash_invlist> will be NULL, indicating that we | |
3123 | * didn't fetch a swash */ | |
3124 | if (swash_invlist) { | |
3125 | ||
3126 | /* Add the passed-in inversion list, which invalidates the one | |
3127 | * already stored in the swash */ | |
3128 | invlist_in_swash_is_valid = FALSE; | |
3129 | _invlist_union(invlist, swash_invlist, &swash_invlist); | |
3130 | } | |
3131 | else { | |
3132 | ||
87367d5f KW |
3133 | /* Here, there is no swash already. Set up a minimal one, if |
3134 | * we are going to return a swash */ | |
3135 | if ((int) _invlist_len(invlist) > invlist_swash_boundary) { | |
971d486f | 3136 | swash_hv = newHV(); |
4aca0fe6 | 3137 | retval = newRV_noinc(MUTABLE_SV(swash_hv)); |
87367d5f | 3138 | } |
9a53f6cf KW |
3139 | swash_invlist = invlist; |
3140 | } | |
9a53f6cf KW |
3141 | } |
3142 | ||
3143 | /* Here, we have computed the union of all the passed-in data. It may | |
3144 | * be that there was an inversion list in the swash which didn't get | |
3145 | * touched; otherwise save the one computed one */ | |
87367d5f KW |
3146 | if (! invlist_in_swash_is_valid |
3147 | && (int) _invlist_len(swash_invlist) > invlist_swash_boundary) | |
3148 | { | |
5c9f4bd2 | 3149 | if (! hv_stores(MUTABLE_HV(SvRV(retval)), "V", swash_invlist)) |
69794297 KW |
3150 | { |
3151 | Perl_croak(aTHX_ "panic: hv_store() unexpectedly failed"); | |
3152 | } | |
cc34d8c5 FC |
3153 | /* We just stole a reference count. */ |
3154 | if (swash_invlist_unclaimed) swash_invlist_unclaimed = FALSE; | |
3155 | else SvREFCNT_inc_simple_void_NN(swash_invlist); | |
9a53f6cf | 3156 | } |
87367d5f | 3157 | |
c41800a8 | 3158 | /* Use the inversion list stand-alone if small enough */ |
87367d5f KW |
3159 | if ((int) _invlist_len(swash_invlist) <= invlist_swash_boundary) { |
3160 | SvREFCNT_dec(retval); | |
02c85471 FC |
3161 | if (!swash_invlist_unclaimed) |
3162 | SvREFCNT_inc_simple_void_NN(swash_invlist); | |
3163 | retval = newRV_noinc(swash_invlist); | |
87367d5f | 3164 | } |
36eb48b4 KW |
3165 | } |
3166 | ||
a0ed51b3 LW |
3167 | return retval; |
3168 | } | |
3169 | ||
035d37be JH |
3170 | |
3171 | /* This API is wrong for special case conversions since we may need to | |
3172 | * return several Unicode characters for a single Unicode character | |
3173 | * (see lib/unicore/SpecCase.txt) The SWASHGET in lib/utf8_heavy.pl is | |
3174 | * the lower-level routine, and it is similarly broken for returning | |
38684baa KW |
3175 | * multiple values. --jhi |
3176 | * For those, you should use to_utf8_case() instead */ | |
b0e3252e | 3177 | /* Now SWASHGET is recasted into S_swatch_get in this file. */ |
680c470c TS |
3178 | |
3179 | /* Note: | |
3180 | * Returns the value of property/mapping C<swash> for the first character | |
3181 | * of the string C<ptr>. If C<do_utf8> is true, the string C<ptr> is | |
3d0f8846 KW |
3182 | * assumed to be in well-formed utf8. If C<do_utf8> is false, the string C<ptr> |
3183 | * is assumed to be in native 8-bit encoding. Caches the swatch in C<swash>. | |
af2af982 KW |
3184 | * |
3185 | * A "swash" is a hash which contains initially the keys/values set up by | |
3186 | * SWASHNEW. The purpose is to be able to completely represent a Unicode | |
3187 | * property for all possible code points. Things are stored in a compact form | |
3188 | * (see utf8_heavy.pl) so that calculation is required to find the actual | |
3189 | * property value for a given code point. As code points are looked up, new | |
3190 | * key/value pairs are added to the hash, so that the calculation doesn't have | |
3191 | * to ever be re-done. Further, each calculation is done, not just for the | |
3192 | * desired one, but for a whole block of code points adjacent to that one. | |
3193 | * For binary properties on ASCII machines, the block is usually for 64 code | |
3194 | * points, starting with a code point evenly divisible by 64. Thus if the | |
3195 | * property value for code point 257 is requested, the code goes out and | |
3196 | * calculates the property values for all 64 code points between 256 and 319, | |
3197 | * and stores these as a single 64-bit long bit vector, called a "swatch", | |
3198 | * under the key for code point 256. The key is the UTF-8 encoding for code | |
3199 | * point 256, minus the final byte. Thus, if the length of the UTF-8 encoding | |
3200 | * for a code point is 13 bytes, the key will be 12 bytes long. If the value | |
3201 | * for code point 258 is then requested, this code realizes that it would be | |
3202 | * stored under the key for 256, and would find that value and extract the | |
3203 | * relevant bit, offset from 256. | |
3204 | * | |
3205 | * Non-binary properties are stored in as many bits as necessary to represent | |
3206 | * their values (32 currently, though the code is more general than that), not | |
3207 | * as single bits, but the principal is the same: the value for each key is a | |
3208 | * vector that encompasses the property values for all code points whose UTF-8 | |
3209 | * representations are represented by the key. That is, for all code points | |
3210 | * whose UTF-8 representations are length N bytes, and the key is the first N-1 | |
3211 | * bytes of that. | |
680c470c | 3212 | */ |
a0ed51b3 | 3213 | UV |
680c470c | 3214 | Perl_swash_fetch(pTHX_ SV *swash, const U8 *ptr, bool do_utf8) |
a0ed51b3 | 3215 | { |
27da23d5 | 3216 | dVAR; |
ef8f7699 | 3217 | HV *const hv = MUTABLE_HV(SvRV(swash)); |
3568d838 JH |
3218 | U32 klen; |
3219 | U32 off; | |
a0ed51b3 | 3220 | STRLEN slen; |
7d85a32c | 3221 | STRLEN needents; |
cfd0369c | 3222 | const U8 *tmps = NULL; |
a0ed51b3 | 3223 | U32 bit; |
979f2922 | 3224 | SV *swatch; |
08fb1ac5 | 3225 | const U8 c = *ptr; |
3568d838 | 3226 | |
7918f24d NC |
3227 | PERL_ARGS_ASSERT_SWASH_FETCH; |
3228 | ||
87367d5f KW |
3229 | /* If it really isn't a hash, it isn't really swash; must be an inversion |
3230 | * list */ | |
3231 | if (SvTYPE(hv) != SVt_PVHV) { | |
3232 | return _invlist_contains_cp((SV*)hv, | |
3233 | (do_utf8) | |
3234 | ? valid_utf8_to_uvchr(ptr, NULL) | |
3235 | : c); | |
3236 | } | |
3237 | ||
08fb1ac5 KW |
3238 | /* We store the values in a "swatch" which is a vec() value in a swash |
3239 | * hash. Code points 0-255 are a single vec() stored with key length | |
3240 | * (klen) 0. All other code points have a UTF-8 representation | |
3241 | * 0xAA..0xYY,0xZZ. A vec() is constructed containing all of them which | |
3242 | * share 0xAA..0xYY, which is the key in the hash to that vec. So the key | |
3243 | * length for them is the length of the encoded char - 1. ptr[klen] is the | |
3244 | * final byte in the sequence representing the character */ | |
3245 | if (!do_utf8 || UTF8_IS_INVARIANT(c)) { | |
3246 | klen = 0; | |
3247 | needents = 256; | |
3248 | off = c; | |
3568d838 | 3249 | } |
08fb1ac5 KW |
3250 | else if (UTF8_IS_DOWNGRADEABLE_START(c)) { |
3251 | klen = 0; | |
3252 | needents = 256; | |
3253 | off = TWO_BYTE_UTF8_TO_NATIVE(c, *(ptr + 1)); | |
979f2922 TS |
3254 | } |
3255 | else { | |
08fb1ac5 KW |
3256 | klen = UTF8SKIP(ptr) - 1; |
3257 | ||
3258 | /* Each vec() stores 2**UTF_ACCUMULATION_SHIFT values. The offset into | |
3259 | * the vec is the final byte in the sequence. (In EBCDIC this is | |
3260 | * converted to I8 to get consecutive values.) To help you visualize | |
3261 | * all this: | |
3262 | * Straight 1047 After final byte | |
3263 | * UTF-8 UTF-EBCDIC I8 transform | |
3264 | * U+0400: \xD0\x80 \xB8\x41\x41 \xB8\x41\xA0 | |
3265 | * U+0401: \xD0\x81 \xB8\x41\x42 \xB8\x41\xA1 | |
3266 | * ... | |
3267 | * U+0409: \xD0\x89 \xB8\x41\x4A \xB8\x41\xA9 | |
3268 | * U+040A: \xD0\x8A \xB8\x41\x51 \xB8\x41\xAA | |
3269 | * ... | |
3270 | * U+0412: \xD0\x92 \xB8\x41\x59 \xB8\x41\xB2 | |
3271 | * U+0413: \xD0\x93 \xB8\x41\x62 \xB8\x41\xB3 | |
3272 | * ... | |
3273 | * U+041B: \xD0\x9B \xB8\x41\x6A \xB8\x41\xBB | |
3274 | * U+041C: \xD0\x9C \xB8\x41\x70 \xB8\x41\xBC | |
3275 | * ... | |
3276 | * U+041F: \xD0\x9F \xB8\x41\x73 \xB8\x41\xBF | |
3277 | * U+0420: \xD0\xA0 \xB8\x42\x41 \xB8\x42\x41 | |
3278 | * | |
3279 | * (There are no discontinuities in the elided (...) entries.) | |
3280 | * The UTF-8 key for these 33 code points is '\xD0' (which also is the | |
3281 | * key for the next 31, up through U+043F, whose UTF-8 final byte is | |
3282 | * \xBF). Thus in UTF-8, each key is for a vec() for 64 code points. | |
3283 | * The final UTF-8 byte, which ranges between \x80 and \xBF, is an | |
3284 | * index into the vec() swatch (after subtracting 0x80, which we | |
3285 | * actually do with an '&'). | |
3286 | * In UTF-EBCDIC, each key is for a 32 code point vec(). The first 32 | |
3287 | * code points above have key '\xB8\x41'. The final UTF-EBCDIC byte has | |
