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" |
b992490d | 34 | #include "invlist_inline.h" |
a0ed51b3 | 35 | |
806547a7 | 36 | static const char malformed_text[] = "Malformed UTF-8 character"; |
27da23d5 | 37 | static const char unees[] = |
806547a7 | 38 | "Malformed UTF-8 character (unexpected end of string)"; |
fb7e7255 KW |
39 | |
40 | /* Be sure to synchronize this message with the similar one in regcomp.c */ | |
760c7c2f | 41 | static const char cp_above_legal_max[] = |
d22ec717 KW |
42 | "Use of code point 0x%" UVXf " is not allowed; the" |
43 | " permissible max is 0x%" UVXf; | |
760c7c2f | 44 | |
d22ec717 | 45 | #define MAX_EXTERNALLY_LEGAL_CP ((UV) (IV_MAX)) |
901b21bf | 46 | |
48ef279e | 47 | /* |
ccfc67b7 | 48 | =head1 Unicode Support |
7fefc6c1 | 49 | These are various utility functions for manipulating UTF8-encoded |
72d33970 | 50 | strings. For the uninitiated, this is a method of representing arbitrary |
61296642 | 51 | Unicode characters as a variable number of bytes, in such a way that |
56da48f7 DM |
52 | characters in the ASCII range are unmodified, and a zero byte never appears |
53 | within non-zero characters. | |
166f8a29 | 54 | |
eaf7a4d2 CS |
55 | =cut |
56 | */ | |
57 | ||
9cbfb8ab KW |
58 | void |
59 | Perl__force_out_malformed_utf8_message(pTHX_ | |
60 | const U8 *const p, /* First byte in UTF-8 sequence */ | |
61 | const U8 * const e, /* Final byte in sequence (may include | |
62 | multiple chars */ | |
63 | const U32 flags, /* Flags to pass to utf8n_to_uvchr(), | |
64 | usually 0, or some DISALLOW flags */ | |
65 | const bool die_here) /* If TRUE, this function does not return */ | |
66 | { | |
67 | /* This core-only function is to be called when a malformed UTF-8 character | |
68 | * is found, in order to output the detailed information about the | |
69 | * malformation before dieing. The reason it exists is for the occasions | |
70 | * when such a malformation is fatal, but warnings might be turned off, so | |
71 | * that normally they would not be actually output. This ensures that they | |
72 | * do get output. Because a sequence may be malformed in more than one | |
73 | * way, multiple messages may be generated, so we can't make them fatal, as | |
74 | * that would cause the first one to die. | |
75 | * | |
76 | * Instead we pretend -W was passed to perl, then die afterwards. The | |
77 | * flexibility is here to return to the caller so they can finish up and | |
78 | * die themselves */ | |
79 | U32 errors; | |
80 | ||
81 | PERL_ARGS_ASSERT__FORCE_OUT_MALFORMED_UTF8_MESSAGE; | |
82 | ||
83 | ENTER; | |
c15a80f3 | 84 | SAVEI8(PL_dowarn); |
9cbfb8ab KW |
85 | SAVESPTR(PL_curcop); |
86 | ||
87 | PL_dowarn = G_WARN_ALL_ON|G_WARN_ON; | |
88 | if (PL_curcop) { | |
89 | PL_curcop->cop_warnings = pWARN_ALL; | |
90 | } | |
91 | ||
92 | (void) utf8n_to_uvchr_error(p, e - p, NULL, flags & ~UTF8_CHECK_ONLY, &errors); | |
93 | ||
94 | LEAVE; | |
95 | ||
96 | if (! errors) { | |
97 | Perl_croak(aTHX_ "panic: _force_out_malformed_utf8_message should" | |
98 | " be called only when there are errors found"); | |
99 | } | |
100 | ||
101 | if (die_here) { | |