3288 | * dicontinuities which go away by transforming it into I8, and we | |
3289 | * effectively subtract 0xA0 to get the index. */ | |
979f2922 | 3290 | needents = (1 << UTF_ACCUMULATION_SHIFT); |
bc3632a8 | 3291 | off = NATIVE_UTF8_TO_I8(ptr[klen]) & UTF_CONTINUATION_MASK; |
979f2922 | 3292 | } |
7d85a32c | 3293 | |
a0ed51b3 LW |
3294 | /* |
3295 | * This single-entry cache saves about 1/3 of the utf8 overhead in test | |
3296 | * suite. (That is, only 7-8% overall over just a hash cache. Still, | |
3297 | * it's nothing to sniff at.) Pity we usually come through at least | |
3298 | * two function calls to get here... | |
3299 | * | |
3300 | * NB: this code assumes that swatches are never modified, once generated! | |
3301 | */ | |
3302 | ||
3568d838 | 3303 | if (hv == PL_last_swash_hv && |
a0ed51b3 | 3304 | klen == PL_last_swash_klen && |
27da23d5 | 3305 | (!klen || memEQ((char *)ptr, (char *)PL_last_swash_key, klen)) ) |
a0ed51b3 LW |
3306 | { |
3307 | tmps = PL_last_swash_tmps; | |
3308 | slen = PL_last_swash_slen; | |
3309 | } | |
3310 | else { | |
3311 | /* Try our second-level swatch cache, kept in a hash. */ | |
e1ec3a88 | 3312 | SV** svp = hv_fetch(hv, (const char*)ptr, klen, FALSE); |
a0ed51b3 | 3313 | |
b0e3252e | 3314 | /* If not cached, generate it via swatch_get */ |
979f2922 | 3315 | if (!svp || !SvPOK(*svp) |
08fb1ac5 KW |
3316 | || !(tmps = (const U8*)SvPV_const(*svp, slen))) |
3317 | { | |
3318 | if (klen) { | |
3319 | const UV code_point = valid_utf8_to_uvchr(ptr, NULL); | |
3320 | swatch = swatch_get(swash, | |
3321 | code_point & ~((UV)needents - 1), | |
3322 | needents); | |
3323 | } | |
3324 | else { /* For the first 256 code points, the swatch has a key of | |
3325 | length 0 */ | |
3326 | swatch = swatch_get(swash, 0, needents); | |
3327 | } | |
979f2922 | 3328 | |
923e4eb5 | 3329 | if (IN_PERL_COMPILETIME) |
623e6609 | 3330 | CopHINTS_set(PL_curcop, PL_hints); |
a0ed51b3 | 3331 | |
979f2922 | 3332 | svp = hv_store(hv, (const char *)ptr, klen, swatch, 0); |
a0ed51b3 | 3333 | |
979f2922 TS |
3334 | if (!svp || !(tmps = (U8*)SvPV(*svp, slen)) |
3335 | || (slen << 3) < needents) | |
5637ef5b NC |
3336 | Perl_croak(aTHX_ "panic: swash_fetch got improper swatch, " |
3337 | "svp=%p, tmps=%p, slen=%"UVuf", needents=%"UVuf, | |
3338 | svp, tmps, (UV)slen, (UV)needents); | |
a0ed51b3 LW |
3339 | } |
3340 | ||
3341 | PL_last_swash_hv = hv; | |
16d8f38a | 3342 | assert(klen <= sizeof(PL_last_swash_key)); |
eac04b2e | 3343 | PL_last_swash_klen = (U8)klen; |
cfd0369c NC |
3344 | /* FIXME change interpvar.h? */ |
3345 | PL_last_swash_tmps = (U8 *) tmps; | |
a0ed51b3 LW |
3346 | PL_last_swash_slen = slen; |
3347 | if (klen) | |
3348 | Copy(ptr, PL_last_swash_key, klen, U8); | |
3349 | } | |
3350 | ||
9faf8d75 | 3351 | switch ((int)((slen << 3) / needents)) { |
a0ed51b3 LW |
3352 | case 1: |
3353 | bit = 1 << (off & 7); | |
3354 | off >>= 3; | |
3355 | return (tmps[off] & bit) != 0; | |
3356 | case 8: | |
3357 | return tmps[off]; | |
3358 | case 16: | |
3359 | off <<= 1; | |
3360 | return (tmps[off] << 8) + tmps[off + 1] ; | |
3361 | case 32: | |
3362 | off <<= 2; | |
3363 | return (tmps[off] << 24) + (tmps[off+1] << 16) + (tmps[off+2] << 8) + tmps[off + 3] ; | |
3364 | } | |
5637ef5b NC |
3365 | Perl_croak(aTHX_ "panic: swash_fetch got swatch of unexpected bit width, " |
3366 | "slen=%"UVuf", needents=%"UVuf, (UV)slen, (UV)needents); | |
670f1322 | 3367 | NORETURN_FUNCTION_END; |
a0ed51b3 | 3368 | } |
2b9d42f0 | 3369 | |
319009ee KW |
3370 | /* Read a single line of the main body of the swash input text. These are of |
3371 | * the form: | |
3372 | * 0053 0056 0073 | |
3373 | * where each number is hex. The first two numbers form the minimum and | |
3374 | * maximum of a range, and the third is the value associated with the range. | |
3375 | * Not all swashes should have a third number | |
3376 | * | |
3377 | * On input: l points to the beginning of the line to be examined; it points | |
3378 | * to somewhere in the string of the whole input text, and is | |
3379 | * terminated by a \n or the null string terminator. | |
3380 | * lend points to the null terminator of that string | |
3381 | * wants_value is non-zero if the swash expects a third number | |
3382 | * typestr is the name of the swash's mapping, like 'ToLower' | |
3383 | * On output: *min, *max, and *val are set to the values read from the line. | |
3384 | * returns a pointer just beyond the line examined. If there was no | |
3385 | * valid min number on the line, returns lend+1 | |
3386 | */ | |
3387 | ||
3388 | STATIC U8* | |
3389 | S_swash_scan_list_line(pTHX_ U8* l, U8* const lend, UV* min, UV* max, UV* val, | |
3390 | const bool wants_value, const U8* const typestr) | |
3391 | { | |
3392 | const int typeto = typestr[0] == 'T' && typestr[1] == 'o'; | |
3393 | STRLEN numlen; /* Length of the number */ | |
02470786 KW |
3394 | I32 flags = PERL_SCAN_SILENT_ILLDIGIT |
3395 | | PERL_SCAN_DISALLOW_PREFIX | |
3396 | | PERL_SCAN_SILENT_NON_PORTABLE; | |
319009ee KW |
3397 | |
3398 | /* nl points to the next \n in the scan */ | |
3399 | U8* const nl = (U8*)memchr(l, '\n', lend - l); | |
3400 | ||
3401 | /* Get the first number on the line: the range minimum */ | |
3402 | numlen = lend - l; | |
3403 | *min = grok_hex((char *)l, &numlen, &flags, NULL); | |
3404 | if (numlen) /* If found a hex number, position past it */ | |
3405 | l += numlen; | |
3406 | else if (nl) { /* Else, go handle next line, if any */ | |
3407 | return nl + 1; /* 1 is length of "\n" */ | |
3408 | } | |
3409 | else { /* Else, no next line */ | |
3410 | return lend + 1; /* to LIST's end at which \n is not found */ | |
3411 | } | |
3412 | ||
3413 | /* The max range value follows, separated by a BLANK */ | |
3414 | if (isBLANK(*l)) { | |
3415 | ++l; | |
02470786 KW |
3416 | flags = PERL_SCAN_SILENT_ILLDIGIT |
3417 | | PERL_SCAN_DISALLOW_PREFIX | |
3418 | | PERL_SCAN_SILENT_NON_PORTABLE; | |
319009ee KW |
3419 | numlen = lend - l; |
3420 | *max = grok_hex((char *)l, &numlen, &flags, NULL); | |
3421 | if (numlen) | |
3422 | l += numlen; | |
3423 | else /* If no value here, it is a single element range */ | |
3424 | *max = *min; | |
3425 | ||
3426 | /* Non-binary tables have a third entry: what the first element of the | |
24303724 | 3427 | * range maps to. The map for those currently read here is in hex */ |
319009ee KW |
3428 | if (wants_value) { |
3429 | if (isBLANK(*l)) { | |
3430 | ++l; | |
f2a7d0fc KW |
3431 | flags = PERL_SCAN_SILENT_ILLDIGIT |
3432 | | PERL_SCAN_DISALLOW_PREFIX | |
3433 | | PERL_SCAN_SILENT_NON_PORTABLE; | |
3434 | numlen = lend - l; | |
3435 | *val = grok_hex((char *)l, &numlen, &flags, NULL); | |
3436 | if (numlen) | |
3437 | l += numlen; | |
3438 | else | |
3439 | *val = 0; | |
319009ee KW |
3440 | } |
3441 | else { | |
3442 | *val = 0; | |
3443 | if (typeto) { | |
dcbac5bb | 3444 | /* diag_listed_as: To%s: illegal mapping '%s' */ |
319009ee KW |
3445 | Perl_croak(aTHX_ "%s: illegal mapping '%s'", |
3446 | typestr, l); | |
3447 | } | |
3448 | } | |
3449 | } | |
3450 | else | |
3451 | *val = 0; /* bits == 1, then any val should be ignored */ | |
3452 | } | |
3453 | else { /* Nothing following range min, should be single element with no | |
3454 | mapping expected */ | |
3455 | *max = *min; | |
3456 | if (wants_value) { | |
3457 | *val = 0; | |
3458 | if (typeto) { | |
dcbac5bb | 3459 | /* diag_listed_as: To%s: illegal mapping '%s' */ |
319009ee KW |
3460 | Perl_croak(aTHX_ "%s: illegal mapping '%s'", typestr, l); |
3461 | } | |
3462 | } | |
3463 | else | |
3464 | *val = 0; /* bits == 1, then val should be ignored */ | |
3465 | } | |
3466 | ||
3467 | /* Position to next line if any, or EOF */ | |
3468 | if (nl) | |
3469 | l = nl + 1; | |
3470 | else | |
3471 | l = lend; | |
3472 | ||
3473 | return l; | |
3474 | } | |
3475 | ||
979f2922 TS |
3476 | /* Note: |
3477 | * Returns a swatch (a bit vector string) for a code point sequence | |
3478 | * that starts from the value C<start> and comprises the number C<span>. | |
3479 | * A C<swash> must be an object created by SWASHNEW (see lib/utf8_heavy.pl). | |
3480 | * Should be used via swash_fetch, which will cache the swatch in C<swash>. | |
3481 | */ | |
3482 | STATIC SV* | |
b0e3252e | 3483 | S_swatch_get(pTHX_ SV* swash, UV start, UV span) |
979f2922 TS |
3484 | { |
3485 | SV *swatch; | |
77f9f126 | 3486 | U8 *l, *lend, *x, *xend, *s, *send; |
979f2922 | 3487 | STRLEN lcur, xcur, scur; |
ef8f7699 | 3488 | HV *const hv = MUTABLE_HV(SvRV(swash)); |
5c9f4bd2 | 3489 | SV** const invlistsvp = hv_fetchs(hv, "V", FALSE); |
36eb48b4 | 3490 | |
88d45d28 KW |
3491 | SV** listsvp = NULL; /* The string containing the main body of the table */ |
3492 | SV** extssvp = NULL; | |
3493 | SV** invert_it_svp = NULL; | |
3494 | U8* typestr = NULL; | |
786861f5 KW |
3495 | STRLEN bits; |
3496 | STRLEN octets; /* if bits == 1, then octets == 0 */ | |
3497 | UV none; | |
3498 | UV end = start + span; | |
972dd592 | 3499 | |
36eb48b4 | 3500 | if (invlistsvp == NULL) { |
786861f5 KW |
3501 | SV** const bitssvp = hv_fetchs(hv, "BITS", FALSE); |
3502 | SV** const nonesvp = hv_fetchs(hv, "NONE", FALSE); | |
3503 | SV** const typesvp = hv_fetchs(hv, "TYPE", FALSE); | |
3504 | extssvp = hv_fetchs(hv, "EXTRAS", FALSE); | |
3505 | listsvp = hv_fetchs(hv, "LIST", FALSE); | |
3506 | invert_it_svp = hv_fetchs(hv, "INVERT_IT", FALSE); | |
3507 | ||
3508 | bits = SvUV(*bitssvp); | |
3509 | none = SvUV(*nonesvp); | |
3510 | typestr = (U8*)SvPV_nolen(*typesvp); | |
3511 | } | |
36eb48b4 KW |
3512 | else { |
3513 | bits = 1; | |
3514 | none = 0; | |
3515 | } | |
786861f5 | 3516 | octets = bits >> 3; /* if bits == 1, then octets == 0 */ |
979f2922 | 3517 | |
b0e3252e | 3518 | PERL_ARGS_ASSERT_SWATCH_GET; |
7918f24d | 3519 | |
979f2922 | 3520 | if (bits != 1 && bits != 8 && bits != 16 && bits != 32) { |
b0e3252e | 3521 | Perl_croak(aTHX_ "panic: swatch_get doesn't expect bits %"UVuf, |
660a4616 | 3522 | (UV)bits); |
979f2922 TS |
3523 | } |
3524 | ||
84ea5ef6 KW |
3525 | /* If overflowed, use the max possible */ |
3526 | if (end < start) { | |
3527 | end = UV_MAX; | |
3528 | span = end - start; | |
3529 | } | |
3530 | ||
979f2922 | 3531 | /* create and initialize $swatch */ |
979f2922 | 3532 | scur = octets ? (span * octets) : (span + 7) / 8; |
e524fe40 NC |
3533 | swatch = newSV(scur); |
3534 | SvPOK_on(swatch); | |
979f2922 TS |
3535 | s = (U8*)SvPVX(swatch); |
3536 | if (octets && none) { | |
0bd48802 | 3537 | const U8* const e = s + scur; |
979f2922 TS |
3538 | while (s < e) { |
3539 | if (bits == 8) | |
3540 | *s++ = (U8)(none & 0xff); | |
3541 | else if (bits == 16) { | |
3542 | *s++ = (U8)((none >> 8) & 0xff); | |
3543 | *s++ = (U8)( none & 0xff); | |
3544 | } | |
3545 | else if (bits == 32) { | |
3546 | *s++ = (U8)((none >> 24) & 0xff); | |
3547 | *s++ = (U8)((none >> 16) & 0xff); | |
3548 | *s++ = (U8)((none >> 8) & 0xff); | |
3549 | *s++ = (U8)( none & 0xff); | |
3550 | } | |
3551 | } | |
3552 | *s = '\0'; | |
3553 | } | |
3554 | else { | |
3555 | (void)memzero((U8*)s, scur + 1); | |
3556 | } | |
3557 | SvCUR_set(swatch, scur); | |
3558 | s = (U8*)SvPVX(swatch); | |
3559 | ||
36eb48b4 KW |
3560 | if (invlistsvp) { /* If has an inversion list set up use that */ |
3561 | _invlist_populate_swatch(*invlistsvp, start, end, s); | |
3562 | return swatch; | |
3563 | } | |
3564 | ||
3565 | /* read $swash->{LIST} */ | |
979f2922 TS |
3566 | l = (U8*)SvPV(*listsvp, lcur); |
3567 | lend = l + lcur; | |
3568 | while (l < lend) { | |
8ed25d53 | 3569 | UV min, max, val, upper; |
319009ee KW |
3570 | l = S_swash_scan_list_line(aTHX_ l, lend, &min, &max, &val, |
3571 | cBOOL(octets), typestr); | |
3572 | if (l > lend) { | |
979f2922 TS |
3573 | break; |
3574 | } | |
3575 | ||
972dd592 | 3576 | /* If looking for something beyond this range, go try the next one */ |
979f2922 TS |
3577 | if (max < start) |
3578 | continue; | |
3579 | ||
8ed25d53 KW |
3580 | /* <end> is generally 1 beyond where we want to set things, but at the |
3581 | * platform's infinity, where we can't go any higher, we want to | |
3582 | * include the code point at <end> */ | |
3583 | upper = (max < end) | |
3584 | ? max | |
3585 | : (max != UV_MAX || end != UV_MAX) | |
3586 | ? end - 1 | |
3587 | : end; | |
3588 | ||
979f2922 | 3589 | if (octets) { |
35da51f7 | 3590 | UV key; |
979f2922 TS |
3591 | if (min < start) { |
3592 | if (!none || val < none) { | |
3593 | val += start - min; | |
3594 | } | |
3595 | min = start; | |
3596 | } | |
8ed25d53 | 3597 | for (key = min; key <= upper; key++) { |
979f2922 | 3598 | STRLEN offset; |
979f2922 TS |
3599 | /* offset must be non-negative (start <= min <= key < end) */ |
3600 | offset = octets * (key - start); | |
3601 | if (bits == 8) | |
3602 | s[offset] = (U8)(val & 0xff); | |
3603 | else if (bits == 16) { | |
3604 | s[offset ] = (U8)((val >> 8) & 0xff); | |
3605 | s[offset + 1] = (U8)( val & 0xff); | |
3606 | } | |
3607 | else if (bits == 32) { | |
3608 | s[offset ] = (U8)((val >> 24) & 0xff); | |
3609 | s[offset + 1] = (U8)((val >> 16) & 0xff); | |
3610 | s[offset + 2] = (U8)((val >> 8) & 0xff); | |
3611 | s[offset + 3] = (U8)( val & 0xff); | |
3612 | } | |
3613 | ||
3614 | if (!none || val < none) | |
3615 | ++val; | |
3616 | } | |
3617 | } | |
711a919c | 3618 | else { /* bits == 1, then val should be ignored */ |
35da51f7 | 3619 | UV key; |
979f2922 TS |
3620 | if (min < start) |
3621 | min = start; | |
6cb05c12 | 3622 | |
8ed25d53 | 3623 | for (key = min; key <= upper; key++) { |
0bd48802 | 3624 | const STRLEN offset = (STRLEN)(key - start); |
979f2922 TS |
3625 | s[offset >> 3] |= 1 << (offset & 7); |
3626 | } | |
3627 | } | |
3628 | } /* while */ | |
979f2922 | 3629 | |
9479a769 | 3630 | /* Invert if the data says it should be. Assumes that bits == 1 */ |
77f9f126 | 3631 | if (invert_it_svp && SvUV(*invert_it_svp)) { |
0bda3001 KW |
3632 | |
3633 | /* Unicode properties should come with all bits above PERL_UNICODE_MAX | |
3634 | * be 0, and their inversion should also be 0, as we don't succeed any | |
3635 | * Unicode property matches for non-Unicode code points */ | |
3636 | if (start <= PERL_UNICODE_MAX) { | |
3637 | ||
3638 | /* The code below assumes that we never cross the | |
3639 | * Unicode/above-Unicode boundary in a range, as otherwise we would | |
3640 | * have to figure out where to stop flipping the bits. Since this | |
3641 | * boundary is divisible by a large power of 2, and swatches comes | |
3642 | * in small powers of 2, this should be a valid assumption */ | |
3643 | assert(start + span - 1 <= PERL_UNICODE_MAX); | |
3644 | ||
507a8485 KW |
3645 | send = s + scur; |
3646 | while (s < send) { | |
3647 | *s = ~(*s); | |
3648 | s++; | |
3649 | } | |
0bda3001 | 3650 | } |
77f9f126 KW |
3651 | } |
3652 | ||
d73c39c5 KW |
3653 | /* read $swash->{EXTRAS} |
3654 | * This code also copied to swash_to_invlist() below */ | |
979f2922 TS |
3655 | x = (U8*)SvPV(*extssvp, xcur); |
3656 | xend = x + xcur; | |
3657 | while (x < xend) { | |
3658 | STRLEN namelen; | |
3659 | U8 *namestr; | |
3660 | SV** othersvp; | |
3661 | HV* otherhv; | |
3662 | STRLEN otherbits; | |
3663 | SV **otherbitssvp, *other; | |
711a919c | 3664 | U8 *s, *o, *nl; |
979f2922 TS |
3665 | STRLEN slen, olen; |
3666 | ||
35da51f7 | 3667 | const U8 opc = *x++; |
979f2922 TS |
3668 | if (opc == '\n') |
3669 | continue; | |
3670 | ||
3671 | nl = (U8*)memchr(x, '\n', xend - x); | |
3672 | ||
3673 | if (opc != '-' && opc != '+' && opc != '!' && opc != '&') { | |
3674 | if (nl) { | |
3675 | x = nl + 1; /* 1 is length of "\n" */ | |
3676 | continue; | |
3677 | } | |
3678 | else { | |
3679 | x = xend; /* to EXTRAS' end at which \n is not found */ | |
3680 | break; | |
3681 | } | |
3682 | } | |
3683 | ||
3684 | namestr = x; | |
3685 | if (nl) { | |
3686 | namelen = nl - namestr; | |
3687 | x = nl + 1; | |
3688 | } | |
3689 | else { | |
3690 | namelen = xend - namestr; | |
3691 | x = xend; | |
3692 | } | |
3693 | ||
3694 | othersvp = hv_fetch(hv, (char *)namestr, namelen, FALSE); | |
ef8f7699 | 3695 | otherhv = MUTABLE_HV(SvRV(*othersvp)); |
017a3ce5 | 3696 | otherbitssvp = hv_fetchs(otherhv, "BITS", FALSE); |
979f2922 TS |
3697 | otherbits = (STRLEN)SvUV(*otherbitssvp); |
3698 | if (bits < otherbits) | |
5637ef5b NC |
3699 | Perl_croak(aTHX_ "panic: swatch_get found swatch size mismatch, " |
3700 | "bits=%"UVuf", otherbits=%"UVuf, (UV)bits, (UV)otherbits); | |
979f2922 TS |
3701 | |
3702 | /* The "other" swatch must be destroyed after. */ | |
b0e3252e | 3703 | other = swatch_get(*othersvp, start, span); |
979f2922 TS |
3704 | o = (U8*)SvPV(other, olen); |
3705 | ||
3706 | if (!olen) | |
b0e3252e | 3707 | Perl_croak(aTHX_ "panic: swatch_get got improper swatch"); |
979f2922 TS |
3708 | |
3709 | s = (U8*)SvPV(swatch, slen); | |
3710 | if (bits == 1 && otherbits == 1) { | |
3711 | if (slen != olen) | |
5637ef5b NC |
3712 | Perl_croak(aTHX_ "panic: swatch_get found swatch length " |
3713 | "mismatch, slen=%"UVuf", olen=%"UVuf, | |
3714 | (UV)slen, (UV)olen); | |
979f2922 TS |
3715 | |
3716 | switch (opc) { | |
3717 | case '+': | |
3718 | while (slen--) | |
3719 | *s++ |= *o++; | |
3720 | break; | |
3721 | case '!': | |
3722 | while (slen--) | |
3723 | *s++ |= ~*o++; | |
3724 | break; | |
3725 | case '-': | |
3726 | while (slen--) | |
3727 | *s++ &= ~*o++; | |
3728 | break; | |
3729 | case '&': | |
3730 | while (slen--) | |
3731 | *s++ &= *o++; | |
3732 | break; | |