102 | Perl_croak(aTHX_ "Malformed UTF-8 character (fatal)"); | |
103 | } | |
104 | } | |
105 | ||
bb07812e KW |
106 | STATIC HV * |
107 | S_new_msg_hv(pTHX_ const char * const message, /* The message text */ | |
108 | U32 categories, /* Packed warning categories */ | |
109 | U32 flag) /* Flag associated with this message */ | |
110 | { | |
111 | /* Creates, populates, and returns an HV* that describes an error message | |
112 | * for the translators between UTF8 and code point */ | |
113 | ||
114 | SV* msg_sv = newSVpv(message, 0); | |
115 | SV* category_sv = newSVuv(categories); | |
116 | SV* flag_bit_sv = newSVuv(flag); | |
117 | ||
118 | HV* msg_hv = newHV(); | |
119 | ||
120 | PERL_ARGS_ASSERT_NEW_MSG_HV; | |
121 | ||
2b672cf5 KW |
122 | (void) hv_stores(msg_hv, "text", msg_sv); |
123 | (void) hv_stores(msg_hv, "warn_categories", category_sv); | |
124 | (void) hv_stores(msg_hv, "flag_bit", flag_bit_sv); | |
bb07812e KW |
125 | |
126 | return msg_hv; | |
127 | } | |
128 | ||
eaf7a4d2 | 129 | /* |
378516de | 130 | =for apidoc uvoffuni_to_utf8_flags |
eebe1485 | 131 | |
a27992cc | 132 | THIS FUNCTION SHOULD BE USED IN ONLY VERY SPECIALIZED CIRCUMSTANCES. |
de69f3af KW |
133 | Instead, B<Almost all code should use L</uvchr_to_utf8> or |
134 | L</uvchr_to_utf8_flags>>. | |
a27992cc | 135 | |
de69f3af KW |
136 | This function is like them, but the input is a strict Unicode |
137 | (as opposed to native) code point. Only in very rare circumstances should code | |
138 | not be using the native code point. | |
949cf498 | 139 | |
efa9cd84 | 140 | For details, see the description for L</uvchr_to_utf8_flags>. |
949cf498 | 141 | |
eebe1485 SC |
142 | =cut |
143 | */ | |
144 | ||
33f38593 KW |
145 | U8 * |
146 | Perl_uvoffuni_to_utf8_flags(pTHX_ U8 *d, UV uv, const UV flags) | |
147 | { | |
148 | PERL_ARGS_ASSERT_UVOFFUNI_TO_UTF8_FLAGS; | |
149 | ||
150 | return uvoffuni_to_utf8_flags_msgs(d, uv, flags, NULL); | |
151 | } | |
152 | ||
c94c2f39 KW |
153 | /* All these formats take a single UV code point argument */ |
154 | const char surrogate_cp_format[] = "UTF-16 surrogate U+%04" UVXf; | |
155 | const char nonchar_cp_format[] = "Unicode non-character U+%04" UVXf | |
156 | " is not recommended for open interchange"; | |
157 | const char super_cp_format[] = "Code point 0x%" UVXf " is not Unicode," | |
158 | " may not be portable"; | |
57ff5f59 KW |
159 | const char perl_extended_cp_format[] = "Code point 0x%" UVXf " is not" \ |
160 | " Unicode, requires a Perl extension," \ | |
161 | " and so is not portable"; | |
c94c2f39 | 162 | |
33f38593 | 163 | #define HANDLE_UNICODE_SURROGATE(uv, flags, msgs) \ |
8ee1cdcb KW |
164 | STMT_START { \ |
165 | if (flags & UNICODE_WARN_SURROGATE) { \ | |
33f38593 KW |
166 | U32 category = packWARN(WARN_SURROGATE); \ |
167 | const char * format = surrogate_cp_format; \ | |
168 | if (msgs) { \ | |
169 | *msgs = new_msg_hv(Perl_form(aTHX_ format, uv), \ | |
170 | category, \ | |
171 | UNICODE_GOT_SURROGATE); \ | |
172 | } \ | |
173 | else { \ | |
174 | Perl_ck_warner_d(aTHX_ category, format, uv); \ | |
175 | } \ | |
8ee1cdcb KW |
176 | } \ |
177 | if (flags & UNICODE_DISALLOW_SURROGATE) { \ | |
178 | return NULL; \ | |
179 | } \ | |
180 | } STMT_END; | |
181 | ||
33f38593 | 182 | #define HANDLE_UNICODE_NONCHAR(uv, flags, msgs) \ |
8ee1cdcb KW |
183 | STMT_START { \ |
184 | if (flags & UNICODE_WARN_NONCHAR) { \ | |
33f38593 KW |
185 | U32 category = packWARN(WARN_NONCHAR); \ |
186 | const char * format = nonchar_cp_format; \ | |
187 | if (msgs) { \ | |
188 | *msgs = new_msg_hv(Perl_form(aTHX_ format, uv), \ | |
189 | category, \ | |
190 | UNICODE_GOT_NONCHAR); \ | |
191 | } \ | |
192 | else { \ | |
193 | Perl_ck_warner_d(aTHX_ category, format, uv); \ | |
194 | } \ | |
8ee1cdcb KW |
195 | } \ |
196 | if (flags & UNICODE_DISALLOW_NONCHAR) { \ | |
197 | return NULL; \ | |
198 | } \ | |
199 | } STMT_END; | |
200 | ||
ba6ed43c KW |
201 | /* Use shorter names internally in this file */ |
202 | #define SHIFT UTF_ACCUMULATION_SHIFT | |
203 | #undef MARK | |
204 | #define MARK UTF_CONTINUATION_MARK | |
205 | #define MASK UTF_CONTINUATION_MASK | |
206 | ||
33f38593 KW |
207 | /* |
208 | =for apidoc uvchr_to_utf8_flags_msgs | |
209 | ||
210 | THIS FUNCTION SHOULD BE USED IN ONLY VERY SPECIALIZED CIRCUMSTANCES. | |
211 | ||
212 | Most code should use C<L</uvchr_to_utf8_flags>()> rather than call this directly. | |
213 | ||
214 | This function is for code that wants any warning and/or error messages to be | |
215 | returned to the caller rather than be displayed. All messages that would have | |
216 | been displayed if all lexcial warnings are enabled will be returned. | |
217 | ||
218 | It is just like C<L</uvchr_to_utf8_flags>> but it takes an extra parameter | |
219 | placed after all the others, C<msgs>. If this parameter is 0, this function | |
220 | behaves identically to C<L</uvchr_to_utf8_flags>>. Otherwise, C<msgs> should | |
221 | be a pointer to an C<HV *> variable, in which this function creates a new HV to | |
222 | contain any appropriate messages. The hash has three key-value pairs, as | |
223 | follows: | |
224 | ||
225 | =over 4 | |
226 | ||
227 | =item C<text> | |
228 | ||
229 | The text of the message as a C<SVpv>. | |
230 | ||
231 | =item C<warn_categories> | |
232 | ||
233 | The warning category (or categories) packed into a C<SVuv>. | |
234 | ||
235 | =item C<flag> | |
236 | ||
237 | A single flag bit associated with this message, in a C<SVuv>. | |
238 | The bit corresponds to some bit in the C<*errors> return value, | |
239 | such as C<UNICODE_GOT_SURROGATE>. | |
240 | ||
241 | =back | |
242 | ||
243 | It's important to note that specifying this parameter as non-null will cause | |
244 | any warnings this function would otherwise generate to be suppressed, and | |
245 | instead be placed in C<*msgs>. The caller can check the lexical warnings state | |
246 | (or not) when choosing what to do with the returned messages. | |
247 | ||
248 | The caller, of course, is responsible for freeing any returned HV. | |
249 | ||
250 | =cut | |
251 | */ | |
252 | ||
253 | /* Undocumented; we don't want people using this. Instead they should use | |
254 | * uvchr_to_utf8_flags_msgs() */ | |
dfe13c55 | 255 | U8 * |