3733 | default: | |
3734 | break; | |
3735 | } | |
3736 | } | |
711a919c | 3737 | else { |
979f2922 TS |
3738 | STRLEN otheroctets = otherbits >> 3; |
3739 | STRLEN offset = 0; | |
35da51f7 | 3740 | U8* const send = s + slen; |
979f2922 TS |
3741 | |
3742 | while (s < send) { | |
3743 | UV otherval = 0; | |
3744 | ||
3745 | if (otherbits == 1) { | |
3746 | otherval = (o[offset >> 3] >> (offset & 7)) & 1; | |
3747 | ++offset; | |
3748 | } | |
3749 | else { | |
3750 | STRLEN vlen = otheroctets; | |
3751 | otherval = *o++; | |
3752 | while (--vlen) { | |
3753 | otherval <<= 8; | |
3754 | otherval |= *o++; | |
3755 | } | |
3756 | } | |
3757 | ||
711a919c | 3758 | if (opc == '+' && otherval) |
6f207bd3 | 3759 | NOOP; /* replace with otherval */ |
979f2922 TS |
3760 | else if (opc == '!' && !otherval) |
3761 | otherval = 1; | |
3762 | else if (opc == '-' && otherval) | |
3763 | otherval = 0; | |
3764 | else if (opc == '&' && !otherval) | |
3765 | otherval = 0; | |
3766 | else { | |
711a919c | 3767 | s += octets; /* no replacement */ |
979f2922 TS |
3768 | continue; |
3769 | } | |
3770 | ||
3771 | if (bits == 8) | |
3772 | *s++ = (U8)( otherval & 0xff); | |
3773 | else if (bits == 16) { | |
3774 | *s++ = (U8)((otherval >> 8) & 0xff); | |
3775 | *s++ = (U8)( otherval & 0xff); | |
3776 | } | |
3777 | else if (bits == 32) { | |
3778 | *s++ = (U8)((otherval >> 24) & 0xff); | |
3779 | *s++ = (U8)((otherval >> 16) & 0xff); | |
3780 | *s++ = (U8)((otherval >> 8) & 0xff); | |
3781 | *s++ = (U8)( otherval & 0xff); | |
3782 | } | |
3783 | } | |
3784 | } | |
3785 | sv_free(other); /* through with it! */ | |
3786 | } /* while */ | |
3787 | return swatch; | |
3788 | } | |
3789 | ||
064c021d | 3790 | HV* |
4c2e1131 | 3791 | Perl__swash_inversion_hash(pTHX_ SV* const swash) |
064c021d KW |
3792 | { |
3793 | ||
79a2a0e8 | 3794 | /* Subject to change or removal. For use only in regcomp.c and regexec.c |
5662e334 KW |
3795 | * Can't be used on a property that is subject to user override, as it |
3796 | * relies on the value of SPECIALS in the swash which would be set by | |
3797 | * utf8_heavy.pl to the hash in the non-overriden file, and hence is not set | |
3798 | * for overridden properties | |
064c021d KW |
3799 | * |
3800 | * Returns a hash which is the inversion and closure of a swash mapping. | |
3801 | * For example, consider the input lines: | |
3802 | * 004B 006B | |
3803 | * 004C 006C | |
3804 | * 212A 006B | |
3805 | * | |
3806 | * The returned hash would have two keys, the utf8 for 006B and the utf8 for | |
3807 | * 006C. The value for each key is an array. For 006C, the array would | |
17d5c697 | 3808 | * have two elements, the utf8 for itself, and for 004C. For 006B, there |
064c021d KW |
3809 | * would be three elements in its array, the utf8 for 006B, 004B and 212A. |
3810 | * | |
3811 | * Essentially, for any code point, it gives all the code points that map to | |
3812 | * it, or the list of 'froms' for that point. | |
3813 | * | |
5662e334 KW |
3814 | * Currently it ignores any additions or deletions from other swashes, |
3815 | * looking at just the main body of the swash, and if there are SPECIALS | |
3816 | * in the swash, at that hash | |
3817 | * | |
3818 | * The specials hash can be extra code points, and most likely consists of | |
3819 | * maps from single code points to multiple ones (each expressed as a string | |
3820 | * of utf8 characters). This function currently returns only 1-1 mappings. | |
3821 | * However consider this possible input in the specials hash: | |
3822 | * "\xEF\xAC\x85" => "\x{0073}\x{0074}", # U+FB05 => 0073 0074 | |
3823 | * "\xEF\xAC\x86" => "\x{0073}\x{0074}", # U+FB06 => 0073 0074 | |
3824 | * | |
3825 | * Both FB05 and FB06 map to the same multi-char sequence, which we don't | |
3826 | * currently handle. But it also means that FB05 and FB06 are equivalent in | |
3827 | * a 1-1 mapping which we should handle, and this relationship may not be in | |
3828 | * the main table. Therefore this function examines all the multi-char | |
3829 | * sequences and adds the 1-1 mappings that come out of that. */ | |
064c021d KW |
3830 | |
3831 | U8 *l, *lend; | |
3832 | STRLEN lcur; | |
3833 | HV *const hv = MUTABLE_HV(SvRV(swash)); | |
3834 | ||
923b6d4e KW |
3835 | /* The string containing the main body of the table. This will have its |
3836 | * assertion fail if the swash has been converted to its inversion list */ | |
064c021d KW |
3837 | SV** const listsvp = hv_fetchs(hv, "LIST", FALSE); |
3838 | ||
3839 | SV** const typesvp = hv_fetchs(hv, "TYPE", FALSE); | |
3840 | SV** const bitssvp = hv_fetchs(hv, "BITS", FALSE); | |
3841 | SV** const nonesvp = hv_fetchs(hv, "NONE", FALSE); | |
3842 | /*SV** const extssvp = hv_fetchs(hv, "EXTRAS", FALSE);*/ | |
3843 | const U8* const typestr = (U8*)SvPV_nolen(*typesvp); | |
3844 | const STRLEN bits = SvUV(*bitssvp); | |
3845 | const STRLEN octets = bits >> 3; /* if bits == 1, then octets == 0 */ | |
3846 | const UV none = SvUV(*nonesvp); | |
5662e334 | 3847 | SV **specials_p = hv_fetchs(hv, "SPECIALS", 0); |
064c021d KW |
3848 | |
3849 | HV* ret = newHV(); | |
3850 | ||
3851 | PERL_ARGS_ASSERT__SWASH_INVERSION_HASH; | |
3852 | ||
3853 | /* Must have at least 8 bits to get the mappings */ | |
3854 | if (bits != 8 && bits != 16 && bits != 32) { | |
3855 | Perl_croak(aTHX_ "panic: swash_inversion_hash doesn't expect bits %"UVuf, | |
3856 | (UV)bits); | |
3857 | } | |
3858 | ||
5662e334 KW |
3859 | if (specials_p) { /* It might be "special" (sometimes, but not always, a |
3860 | mapping to more than one character */ | |
3861 | ||
3862 | /* Construct an inverse mapping hash for the specials */ | |
3863 | HV * const specials_hv = MUTABLE_HV(SvRV(*specials_p)); | |
3864 | HV * specials_inverse = newHV(); | |
3865 | char *char_from; /* the lhs of the map */ | |
3866 | I32 from_len; /* its byte length */ | |
3867 | char *char_to; /* the rhs of the map */ | |
3868 | I32 to_len; /* its byte length */ | |
3869 | SV *sv_to; /* and in a sv */ | |
3870 | AV* from_list; /* list of things that map to each 'to' */ | |
3871 | ||
3872 | hv_iterinit(specials_hv); | |
3873 | ||
3874 | /* The keys are the characters (in utf8) that map to the corresponding | |
3875 | * utf8 string value. Iterate through the list creating the inverse | |
3876 | * list. */ | |
3877 | while ((sv_to = hv_iternextsv(specials_hv, &char_from, &from_len))) { | |
3878 | SV** listp; | |
3879 | if (! SvPOK(sv_to)) { | |
5637ef5b NC |
3880 | Perl_croak(aTHX_ "panic: value returned from hv_iternextsv() " |
3881 | "unexpectedly is not a string, flags=%lu", | |
3882 | (unsigned long)SvFLAGS(sv_to)); | |
5662e334 | 3883 | } |
4b88fb76 | 3884 | /*DEBUG_U(PerlIO_printf(Perl_debug_log, "Found mapping from %"UVXf", First char of to is %"UVXf"\n", valid_utf8_to_uvchr((U8*) char_from, 0), valid_utf8_to_uvchr((U8*) SvPVX(sv_to), 0)));*/ |
5662e334 KW |
3885 | |
3886 | /* Each key in the inverse list is a mapped-to value, and the key's | |
3887 | * hash value is a list of the strings (each in utf8) that map to | |
3888 | * it. Those strings are all one character long */ | |
3889 | if ((listp = hv_fetch(specials_inverse, | |
3890 | SvPVX(sv_to), | |
3891 | SvCUR(sv_to), 0))) | |
3892 | { | |
3893 | from_list = (AV*) *listp; | |
3894 | } | |
3895 | else { /* No entry yet for it: create one */ | |
3896 | from_list = newAV(); | |
3897 | if (! hv_store(specials_inverse, | |
3898 | SvPVX(sv_to), | |
3899 | SvCUR(sv_to), | |
3900 | (SV*) from_list, 0)) | |
3901 | { | |
3902 | Perl_croak(aTHX_ "panic: hv_store() unexpectedly failed"); | |
3903 | } | |
3904 | } | |
3905 | ||
3906 | /* Here have the list associated with this 'to' (perhaps newly | |
3907 | * created and empty). Just add to it. Note that we ASSUME that | |
3908 | * the input is guaranteed to not have duplications, so we don't | |
3909 | * check for that. Duplications just slow down execution time. */ | |
3910 | av_push(from_list, newSVpvn_utf8(char_from, from_len, TRUE)); | |
3911 | } | |
3912 | ||
3913 | /* Here, 'specials_inverse' contains the inverse mapping. Go through | |
3914 | * it looking for cases like the FB05/FB06 examples above. There would | |
3915 | * be an entry in the hash like | |
3916 | * 'st' => [ FB05, FB06 ] | |
3917 | * In this example we will create two lists that get stored in the | |
3918 | * returned hash, 'ret': | |
3919 | * FB05 => [ FB05, FB06 ] | |
3920 | * FB06 => [ FB05, FB06 ] | |
3921 | * | |
3922 | * Note that there is nothing to do if the array only has one element. | |
3923 | * (In the normal 1-1 case handled below, we don't have to worry about | |
3924 | * two lists, as everything gets tied to the single list that is | |
3925 | * generated for the single character 'to'. But here, we are omitting | |