33f38593 | 256 | Perl_uvoffuni_to_utf8_flags_msgs(pTHX_ U8 *d, UV uv, const UV flags, HV** msgs) |
a0ed51b3 | 257 | { |
33f38593 KW |
258 | PERL_ARGS_ASSERT_UVOFFUNI_TO_UTF8_FLAGS_MSGS; |
259 | ||
260 | if (msgs) { | |
261 | *msgs = NULL; | |
262 | } | |
7918f24d | 263 | |
2d1545e5 | 264 | if (OFFUNI_IS_INVARIANT(uv)) { |
4c8cd605 | 265 | *d++ = LATIN1_TO_NATIVE(uv); |
d9432125 KW |
266 | return d; |
267 | } | |
facc1dc2 | 268 | |
3ea68d71 | 269 | if (uv <= MAX_UTF8_TWO_BYTE) { |
facc1dc2 KW |
270 | *d++ = I8_TO_NATIVE_UTF8(( uv >> SHIFT) | UTF_START_MARK(2)); |
271 | *d++ = I8_TO_NATIVE_UTF8(( uv & MASK) | MARK); | |
3ea68d71 KW |
272 | return d; |
273 | } | |
d9432125 | 274 | |
ba6ed43c KW |
275 | /* Not 2-byte; test for and handle 3-byte result. In the test immediately |
276 | * below, the 16 is for start bytes E0-EF (which are all the possible ones | |
277 | * for 3 byte characters). The 2 is for 2 continuation bytes; these each | |
278 | * contribute SHIFT bits. This yields 0x4000 on EBCDIC platforms, 0x1_0000 | |
279 | * on ASCII; so 3 bytes covers the range 0x400-0x3FFF on EBCDIC; | |
280 | * 0x800-0xFFFF on ASCII */ | |
281 | if (uv < (16 * (1U << (2 * SHIFT)))) { | |
282 | *d++ = I8_TO_NATIVE_UTF8(( uv >> ((3 - 1) * SHIFT)) | UTF_START_MARK(3)); | |
283 | *d++ = I8_TO_NATIVE_UTF8(((uv >> ((2 - 1) * SHIFT)) & MASK) | MARK); | |
284 | *d++ = I8_TO_NATIVE_UTF8(( uv /* (1 - 1) */ & MASK) | MARK); | |
285 | ||
286 | #ifndef EBCDIC /* These problematic code points are 4 bytes on EBCDIC, so | |
287 | aren't tested here */ | |
288 | /* The most likely code points in this range are below the surrogates. | |
289 | * Do an extra test to quickly exclude those. */ | |
290 | if (UNLIKELY(uv >= UNICODE_SURROGATE_FIRST)) { | |
291 | if (UNLIKELY( UNICODE_IS_32_CONTIGUOUS_NONCHARS(uv) | |
292 | || UNICODE_IS_END_PLANE_NONCHAR_GIVEN_NOT_SUPER(uv))) | |
293 | { | |
33f38593 | 294 | HANDLE_UNICODE_NONCHAR(uv, flags, msgs); |
8ee1cdcb KW |
295 | } |
296 | else if (UNLIKELY(UNICODE_IS_SURROGATE(uv))) { | |
33f38593 | 297 | HANDLE_UNICODE_SURROGATE(uv, flags, msgs); |
760c7c2f | 298 | } |
ba6ed43c KW |
299 | } |
300 | #endif | |
301 | return d; | |
302 | } | |
303 | ||
304 | /* Not 3-byte; that means the code point is at least 0x1_0000 on ASCII | |
305 | * platforms, and 0x4000 on EBCDIC. There are problematic cases that can | |
306 | * happen starting with 4-byte characters on ASCII platforms. We unify the | |
307 | * code for these with EBCDIC, even though some of them require 5-bytes on | |
308 | * those, because khw believes the code saving is worth the very slight | |
309 | * performance hit on these high EBCDIC code points. */ | |
310 | ||
311 | if (UNLIKELY(UNICODE_IS_SUPER(uv))) { | |
d22ec717 KW |
312 | if (UNLIKELY(uv > MAX_EXTERNALLY_LEGAL_CP)) { |
313 | Perl_croak(aTHX_ cp_above_legal_max, uv, MAX_EXTERNALLY_LEGAL_CP); | |
a5bf80e0 | 314 | } |
33f38593 KW |
315 | if ( (flags & UNICODE_WARN_SUPER) |
316 | || ( (flags & UNICODE_WARN_PERL_EXTENDED) | |
0a8a1a5b | 317 | && UNICODE_IS_PERL_EXTENDED(uv))) |
a5bf80e0 | 318 | { |
33f38593 KW |
319 | const char * format = super_cp_format; |
320 | U32 category = packWARN(WARN_NON_UNICODE); | |
321 | U32 flag = UNICODE_GOT_SUPER; | |
322 | ||
323 | /* Choose the more dire applicable warning */ | |
324 | if (UNICODE_IS_PERL_EXTENDED(uv)) { | |
325 | format = perl_extended_cp_format; | |
326 | if (flags & (UNICODE_WARN_PERL_EXTENDED | |
327 | |UNICODE_DISALLOW_PERL_EXTENDED)) | |
328 | { | |
329 | flag = UNICODE_GOT_PERL_EXTENDED; | |
330 | } | |
331 | } | |
a5bf80e0 | 332 | |
33f38593 KW |
333 | if (msgs) { |
334 | *msgs = new_msg_hv(Perl_form(aTHX_ format, uv), | |
335 | category, flag); | |
336 | } | |
337 | else { | |
338 | Perl_ck_warner_d(aTHX_ packWARN(WARN_NON_UNICODE), format, uv); | |
339 | } | |
a5bf80e0 | 340 | } |
56576a04 | 341 | if ( (flags & UNICODE_DISALLOW_SUPER) |
0a8a1a5b KW |
342 | || ( (flags & UNICODE_DISALLOW_PERL_EXTENDED) |
343 | && UNICODE_IS_PERL_EXTENDED(uv))) | |
a5bf80e0 KW |
344 | { |
345 | return NULL; | |
346 | } | |
347 | } | |
ba6ed43c | 348 | else if (UNLIKELY(UNICODE_IS_END_PLANE_NONCHAR_GIVEN_NOT_SUPER(uv))) { |
33f38593 | 349 | HANDLE_UNICODE_NONCHAR(uv, flags, msgs); |
507b9800 | 350 | } |
d9432125 | 351 | |
ba6ed43c KW |
352 | /* Test for and handle 4-byte result. In the test immediately below, the |
353 | * 8 is for start bytes F0-F7 (which are all the possible ones for 4 byte | |
354 | * characters). The 3 is for 3 continuation bytes; these each contribute | |
355 | * SHIFT bits. This yields 0x4_0000 on EBCDIC platforms, 0x20_0000 on | |
356 | * ASCII, so 4 bytes covers the range 0x4000-0x3_FFFF on EBCDIC; | |
357 | * 0x1_0000-0x1F_FFFF on ASCII */ | |
358 | if (uv < (8 * (1U << (3 * SHIFT)))) { | |
359 | *d++ = I8_TO_NATIVE_UTF8(( uv >> ((4 - 1) * SHIFT)) | UTF_START_MARK(4)); | |
360 | *d++ = I8_TO_NATIVE_UTF8(((uv >> ((3 - 1) * SHIFT)) & MASK) | MARK); | |
361 | *d++ = I8_TO_NATIVE_UTF8(((uv >> ((2 - 1) * SHIFT)) & MASK) | MARK); | |
362 | *d++ = I8_TO_NATIVE_UTF8(( uv /* (1 - 1) */ & MASK) | MARK); | |
363 | ||
364 | #ifdef EBCDIC /* These were handled on ASCII platforms in the code for 3-byte | |
365 | characters. The end-plane non-characters for EBCDIC were | |
366 | handled just above */ | |
367 | if (UNLIKELY(UNICODE_IS_32_CONTIGUOUS_NONCHARS(uv))) { | |
33f38593 | 368 | HANDLE_UNICODE_NONCHAR(uv, flags, msgs); |
d528804a | 369 | } |
ba6ed43c | 370 | else if (UNLIKELY(UNICODE_IS_SURROGATE(uv))) { |
33f38593 | 371 | HANDLE_UNICODE_SURROGATE(uv, flags, msgs); |
ba6ed43c KW |
372 | } |
373 | #endif | |
374 | ||
375 | return d; | |
376 | } | |
377 | ||
378 | /* Not 4-byte; that means the code point is at least 0x20_0000 on ASCII | |
379 | * platforms, and 0x4000 on EBCDIC. At this point we switch to a loop | |
380 | * format. The unrolled version above turns out to not save all that much | |
381 | * time, and at these high code points (well above the legal Unicode range | |
382 | * on ASCII platforms, and well above anything in common use in EBCDIC), | |
383 | * khw believes that less code outweighs slight performance gains. */ | |
384 | ||
d9432125 | 385 | { |
5aaebcb3 | 386 | STRLEN len = OFFUNISKIP(uv); |
1d72bdf6 NIS |
387 | U8 *p = d+len-1; |
388 | while (p > d) { | |
4c8cd605 | 389 | *p-- = I8_TO_NATIVE_UTF8((uv & UTF_CONTINUATION_MASK) | UTF_CONTINUATION_MARK); |