3926 | * that list, ('st' in the example), so must have multiple lists.) */ | |
3927 | while ((from_list = (AV *) hv_iternextsv(specials_inverse, | |
3928 | &char_to, &to_len))) | |
3929 | { | |
3930 | if (av_len(from_list) > 0) { | |
c70927a6 | 3931 | SSize_t i; |
5662e334 KW |
3932 | |
3933 | /* We iterate over all combinations of i,j to place each code | |
3934 | * point on each list */ | |
3935 | for (i = 0; i <= av_len(from_list); i++) { | |
c70927a6 | 3936 | SSize_t j; |
5662e334 KW |
3937 | AV* i_list = newAV(); |
3938 | SV** entryp = av_fetch(from_list, i, FALSE); | |
3939 | if (entryp == NULL) { | |
3940 | Perl_croak(aTHX_ "panic: av_fetch() unexpectedly failed"); | |
3941 | } | |
3942 | if (hv_fetch(ret, SvPVX(*entryp), SvCUR(*entryp), FALSE)) { | |
3943 | Perl_croak(aTHX_ "panic: unexpected entry for %s", SvPVX(*entryp)); | |
3944 | } | |
3945 | if (! hv_store(ret, SvPVX(*entryp), SvCUR(*entryp), | |
3946 | (SV*) i_list, FALSE)) | |
3947 | { | |
3948 | Perl_croak(aTHX_ "panic: hv_store() unexpectedly failed"); | |
3949 | } | |
3950 | ||
4b88fb76 | 3951 | /* For debugging: UV u = valid_utf8_to_uvchr((U8*) SvPVX(*entryp), 0);*/ |
5662e334 KW |
3952 | for (j = 0; j <= av_len(from_list); j++) { |
3953 | entryp = av_fetch(from_list, j, FALSE); | |
3954 | if (entryp == NULL) { | |
3955 | Perl_croak(aTHX_ "panic: av_fetch() unexpectedly failed"); | |
3956 | } | |
3957 | ||
3958 | /* When i==j this adds itself to the list */ | |
4b88fb76 KW |
3959 | av_push(i_list, newSVuv(utf8_to_uvchr_buf( |
3960 | (U8*) SvPVX(*entryp), | |
3961 | (U8*) SvPVX(*entryp) + SvCUR(*entryp), | |
3962 | 0))); | |
4637d003 | 3963 | /*DEBUG_U(PerlIO_printf(Perl_debug_log, "%s: %d: Adding %"UVXf" to list for %"UVXf"\n", __FILE__, __LINE__, valid_utf8_to_uvchr((U8*) SvPVX(*entryp), 0), u));*/ |
5662e334 KW |
3964 | } |
3965 | } | |
3966 | } | |
3967 | } | |
3968 | SvREFCNT_dec(specials_inverse); /* done with it */ | |
3969 | } /* End of specials */ | |
3970 | ||
064c021d KW |
3971 | /* read $swash->{LIST} */ |
3972 | l = (U8*)SvPV(*listsvp, lcur); | |
3973 | lend = l + lcur; | |
3974 | ||
3975 | /* Go through each input line */ | |
3976 | while (l < lend) { | |
3977 | UV min, max, val; | |
3978 | UV inverse; | |
3979 | l = S_swash_scan_list_line(aTHX_ l, lend, &min, &max, &val, | |
3980 | cBOOL(octets), typestr); | |
3981 | if (l > lend) { | |
3982 | break; | |
3983 | } | |
3984 | ||
3985 | /* Each element in the range is to be inverted */ | |
3986 | for (inverse = min; inverse <= max; inverse++) { | |
3987 | AV* list; | |
064c021d KW |
3988 | SV** listp; |
3989 | IV i; | |
3990 | bool found_key = FALSE; | |
5662e334 | 3991 | bool found_inverse = FALSE; |
064c021d KW |
3992 | |
3993 | /* The key is the inverse mapping */ | |
3994 | char key[UTF8_MAXBYTES+1]; | |
c80e42f3 | 3995 | char* key_end = (char *) uvchr_to_utf8((U8*) key, val); |
064c021d KW |
3996 | STRLEN key_len = key_end - key; |
3997 | ||
064c021d KW |
3998 | /* Get the list for the map */ |
3999 | if ((listp = hv_fetch(ret, key, key_len, FALSE))) { | |
4000 | list = (AV*) *listp; | |
4001 | } | |
4002 | else { /* No entry yet for it: create one */ | |
4003 | list = newAV(); | |
4004 | if (! hv_store(ret, key, key_len, (SV*) list, FALSE)) { | |
4005 | Perl_croak(aTHX_ "panic: hv_store() unexpectedly failed"); | |
4006 | } | |
4007 | } | |
4008 | ||
5662e334 KW |
4009 | /* Look through list to see if this inverse mapping already is |
4010 | * listed, or if there is a mapping to itself already */ | |
508f7cfa | 4011 | for (i = 0; i <= av_len(list); i++) { |
064c021d KW |
4012 | SV** entryp = av_fetch(list, i, FALSE); |
4013 | SV* entry; | |
4014 | if (entryp == NULL) { | |
4015 | Perl_croak(aTHX_ "panic: av_fetch() unexpectedly failed"); | |
4016 | } | |
4017 | entry = *entryp; | |
5662e334 | 4018 | /*DEBUG_U(PerlIO_printf(Perl_debug_log, "list for %"UVXf" contains %"UVXf"\n", val, SvUV(entry)));*/ |
56ca34ca | 4019 | if (SvUV(entry) == val) { |
064c021d | 4020 | found_key = TRUE; |
5662e334 KW |
4021 | } |
4022 | if (SvUV(entry) == inverse) { | |
4023 | found_inverse = TRUE; | |
4024 | } | |
4025 | ||
4026 | /* No need to continue searching if found everything we are | |
4027 | * looking for */ | |
4028 | if (found_key && found_inverse) { | |
064c021d KW |
4029 | break; |
4030 | } | |
4031 | } | |
56ca34ca KW |
4032 | |
4033 | /* Make sure there is a mapping to itself on the list */ | |
064c021d | 4034 | if (! found_key) { |
d397ff6a | 4035 | av_push(list, newSVuv(val)); |
4637d003 | 4036 | /*DEBUG_U(PerlIO_printf(Perl_debug_log, "%s: %d: Adding %"UVXf" to list for %"UVXf"\n", __FILE__, __LINE__, val, val));*/ |
064c021d KW |
4037 | } |
4038 | ||
4039 | ||
4040 | /* Simply add the value to the list */ | |
5662e334 KW |
4041 | if (! found_inverse) { |
4042 | av_push(list, newSVuv(inverse)); | |
4637d003 | 4043 | /*DEBUG_U(PerlIO_printf(Perl_debug_log, "%s: %d: Adding %"UVXf" to list for %"UVXf"\n", __FILE__, __LINE__, inverse, val));*/ |
5662e334 | 4044 | } |
064c021d | 4045 | |
b0e3252e | 4046 | /* swatch_get() increments the value of val for each element in the |
064c021d KW |
4047 | * range. That makes more compact tables possible. You can |
4048 | * express the capitalization, for example, of all consecutive | |
4049 | * letters with a single line: 0061\t007A\t0041 This maps 0061 to | |
4050 | * 0041, 0062 to 0042, etc. I (khw) have never understood 'none', | |
bd3f2f94 | 4051 | * and it's not documented; it appears to be used only in |
b0e3252e | 4052 | * implementing tr//; I copied the semantics from swatch_get(), just |
bd3f2f94 | 4053 | * in case */ |
064c021d KW |
4054 | if (!none || val < none) { |
4055 | ++val; | |
4056 | } | |
4057 | } | |
4058 | } | |
4059 | ||
4060 | return ret; | |
4061 | } | |
4062 | ||
a25abddc | 4063 | SV* |
d764b54e KW |
4064 | Perl__swash_to_invlist(pTHX_ SV* const swash) |
4065 | { | |
4066 | ||
ed92f1b3 KW |
4067 | /* Subject to change or removal. For use only in one place in regcomp.c. |
4068 | * Ownership is given to one reference count in the returned SV* */ | |
d764b54e KW |
4069 | |
4070 | U8 *l, *lend; | |
4071 | char *loc; | |
4072 | STRLEN lcur; | |
4073 | HV *const hv = MUTABLE_HV(SvRV(swash)); | |
4074 | UV elements = 0; /* Number of elements in the inversion list */ | |
b443038a | 4075 | U8 empty[] = ""; |
923b6d4e KW |
4076 | SV** listsvp; |
4077 | SV** typesvp; | |
4078 | SV** bitssvp; | |
4079 | SV** extssvp; | |
4080 | SV** invert_it_svp; | |
d764b54e | 4081 | |
923b6d4e KW |
4082 | U8* typestr; |
4083 | STRLEN bits; | |
4084 | STRLEN octets; /* if bits == 1, then octets == 0 */ | |
d73c39c5 KW |
4085 | U8 *x, *xend; |
4086 | STRLEN xcur; | |
d764b54e | 4087 | |
a25abddc | 4088 | SV* invlist; |
d764b54e | 4089 | |
b81740c0 KW |
4090 | PERL_ARGS_ASSERT__SWASH_TO_INVLIST; |
4091 | ||
923b6d4e KW |
4092 | /* If not a hash, it must be the swash's inversion list instead */ |
4093 | if (SvTYPE(hv) != SVt_PVHV) { | |
ed92f1b3 | 4094 | return SvREFCNT_inc_simple_NN((SV*) hv); |
923b6d4e KW |
4095 | } |
4096 | ||
4097 | /* The string containing the main body of the table */ | |
4098 | listsvp = hv_fetchs(hv, "LIST", FALSE); | |
4099 | typesvp = hv_fetchs(hv, "TYPE", FALSE); | |
4100 | bitssvp = hv_fetchs(hv, "BITS", FALSE); | |
4101 | extssvp = hv_fetchs(hv, "EXTRAS", FALSE); | |
4102 | invert_it_svp = hv_fetchs(hv, "INVERT_IT", FALSE); | |
4103 | ||
4104 | typestr = (U8*)SvPV_nolen(*typesvp); | |
4105 | bits = SvUV(*bitssvp); | |
4106 | octets = bits >> 3; /* if bits == 1, then octets == 0 */ | |
4107 | ||
d764b54e | 4108 | /* read $swash->{LIST} */ |
b443038a KW |
4109 | if (SvPOK(*listsvp)) { |
4110 | l = (U8*)SvPV(*listsvp, lcur); | |
4111 | } | |
4112 | else { | |
4113 | /* LIST legitimately doesn't contain a string during compilation phases | |
4114 | * of Perl itself, before the Unicode tables are generated. In this | |
4115 | * case, just fake things up by creating an empty list */ | |
4116 | l = empty; | |
4117 | lcur = 0; | |
4118 | } | |
d764b54e KW |
4119 | loc = (char *) l; |
4120 | lend = l + lcur; | |
4121 | ||
4122 | /* Scan the input to count the number of lines to preallocate array size | |
4123 | * based on worst possible case, which is each line in the input creates 2 | |
4124 | * elements in the inversion list: 1) the beginning of a range in the list; | |
4125 | * 2) the beginning of a range not in the list. */ | |
4126 | while ((loc = (strchr(loc, '\n'))) != NULL) { | |
4127 | elements += 2; | |
4128 | loc++; | |
4129 | } | |
4130 | ||
4131 | /* If the ending is somehow corrupt and isn't a new line, add another | |
4132 | * element for the final range that isn't in the inversion list */ | |
fd05e003 KW |
4133 | if (! (*lend == '\n' |
4134 | || (*lend == '\0' && (lcur == 0 || *(lend - 1) == '\n')))) | |
4135 | { | |
d764b54e KW |
4136 | elements++; |
4137 | } | |
4138 | ||
4139 | invlist = _new_invlist(elements); | |
4140 | ||
4141 | /* Now go through the input again, adding each range to the list */ | |
4142 | while (l < lend) { | |
4143 | UV start, end; | |
4144 | UV val; /* Not used by this function */ | |
4145 | ||
4146 | l = S_swash_scan_list_line(aTHX_ l, lend, &start, &end, &val, | |
4147 | cBOOL(octets), typestr); | |
4148 | ||
4149 | if (l > lend) { | |
4150 | break; | |
4151 | } | |
4152 | ||
9d501133 | 4153 | invlist = _add_range_to_invlist(invlist, start, end); |
d764b54e KW |
4154 | } |
4155 | ||
77f9f126 KW |
4156 | /* Invert if the data says it should be */ |
4157 | if (invert_it_svp && SvUV(*invert_it_svp)) { | |
45bb2768 | 4158 | _invlist_invert_prop(invlist); |
77f9f126 KW |
4159 | } |
4160 | ||
b0e3252e | 4161 | /* This code is copied from swatch_get() |
d73c39c5 KW |
4162 | * read $swash->{EXTRAS} */ |
4163 | x = (U8*)SvPV(*extssvp, xcur); | |
4164 | xend = x + xcur; | |
4165 | while (x < xend) { | |
4166 | STRLEN namelen; | |
4167 | U8 *namestr; | |
4168 | SV** othersvp; | |
4169 | HV* otherhv; | |
4170 | STRLEN otherbits; | |
4171 | SV **otherbitssvp, *other; | |
4172 | U8 *nl; | |
4173 | ||
4174 | const U8 opc = *x++; | |
4175 | if (opc == '\n') | |
4176 | continue; | |
4177 | ||
4178 | nl = (U8*)memchr(x, '\n', xend - x); | |
4179 | ||
4180 | if (opc != '-' && opc != '+' && opc != '!' && opc != '&') { | |
4181 | if (nl) { | |
4182 | x = nl + 1; /* 1 is length of "\n" */ | |
4183 | continue; | |
4184 | } | |
4185 | else { | |
4186 | x = xend; /* to EXTRAS' end at which \n is not found */ | |
4187 | break; | |
4188 | } | |
4189 | } | |
4190 | ||
4191 | namestr = x; | |
4192 | if (nl) { | |
4193 | namelen = nl - namestr; | |
4194 | x = nl + 1; | |
4195 | } | |
4196 | else { | |
4197 | namelen = xend - namestr; | |
4198 | x = xend; | |
4199 | } | |
4200 | ||
4201 | othersvp = hv_fetch(hv, (char *)namestr, namelen, FALSE); | |
4202 | otherhv = MUTABLE_HV(SvRV(*othersvp)); | |
4203 | otherbitssvp = hv_fetchs(otherhv, "BITS", FALSE); | |
4204 | otherbits = (STRLEN)SvUV(*otherbitssvp); | |
4205 | ||
4206 | if (bits != otherbits || bits != 1) { | |
5637ef5b NC |
4207 | Perl_croak(aTHX_ "panic: _swash_to_invlist only operates on boolean " |
4208 | "properties, bits=%"UVuf", otherbits=%"UVuf, | |
4209 | (UV)bits, (UV)otherbits); | |
d73c39c5 KW |
4210 | } |
4211 | ||
4212 | /* The "other" swatch must be destroyed after. */ | |
4213 | other = _swash_to_invlist((SV *)*othersvp); | |
4214 | ||
b0e3252e | 4215 | /* End of code copied from swatch_get() */ |
d73c39c5 KW |
4216 | switch (opc) { |
4217 | case '+': | |
4218 | _invlist_union(invlist, other, &invlist); | |
4219 | break; | |
4220 | case '!': | |
6c46377d | 4221 | _invlist_union_maybe_complement_2nd(invlist, other, TRUE, &invlist); |
d73c39c5 KW |
4222 | break; |
4223 | case '-': | |
4224 | _invlist_subtract(invlist, other, &invlist); | |
4225 | break; | |
4226 | case '&': | |
4227 | _invlist_intersection(invlist, other, &invlist); | |
4228 | break; | |
4229 | default: | |
4230 | break; | |
4231 | } | |
4232 | sv_free(other); /* through with it! */ | |
4233 | } | |
4234 | ||
d764b54e KW |
4235 | return invlist; |
4236 | } | |
4237 | ||
3fdfee00 KW |
4238 | SV* |
4239 | Perl__get_swash_invlist(pTHX_ SV* const swash) | |
4240 | { | |
872dd7e0 | 4241 | SV** ptr; |
3fdfee00 KW |
4242 | |
4243 | PERL_ARGS_ASSERT__GET_SWASH_INVLIST; | |
4244 | ||
87367d5f | 4245 | if (! SvROK(swash)) { |
872dd7e0 KW |
4246 | return NULL; |
4247 | } | |
4248 | ||
87367d5f KW |
4249 | /* If it really isn't a hash, it isn't really swash; must be an inversion |
4250 | * list */ | |
4251 | if (SvTYPE(SvRV(swash)) != SVt_PVHV) { | |
4252 | return SvRV(swash); | |
4253 | } | |
872dd7e0 | 4254 | |
87367d5f | 4255 | ptr = hv_fetchs(MUTABLE_HV(SvRV(swash)), "V", FALSE); |
3fdfee00 KW |
4256 | if (! ptr) { |
4257 | return NULL; | |
4258 | } | |
4259 | ||
4260 | return *ptr; | |
4261 | } | |
4262 | ||
0876b9a0 | 4263 | bool |
5aaab254 | 4264 | Perl_check_utf8_print(pTHX_ const U8* s, const STRLEN len) |
0876b9a0 KW |
4265 | { |
4266 | /* May change: warns if surrogates, non-character code points, or | |
af2af982 KW |
4267 | * non-Unicode code points are in s which has length len bytes. Returns |
4268 | * TRUE if none found; FALSE otherwise. The only other validity check is | |
4269 | * to make sure that this won't exceed the string's length */ | |
0876b9a0 KW |
4270 | |
4271 | const U8* const e = s + len; | |
4272 | bool ok = TRUE; | |
4273 | ||
4274 | PERL_ARGS_ASSERT_CHECK_UTF8_PRINT; | |
4275 | ||
4276 | while (s < e) { | |
4277 | if (UTF8SKIP(s) > len) { | |
4278 | Perl_ck_warner_d(aTHX_ packWARN(WARN_UTF8), | |
4279 | "%s in %s", unees, PL_op ? OP_DESC(PL_op) : "print"); | |
4280 | return FALSE; | |
4281 | } | |
732fbc05 | 4282 | if (UNLIKELY(*s >= UTF8_FIRST_PROBLEMATIC_CODE_POINT_FIRST_BYTE)) { |
0876b9a0 KW |
4283 | STRLEN char_len; |
4284 | if (UTF8_IS_SUPER(s)) { | |
8457b38f | 4285 | if (ckWARN_d(WARN_NON_UNICODE)) { |
4b88fb76 | 4286 | UV uv = utf8_to_uvchr_buf(s, e, &char_len); |
8457b38f KW |
4287 | Perl_warner(aTHX_ packWARN(WARN_NON_UNICODE), |
4288 | "Code point 0x%04"UVXf" is not Unicode, may not be portable", uv); | |
4289 | ok = FALSE; | |
4290 | } | |
0876b9a0 KW |
4291 | } |
4292 | else if (UTF8_IS_SURROGATE(s)) { | |
8457b38f | 4293 | if (ckWARN_d(WARN_SURROGATE)) { |
4b88fb76 | 4294 | UV uv = utf8_to_uvchr_buf(s, e, &char_len); |
8457b38f KW |
4295 | Perl_warner(aTHX_ packWARN(WARN_SURROGATE), |
4296 | "Unicode surrogate U+%04"UVXf" is illegal in UTF-8", uv); | |
4297 | ok = FALSE; | |
4298 | } | |
0876b9a0 KW |
4299 | } |
4300 | else if | |
8457b38f KW |
4301 | ((UTF8_IS_NONCHAR_GIVEN_THAT_NON_SUPER_AND_GE_PROBLEMATIC(s)) |
4302 | && (ckWARN_d(WARN_NONCHAR))) | |
0876b9a0 | 4303 | { |
4b88fb76 | 4304 | UV uv = utf8_to_uvchr_buf(s, e, &char_len); |
8457b38f | 4305 | Perl_warner(aTHX_ packWARN(WARN_NONCHAR), |
0876b9a0 KW |
4306 | "Unicode non-character U+%04"UVXf" is illegal for open interchange", uv); |
4307 | ok = FALSE; | |
4308 | } | |
4309 | } | |
4310 | s += UTF8SKIP(s); | |
4311 | } | |
4312 | ||
4313 | return ok; | |
4314 | } | |
4315 | ||
0f830e0b | 4316 | /* |
87cea99e | 4317 | =for apidoc pv_uni_display |
d2cc3551 | 4318 | |
a1433954 KW |
4319 | Build to the scalar C<dsv> a displayable version of the string C<spv>, |
4320 | length C<len>, the displayable version being at most C<pvlim> bytes long | |
d2cc3551 | 4321 | (if longer, the rest is truncated and "..." will be appended). |
0a2ef054 | 4322 | |
a1433954 | 4323 | The C<flags> argument can have UNI_DISPLAY_ISPRINT set to display |
00e86452 | 4324 | isPRINT()able characters as themselves, UNI_DISPLAY_BACKSLASH |
0a2ef054 JH |
4325 | to display the \\[nrfta\\] as the backslashed versions (like '\n') |
4326 | (UNI_DISPLAY_BACKSLASH is preferred over UNI_DISPLAY_ISPRINT for \\). | |
4327 | UNI_DISPLAY_QQ (and its alias UNI_DISPLAY_REGEX) have both | |
4328 | UNI_DISPLAY_BACKSLASH and UNI_DISPLAY_ISPRINT turned on. | |
4329 | ||
a1433954 | 4330 | The pointer to the PV of the C<dsv> is returned. |
d2cc3551 JH |
4331 | |
4332 | =cut */ | |
e6b2e755 | 4333 | char * |
e1ec3a88 | 4334 | Perl_pv_uni_display(pTHX_ SV *dsv, const U8 *spv, STRLEN len, STRLEN pvlim, UV flags) |
e6b2e755 JH |
4335 | { |
4336 | int truncated = 0; | |
e1ec3a88 | 4337 | const char *s, *e; |
e6b2e755 | 4338 | |
7918f24d NC |
4339 | PERL_ARGS_ASSERT_PV_UNI_DISPLAY; |
4340 | ||
76f68e9b | 4341 | sv_setpvs(dsv, ""); |
7fddd944 | 4342 | SvUTF8_off(dsv); |
e1ec3a88 | 4343 | for (s = (const char *)spv, e = s + len; s < e; s += UTF8SKIP(s)) { |
e6b2e755 | 4344 | UV u; |
a49f32c6 NC |
4345 | /* This serves double duty as a flag and a character to print after |
4346 | a \ when flags & UNI_DISPLAY_BACKSLASH is true. | |
4347 | */ | |
4348 | char ok = 0; | |
c728cb41 | 4349 | |
e6b2e755 JH |
4350 | if (pvlim && SvCUR(dsv) >= pvlim) { |
4351 | truncated++; | |
4352 | break; | |
4353 | } | |
4b88fb76 | 4354 | u = utf8_to_uvchr_buf((U8*)s, (U8*)e, 0); |
c728cb41 | 4355 | if (u < 256) { |
a3b680e6 | 4356 | const unsigned char c = (unsigned char)u & 0xFF; |
0bd48802 | 4357 | if (flags & UNI_DISPLAY_BACKSLASH) { |
a49f32c6 | 4358 | switch (c) { |
c728cb41 | 4359 | case '\n': |
a49f32c6 | 4360 | ok = 'n'; break; |
c728cb41 | 4361 | case '\r': |
a49f32c6 | 4362 | ok = 'r'; break; |
c728cb41 | 4363 | case '\t': |
a49f32c6 | 4364 | ok = 't'; break; |
c728cb41 | 4365 | case '\f': |
a49f32c6 | 4366 | ok = 'f'; break; |
c728cb41 | 4367 | case '\a': |
a49f32c6 | 4368 | ok = 'a'; break; |
c728cb41 | 4369 | case '\\': |