1d72bdf6 NIS |
390 | uv >>= UTF_ACCUMULATION_SHIFT; |
391 | } | |
4c8cd605 | 392 | *p = I8_TO_NATIVE_UTF8((uv & UTF_START_MASK(len)) | UTF_START_MARK(len)); |
1d72bdf6 NIS |
393 | return d+len; |
394 | } | |
a0ed51b3 | 395 | } |
a5bf80e0 | 396 | |
646ca15d | 397 | /* |
07693fe6 KW |
398 | =for apidoc uvchr_to_utf8 |
399 | ||
bcb1a2d4 | 400 | Adds the UTF-8 representation of the native code point C<uv> to the end |
f2fc1b45 | 401 | of the string C<d>; C<d> should have at least C<UVCHR_SKIP(uv)+1> (up to |
c749c9fd KW |
402 | C<UTF8_MAXBYTES+1>) free bytes available. The return value is the pointer to |
403 | the byte after the end of the new character. In other words, | |
07693fe6 KW |
404 | |
405 | d = uvchr_to_utf8(d, uv); | |
406 | ||
407 | is the recommended wide native character-aware way of saying | |
408 | ||
409 | *(d++) = uv; | |
410 | ||
d22ec717 KW |
411 | This function accepts any code point from 0..C<IV_MAX> as input. |
412 | C<IV_MAX> is typically 0x7FFF_FFFF in a 32-bit word. | |
760c7c2f KW |
413 | |
414 | It is possible to forbid or warn on non-Unicode code points, or those that may | |
415 | be problematic by using L</uvchr_to_utf8_flags>. | |
de69f3af | 416 | |
07693fe6 KW |
417 | =cut |
418 | */ | |
419 | ||
de69f3af KW |
420 | /* This is also a macro */ |
421 | PERL_CALLCONV U8* Perl_uvchr_to_utf8(pTHX_ U8 *d, UV uv); | |
422 | ||
07693fe6 KW |
423 | U8 * |
424 | Perl_uvchr_to_utf8(pTHX_ U8 *d, UV uv) | |
425 | { | |
de69f3af | 426 | return uvchr_to_utf8(d, uv); |
07693fe6 KW |
427 | } |
428 | ||
de69f3af KW |
429 | /* |
430 | =for apidoc uvchr_to_utf8_flags | |
431 | ||
432 | Adds the UTF-8 representation of the native code point C<uv> to the end | |
f2fc1b45 | 433 | of the string C<d>; C<d> should have at least C<UVCHR_SKIP(uv)+1> (up to |
c749c9fd KW |
434 | C<UTF8_MAXBYTES+1>) free bytes available. The return value is the pointer to |
435 | the byte after the end of the new character. In other words, | |
de69f3af KW |
436 | |
437 | d = uvchr_to_utf8_flags(d, uv, flags); | |
438 | ||
439 | or, in most cases, | |
440 | ||
441 | d = uvchr_to_utf8_flags(d, uv, 0); | |
442 | ||
443 | This is the Unicode-aware way of saying | |
444 | ||
445 | *(d++) = uv; | |
446 | ||
d22ec717 KW |
447 | If C<flags> is 0, this function accepts any code point from 0..C<IV_MAX> as |
448 | input. C<IV_MAX> is typically 0x7FFF_FFFF in a 32-bit word. | |
760c7c2f KW |
449 | |
450 | Specifying C<flags> can further restrict what is allowed and not warned on, as | |
451 | follows: | |
de69f3af | 452 | |
796b6530 | 453 | If C<uv> is a Unicode surrogate code point and C<UNICODE_WARN_SURROGATE> is set, |
7ee537e6 KW |
454 | the function will raise a warning, provided UTF8 warnings are enabled. If |
455 | instead C<UNICODE_DISALLOW_SURROGATE> is set, the function will fail and return | |
456 | NULL. If both flags are set, the function will both warn and return NULL. | |
de69f3af | 457 | |
760c7c2f KW |
458 | Similarly, the C<UNICODE_WARN_NONCHAR> and C<UNICODE_DISALLOW_NONCHAR> flags |
459 | affect how the function handles a Unicode non-character. | |