a49f32c6 | 4370 | ok = '\\'; break; |
c728cb41 JH |
4371 | default: break; |
4372 | } | |
a49f32c6 | 4373 | if (ok) { |
88c9ea1e | 4374 | const char string = ok; |
76f68e9b | 4375 | sv_catpvs(dsv, "\\"); |
5e7aa789 | 4376 | sv_catpvn(dsv, &string, 1); |
a49f32c6 | 4377 | } |
c728cb41 | 4378 | } |
00e86452 | 4379 | /* isPRINT() is the locale-blind version. */ |
a49f32c6 | 4380 | if (!ok && (flags & UNI_DISPLAY_ISPRINT) && isPRINT(c)) { |
88c9ea1e | 4381 | const char string = c; |
5e7aa789 | 4382 | sv_catpvn(dsv, &string, 1); |
a49f32c6 | 4383 | ok = 1; |
0a2ef054 | 4384 | } |
c728cb41 JH |
4385 | } |
4386 | if (!ok) | |
9e55ce06 | 4387 | Perl_sv_catpvf(aTHX_ dsv, "\\x{%"UVxf"}", u); |
e6b2e755 JH |
4388 | } |
4389 | if (truncated) | |
396482e1 | 4390 | sv_catpvs(dsv, "..."); |
48ef279e | 4391 | |
e6b2e755 JH |
4392 | return SvPVX(dsv); |
4393 | } | |
2b9d42f0 | 4394 | |
d2cc3551 | 4395 | /* |
87cea99e | 4396 | =for apidoc sv_uni_display |
d2cc3551 | 4397 | |
a1433954 KW |
4398 | Build to the scalar C<dsv> a displayable version of the scalar C<sv>, |
4399 | the displayable version being at most C<pvlim> bytes long | |
d2cc3551 | 4400 | (if longer, the rest is truncated and "..." will be appended). |
0a2ef054 | 4401 | |
a1433954 | 4402 | The C<flags> argument is as in L</pv_uni_display>(). |
0a2ef054 | 4403 | |
a1433954 | 4404 | The pointer to the PV of the C<dsv> is returned. |
d2cc3551 | 4405 | |
d4c19fe8 AL |
4406 | =cut |
4407 | */ | |
e6b2e755 JH |
4408 | char * |
4409 | Perl_sv_uni_display(pTHX_ SV *dsv, SV *ssv, STRLEN pvlim, UV flags) | |
4410 | { | |
8cdde9f8 NC |
4411 | const char * const ptr = |
4412 | isREGEXP(ssv) ? RX_WRAPPED((REGEXP*)ssv) : SvPVX_const(ssv); | |
4413 | ||
7918f24d NC |
4414 | PERL_ARGS_ASSERT_SV_UNI_DISPLAY; |
4415 | ||
8cdde9f8 | 4416 | return Perl_pv_uni_display(aTHX_ dsv, (const U8*)ptr, |
cfd0369c | 4417 | SvCUR(ssv), pvlim, flags); |
701a277b JH |
4418 | } |
4419 | ||
d2cc3551 | 4420 | /* |
e6226b18 | 4421 | =for apidoc foldEQ_utf8 |
d2cc3551 | 4422 | |
a1433954 | 4423 | Returns true if the leading portions of the strings C<s1> and C<s2> (either or both |
e6226b18 | 4424 | of which may be in UTF-8) are the same case-insensitively; false otherwise. |
d51c1b21 | 4425 | How far into the strings to compare is determined by other input parameters. |
8b35872c | 4426 | |
a1433954 KW |
4427 | If C<u1> is true, the string C<s1> is assumed to be in UTF-8-encoded Unicode; |
4428 | otherwise it is assumed to be in native 8-bit encoding. Correspondingly for C<u2> | |
4429 | with respect to C<s2>. | |
8b35872c | 4430 | |
a1433954 KW |
4431 | If the byte length C<l1> is non-zero, it says how far into C<s1> to check for fold |
4432 | equality. In other words, C<s1>+C<l1> will be used as a goal to reach. The | |
8b35872c | 4433 | scan will not be considered to be a match unless the goal is reached, and |
a1433954 KW |
4434 | scanning won't continue past that goal. Correspondingly for C<l2> with respect to |
4435 | C<s2>. | |
4436 | ||
4437 | If C<pe1> is non-NULL and the pointer it points to is not NULL, that pointer is | |
03bb5c85 KW |
4438 | considered an end pointer to the position 1 byte past the maximum point |
4439 | in C<s1> beyond which scanning will not continue under any circumstances. | |
4440 | (This routine assumes that UTF-8 encoded input strings are not malformed; | |
4441 | malformed input can cause it to read past C<pe1>). | |
4442 | This means that if both C<l1> and C<pe1> are specified, and C<pe1> | |
a1433954 KW |
4443 | is less than C<s1>+C<l1>, the match will never be successful because it can |
4444 | never | |
d51c1b21 | 4445 | get as far as its goal (and in fact is asserted against). Correspondingly for |
a1433954 | 4446 | C<pe2> with respect to C<s2>. |
8b35872c | 4447 | |
a1433954 KW |
4448 | At least one of C<s1> and C<s2> must have a goal (at least one of C<l1> and |
4449 | C<l2> must be non-zero), and if both do, both have to be | |
8b35872c KW |
4450 | reached for a successful match. Also, if the fold of a character is multiple |
4451 | characters, all of them must be matched (see tr21 reference below for | |
4452 | 'folding'). | |
4453 | ||
a1433954 KW |
4454 | Upon a successful match, if C<pe1> is non-NULL, |
4455 | it will be set to point to the beginning of the I<next> character of C<s1> | |
4456 | beyond what was matched. Correspondingly for C<pe2> and C<s2>. | |
d2cc3551 JH |
4457 | |
4458 | For case-insensitiveness, the "casefolding" of Unicode is used | |
4459 | instead of upper/lowercasing both the characters, see | |
a1433954 | 4460 | L<http://www.unicode.org/unicode/reports/tr21/> (Case Mappings). |
d2cc3551 JH |
4461 | |
4462 | =cut */ | |
a33c29bc KW |
4463 | |
4464 | /* A flags parameter has been added which may change, and hence isn't | |
4465 | * externally documented. Currently it is: | |
4466 | * 0 for as-documented above | |
4467 | * FOLDEQ_UTF8_NOMIX_ASCII meaning that if a non-ASCII character folds to an | |
4468 | ASCII one, to not match | |
5e64d0fa KW |
4469 | * FOLDEQ_UTF8_LOCALE meaning that locale rules are to be used for code |
4470 | * points below 256; unicode rules for above 255; and | |
4471 | * folds that cross those boundaries are disallowed, | |
4472 | * like the NOMIX_ASCII option | |
18f762c3 KW |
4473 | * FOLDEQ_S1_ALREADY_FOLDED s1 has already been folded before calling this |
4474 | * routine. This allows that step to be skipped. | |
4475 | * FOLDEQ_S2_ALREADY_FOLDED Similarly. | |
a33c29bc | 4476 | */ |
701a277b | 4477 | I32 |
5aaab254 | 4478 | Perl_foldEQ_utf8_flags(pTHX_ const char *s1, char **pe1, UV l1, bool u1, const char *s2, char **pe2, UV l2, bool u2, U32 flags) |
332ddc25 | 4479 | { |
8b35872c | 4480 | dVAR; |
eb578fdb KW |
4481 | const U8 *p1 = (const U8*)s1; /* Point to current char */ |
4482 | const U8 *p2 = (const U8*)s2; | |
4483 | const U8 *g1 = NULL; /* goal for s1 */ | |
4484 | const U8 *g2 = NULL; | |
4485 | const U8 *e1 = NULL; /* Don't scan s1 past this */ | |
4486 | U8 *f1 = NULL; /* Point to current folded */ | |
4487 | const U8 *e2 = NULL; | |
4488 | U8 *f2 = NULL; | |
48ef279e | 4489 | STRLEN n1 = 0, n2 = 0; /* Number of bytes in current char */ |
8b35872c KW |
4490 | U8 foldbuf1[UTF8_MAXBYTES_CASE+1]; |
4491 | U8 foldbuf2[UTF8_MAXBYTES_CASE+1]; | |
8b35872c | 4492 | |
eda9cac1 | 4493 | PERL_ARGS_ASSERT_FOLDEQ_UTF8_FLAGS; |
8b35872c | 4494 | |
18f762c3 | 4495 | assert( ! ((flags & (FOLDEQ_UTF8_NOMIX_ASCII | FOLDEQ_UTF8_LOCALE)) |
b08f1bd5 KW |
4496 | && (flags & (FOLDEQ_S1_ALREADY_FOLDED | FOLDEQ_S2_ALREADY_FOLDED)))); |
4497 | /* The algorithm is to trial the folds without regard to the flags on | |
4498 | * the first line of the above assert(), and then see if the result | |
4499 | * violates them. This means that the inputs can't be pre-folded to a | |
4500 | * violating result, hence the assert. This could be changed, with the | |
4501 | * addition of extra tests here for the already-folded case, which would | |
4502 | * slow it down. That cost is more than any possible gain for when these | |
4503 | * flags are specified, as the flags indicate /il or /iaa matching which | |
4504 | * is less common than /iu, and I (khw) also believe that real-world /il | |
4505 | * and /iaa matches are most likely to involve code points 0-255, and this | |
4506 | * function only under rare conditions gets called for 0-255. */ | |
18f762c3 | 4507 | |
8b35872c | 4508 | if (pe1) { |
48ef279e | 4509 | e1 = *(U8**)pe1; |
8b35872c KW |
4510 | } |
4511 | ||
4512 | if (l1) { | |
48ef279e | 4513 | g1 = (const U8*)s1 + l1; |
8b35872c KW |
4514 | } |
4515 | ||
4516 | if (pe2) { | |
48ef279e | 4517 | e2 = *(U8**)pe2; |
8b35872c KW |
4518 | } |
4519 | ||
4520 | if (l2) { | |
48ef279e | 4521 | g2 = (const U8*)s2 + l2; |
8b35872c KW |
4522 | } |
4523 | ||
4524 | /* Must have at least one goal */ | |
4525 | assert(g1 || g2); | |
4526 | ||
4527 | if (g1) { | |
4528 | ||
48ef279e KW |
4529 | /* Will never match if goal is out-of-bounds */ |
4530 | assert(! e1 || e1 >= g1); | |
8b35872c | 4531 | |
48ef279e KW |
4532 | /* Here, there isn't an end pointer, or it is beyond the goal. We |
4533 | * only go as far as the goal */ | |
4534 | e1 = g1; | |
8b35872c | 4535 | } |
313b38e5 NC |
4536 | else { |
4537 | assert(e1); /* Must have an end for looking at s1 */ | |
4538 | } | |
8b35872c KW |
4539 | |
4540 | /* Same for goal for s2 */ | |
4541 | if (g2) { | |
48ef279e KW |
4542 | assert(! e2 || e2 >= g2); |
4543 | e2 = g2; | |
8b35872c | 4544 | } |
313b38e5 NC |
4545 | else { |
4546 | assert(e2); | |
4547 | } | |
8b35872c | 4548 | |
18f762c3 KW |
4549 | /* If both operands are already folded, we could just do a memEQ on the |
4550 | * whole strings at once, but it would be better if the caller realized | |