93e6dbd6 | 460 | |
760c7c2f KW |
461 | And likewise, the C<UNICODE_WARN_SUPER> and C<UNICODE_DISALLOW_SUPER> flags |
462 | affect the handling of code points that are above the Unicode maximum of | |
463 | 0x10FFFF. Languages other than Perl may not be able to accept files that | |
464 | contain these. | |
93e6dbd6 KW |
465 | |
466 | The flag C<UNICODE_WARN_ILLEGAL_INTERCHANGE> selects all three of | |
467 | the above WARN flags; and C<UNICODE_DISALLOW_ILLEGAL_INTERCHANGE> selects all | |
ecc1615f KW |
468 | three DISALLOW flags. C<UNICODE_DISALLOW_ILLEGAL_INTERCHANGE> restricts the |
469 | allowed inputs to the strict UTF-8 traditionally defined by Unicode. | |
470 | Similarly, C<UNICODE_WARN_ILLEGAL_C9_INTERCHANGE> and | |
471 | C<UNICODE_DISALLOW_ILLEGAL_C9_INTERCHANGE> are shortcuts to select the | |
472 | above-Unicode and surrogate flags, but not the non-character ones, as | |
473 | defined in | |
474 | L<Unicode Corrigendum #9|http://www.unicode.org/versions/corrigendum9.html>. | |
475 | See L<perlunicode/Noncharacter code points>. | |
93e6dbd6 | 476 | |
57ff5f59 KW |
477 | Extremely high code points were never specified in any standard, and require an |
478 | extension to UTF-8 to express, which Perl does. It is likely that programs | |
479 | written in something other than Perl would not be able to read files that | |
480 | contain these; nor would Perl understand files written by something that uses a | |
481 | different extension. For these reasons, there is a separate set of flags that | |
482 | can warn and/or disallow these extremely high code points, even if other | |
483 | above-Unicode ones are accepted. They are the C<UNICODE_WARN_PERL_EXTENDED> | |
484 | and C<UNICODE_DISALLOW_PERL_EXTENDED> flags. For more information see | |
485 | L</C<UTF8_GOT_PERL_EXTENDED>>. Of course C<UNICODE_DISALLOW_SUPER> will | |
486 | treat all above-Unicode code points, including these, as malformations. (Note | |
487 | that the Unicode standard considers anything above 0x10FFFF to be illegal, but | |
488 | there are standards predating it that allow up to 0x7FFF_FFFF (2**31 -1)) | |
489 | ||
490 | A somewhat misleadingly named synonym for C<UNICODE_WARN_PERL_EXTENDED> is | |
491 | retained for backward compatibility: C<UNICODE_WARN_ABOVE_31_BIT>. Similarly, | |
492 | C<UNICODE_DISALLOW_ABOVE_31_BIT> is usable instead of the more accurately named | |
7c4a22ed KW |
493 | C<UNICODE_DISALLOW_PERL_EXTENDED>. The names are misleading because on EBCDIC |
494 | platforms,these flags can apply to code points that actually do fit in 31 bits. | |
495 | The new names accurately describe the situation in all cases. | |
de69f3af | 496 | |
de69f3af KW |
497 | =cut |
498 | */ | |
499 | ||
500 | /* This is also a macro */ | |
501 | PERL_CALLCONV U8* Perl_uvchr_to_utf8_flags(pTHX_ U8 *d, UV uv, UV flags); | |
502 | ||
07693fe6 KW |
503 | U8 * |
504 | Perl_uvchr_to_utf8_flags(pTHX_ U8 *d, UV uv, UV flags) | |
505 | { | |
de69f3af | 506 | return uvchr_to_utf8_flags(d, uv, flags); |
07693fe6 KW |
507 | } |
508 | ||
57ff5f59 KW |
509 | #ifndef UV_IS_QUAD |
510 | ||
e050c007 KW |
511 | STATIC int |