4551 | * this and didn't even call us */ | |
4552 | ||
8b35872c KW |
4553 | /* Look through both strings, a character at a time */ |
4554 | while (p1 < e1 && p2 < e2) { | |
4555 | ||
d51c1b21 | 4556 | /* If at the beginning of a new character in s1, get its fold to use |
5e64d0fa KW |
4557 | * and the length of the fold. (exception: locale rules just get the |
4558 | * character to a single byte) */ | |
48ef279e | 4559 | if (n1 == 0) { |
18f762c3 KW |
4560 | if (flags & FOLDEQ_S1_ALREADY_FOLDED) { |
4561 | f1 = (U8 *) p1; | |
4562 | n1 = UTF8SKIP(f1); | |
18f762c3 KW |
4563 | } |
4564 | else { | |
a6d5f321 KW |
4565 | /* If in locale matching, we use two sets of rules, depending |
4566 | * on if the code point is above or below 255. Here, we test | |
4567 | * for and handle locale rules */ | |
1b4059d5 | 4568 | if ((flags & FOLDEQ_UTF8_LOCALE) |
a4f12ed7 | 4569 | && (! u1 || ! UTF8_IS_ABOVE_LATIN1(*p1))) |
5e64d0fa | 4570 | { |
1b4059d5 | 4571 | /* There is no mixing of code points above and below 255. */ |
a4f12ed7 | 4572 | if (u2 && UTF8_IS_ABOVE_LATIN1(*p2)) { |
1b4059d5 KW |
4573 | return 0; |
4574 | } | |
5e64d0fa | 4575 | |
1b4059d5 KW |
4576 | /* We handle locale rules by converting, if necessary, the |
4577 | * code point to a single byte. */ | |
4578 | if (! u1 || UTF8_IS_INVARIANT(*p1)) { | |
4579 | *foldbuf1 = *p1; | |
4580 | } | |
4581 | else { | |
94bb8c36 | 4582 | *foldbuf1 = TWO_BYTE_UTF8_TO_NATIVE(*p1, *(p1 + 1)); |
1b4059d5 KW |
4583 | } |
4584 | n1 = 1; | |
5e64d0fa | 4585 | } |
a6d5f321 KW |
4586 | else if (isASCII(*p1)) { /* Note, that here won't be both |
4587 | ASCII and using locale rules */ | |
1b4059d5 KW |
4588 | |
4589 | /* If trying to mix non- with ASCII, and not supposed to, | |
4590 | * fail */ | |
4591 | if ((flags & FOLDEQ_UTF8_NOMIX_ASCII) && ! isASCII(*p2)) { | |
4592 | return 0; | |
4593 | } | |
4594 | n1 = 1; | |
d22b930b | 4595 | *foldbuf1 = toFOLD(*p1); |
5e64d0fa | 4596 | } |
1b4059d5 KW |
4597 | else if (u1) { |
4598 | to_utf8_fold(p1, foldbuf1, &n1); | |
a33c29bc | 4599 | } |
9cef8533 | 4600 | else { /* Not utf8, get utf8 fold */ |
f4cd282c | 4601 | to_uni_fold(*p1, foldbuf1, &n1); |
1b4059d5 KW |
4602 | } |
4603 | f1 = foldbuf1; | |
18f762c3 | 4604 | } |
48ef279e | 4605 | } |
8b35872c | 4606 | |
48ef279e | 4607 | if (n2 == 0) { /* Same for s2 */ |
18f762c3 KW |
4608 | if (flags & FOLDEQ_S2_ALREADY_FOLDED) { |
4609 | f2 = (U8 *) p2; | |
4610 | n2 = UTF8SKIP(f2); | |
4611 | } | |
4612 | else { | |
227968da | 4613 | if ((flags & FOLDEQ_UTF8_LOCALE) |
a4f12ed7 | 4614 | && (! u2 || ! UTF8_IS_ABOVE_LATIN1(*p2))) |
5e64d0fa | 4615 | { |
227968da KW |
4616 | /* Here, the next char in s2 is < 256. We've already |
4617 | * worked on s1, and if it isn't also < 256, can't match */ | |
a4f12ed7 | 4618 | if (u1 && UTF8_IS_ABOVE_LATIN1(*p1)) { |
227968da KW |
4619 | return 0; |
4620 | } | |
4621 | if (! u2 || UTF8_IS_INVARIANT(*p2)) { | |
4622 | *foldbuf2 = *p2; | |
4623 | } | |
4624 | else { | |
94bb8c36 | 4625 | *foldbuf2 = TWO_BYTE_UTF8_TO_NATIVE(*p2, *(p2 + 1)); |
227968da KW |
4626 | } |
4627 | ||
4628 | /* Use another function to handle locale rules. We've made | |
4629 | * sure that both characters to compare are single bytes */ | |
4630 | if (! foldEQ_locale((char *) f1, (char *) foldbuf2, 1)) { | |
4631 | return 0; | |
4632 | } | |
4633 | n1 = n2 = 0; | |
5e64d0fa | 4634 | } |
227968da | 4635 | else if (isASCII(*p2)) { |
ba9114af | 4636 | if ((flags & FOLDEQ_UTF8_NOMIX_ASCII) && ! isASCII(*p1)) { |
227968da KW |
4637 | return 0; |
4638 | } | |
4639 | n2 = 1; | |
d22b930b | 4640 | *foldbuf2 = toFOLD(*p2); |
5e64d0fa | 4641 | } |
227968da KW |
4642 | else if (u2) { |
4643 | to_utf8_fold(p2, foldbuf2, &n2); | |
5001101e | 4644 | } |
227968da | 4645 | else { |
f4cd282c | 4646 | to_uni_fold(*p2, foldbuf2, &n2); |
a33c29bc | 4647 | } |
227968da | 4648 | f2 = foldbuf2; |
18f762c3 | 4649 | } |
48ef279e | 4650 | } |
8b35872c | 4651 | |
5001101e | 4652 | /* Here f1 and f2 point to the beginning of the strings to compare. |
227968da KW |
4653 | * These strings are the folds of the next character from each input |
4654 | * string, stored in utf8. */ | |
5e64d0fa | 4655 | |
48ef279e KW |
4656 | /* While there is more to look for in both folds, see if they |
4657 | * continue to match */ | |
4658 | while (n1 && n2) { | |
4659 | U8 fold_length = UTF8SKIP(f1); | |
4660 | if (fold_length != UTF8SKIP(f2) | |
4661 | || (fold_length == 1 && *f1 != *f2) /* Short circuit memNE | |
4662 | function call for single | |
a6d5f321 | 4663 | byte */ |
48ef279e KW |
4664 | || memNE((char*)f1, (char*)f2, fold_length)) |
4665 | { | |
e6226b18 | 4666 | return 0; /* mismatch */ |
48ef279e KW |
4667 | } |
4668 | ||
4669 | /* Here, they matched, advance past them */ | |
4670 | n1 -= fold_length; | |
4671 | f1 += fold_length; | |
4672 | n2 -= fold_length; | |
4673 | f2 += fold_length; | |
4674 | } | |
8b35872c | 4675 | |
48ef279e KW |
4676 | /* When reach the end of any fold, advance the input past it */ |
4677 | if (n1 == 0) { | |
4678 | p1 += u1 ? UTF8SKIP(p1) : 1; | |
4679 | } | |
4680 | if (n2 == 0) { | |
4681 | p2 += u2 ? UTF8SKIP(p2) : 1; | |
4682 | } | |
8b35872c KW |
4683 | } /* End of loop through both strings */ |
4684 | ||
4685 | /* A match is defined by each scan that specified an explicit length | |
4686 | * reaching its final goal, and the other not having matched a partial | |
4687 | * character (which can happen when the fold of a character is more than one | |
4688 | * character). */ | |
4689 | if (! ((g1 == 0 || p1 == g1) && (g2 == 0 || p2 == g2)) || n1 || n2) { | |
e6226b18 | 4690 | return 0; |
8b35872c KW |
4691 | } |
4692 | ||
4693 | /* Successful match. Set output pointers */ | |
4694 | if (pe1) { | |
48ef279e | 4695 | *pe1 = (char*)p1; |
8b35872c KW |
4696 | } |
4697 | if (pe2) { | |
48ef279e | 4698 | *pe2 = (char*)p2; |
8b35872c | 4699 | } |
e6226b18 | 4700 | return 1; |
e6b2e755 | 4701 | } |
701a277b | 4702 | |
37e7596b KW |
4703 | /* XXX The next four functions should likely be moved to mathoms.c once all |
4704 | * occurrences of them are removed from the core; some cpan-upstream modules | |
4705 | * still use them */ | |
4706 | ||
4707 | U8 * | |
4708 | Perl_uvuni_to_utf8(pTHX_ U8 *d, UV uv) | |
4709 | { | |
4710 | PERL_ARGS_ASSERT_UVUNI_TO_UTF8; | |
4711 | ||
4712 | return Perl_uvoffuni_to_utf8_flags(aTHX_ d, uv, 0); | |
4713 | } | |
4714 | ||
4715 | UV | |
4716 | Perl_utf8n_to_uvuni(pTHX_ const U8 *s, STRLEN curlen, STRLEN *retlen, U32 flags) | |
4717 | { | |
4718 | PERL_ARGS_ASSERT_UTF8N_TO_UVUNI; | |
4719 | ||
4720 | return NATIVE_TO_UNI(utf8n_to_uvchr(s, curlen, retlen, flags)); | |
4721 | } | |
4722 | ||
4723 | /* | |
4724 | =for apidoc uvuni_to_utf8_flags | |
4725 | ||
4726 | Instead you almost certainly want to use L</uvchr_to_utf8> or | |
4727 | L</uvchr_to_utf8_flags>>. | |
4728 | ||
4729 | This function is a deprecated synonym for L</uvoffuni_to_utf8_flags>, | |
4730 | which itself, while not deprecated, should be used only in isolated | |
4731 | circumstances. These functions were useful for code that wanted to handle | |
4732 | both EBCDIC and ASCII platforms with Unicode properties, but starting in Perl | |
4733 | v5.20, the distinctions between the platforms have mostly been made invisible | |
4734 | to most code, so this function is quite unlikely to be what you want. | |
4735 | ||
4736 | =cut | |
4737 | */ | |
4738 | ||
4739 | U8 * | |
4740 | Perl_uvuni_to_utf8_flags(pTHX_ U8 *d, UV uv, UV flags) | |
4741 | { | |
4742 | PERL_ARGS_ASSERT_UVUNI_TO_UTF8_FLAGS; | |
4743 | ||
4744 | return uvoffuni_to_utf8_flags(d, uv, flags); | |
4745 | } | |
4746 | ||
4747 | /* | |
4748 | =for apidoc utf8n_to_uvuni | |
4749 | ||
4750 | Instead use L</utf8_to_uvchr_buf>, or rarely, L</utf8n_to_uvchr>. | |
4751 | ||
e2660c54 | 4752 | This function was useful for code that wanted to handle both EBCDIC and |
37e7596b KW |
4753 | ASCII platforms with Unicode properties, but starting in Perl v5.20, the |
4754 | distinctions between the platforms have mostly been made invisible to most | |
e2660c54 KW |
4755 | code, so this function is quite unlikely to be what you want. If you do need |
4756 | this precise functionality, use instead | |
4757 | C<L<NATIVE_TO_UNI(utf8_to_uvchr_buf(...))|/utf8_to_uvchr_buf>> | |
4758 | or C<L<NATIVE_TO_UNI(utf8n_to_uvchr(...))|/utf8n_to_uvchr>>. | |
37e7596b KW |
4759 | |
4760 | =cut | |
4761 | */ | |
4762 | ||
a49f32c6 NC |
4763 | /* |
4764 | * Local variables: | |
4765 | * c-indentation-style: bsd | |
4766 | * c-basic-offset: 4 | |
14d04a33 | 4767 | * indent-tabs-mode: nil |
a49f32c6 NC |
4768 | * End: |
4769 | * | |
14d04a33 | 4770 | * ex: set ts=8 sts=4 sw=4 et: |
37442d52 | 4771 | */ |