512 | S_is_utf8_cp_above_31_bits(const U8 * const s, | |
513 | const U8 * const e, | |
514 | const bool consider_overlongs) | |
83dc0f42 KW |
515 | { |
516 | /* Returns TRUE if the first code point represented by the Perl-extended- | |
517 | * UTF-8-encoded string starting at 's', and looking no further than 'e - | |
518 | * 1' doesn't fit into 31 bytes. That is, that if it is >= 2**31. | |
519 | * | |
520 | * The function handles the case where the input bytes do not include all | |
521 | * the ones necessary to represent a full character. That is, they may be | |
522 | * the intial bytes of the representation of a code point, but possibly | |
523 | * the final ones necessary for the complete representation may be beyond | |
524 | * 'e - 1'. | |
525 | * | |
e050c007 KW |
526 | * The function also can handle the case where the input is an overlong |
527 | * sequence. If 'consider_overlongs' is 0, the function assumes the | |
528 | * input is not overlong, without checking, and will return based on that | |
529 | * assumption. If this parameter is 1, the function will go to the trouble | |
530 | * of figuring out if it actually evaluates to above or below 31 bits. | |
83dc0f42 | 531 | * |
e050c007 | 532 | * The sequence is otherwise assumed to be well-formed, without checking. |
83dc0f42 KW |
533 | */ |
534 | ||
e050c007 KW |
535 | const STRLEN len = e - s; |
536 | int is_overlong; | |
537 | ||
538 | PERL_ARGS_ASSERT_IS_UTF8_CP_ABOVE_31_BITS; | |
539 | ||
540 | assert(! UTF8_IS_INVARIANT(*s) && e > s); | |
541 | ||
83dc0f42 KW |
542 | #ifdef EBCDIC |
543 | ||
e050c007 | 544 | PERL_UNUSED_ARG(consider_overlongs); |
83dc0f42 | 545 | |
e050c007 KW |
546 | /* On the EBCDIC code pages we handle, only the native start byte 0xFE can |
547 | * mean a 32-bit or larger code point (0xFF is an invariant). 0xFE can | |
548 | * also be the start byte for a 31-bit code point; we need at least 2 | |
549 | * bytes, and maybe up through 8 bytes, to determine that. (It can also be | |
550 | * the start byte for an overlong sequence, but for 30-bit or smaller code | |
551 | * points, so we don't have to worry about overlongs on EBCDIC.) */ | |
552 | if (*s != 0xFE) { | |
553 | return 0; | |
554 | } | |
83dc0f42 | 555 | |
e050c007 KW |
556 | if (len == 1) { |
557 | return -1; | |
558 | } | |
83dc0f42 | 559 | |
e050c007 | 560 | #else |
83dc0f42 | 561 | |
e050c007 KW |
562 | /* On ASCII, FE and FF are the only start bytes that can evaluate to |
563 | * needing more than 31 bits. */ | |
564 | if (LIKELY(*s < 0xFE)) { | |
565 | return 0; | |
566 | } | |
83dc0f42 | 567 | |
e050c007 KW |
568 | /* What we have left are FE and FF. Both of these require more than 31 |
569 | * bits unless they are for overlongs. */ | |
570 | if (! consider_overlongs) { | |
571 | return 1; | |
572 | } | |
83dc0f42 | 573 | |
e050c007 KW |
574 | /* Here, we have FE or FF. If the input isn't overlong, it evaluates to |
575 | * above 31 bits. But we need more than one byte to discern this, so if | |
576 | * passed just the start byte, it could be an overlong evaluating to | |
577 | * smaller */ | |
578 | if (len == 1) { | |
579 | return -1; | |
580 | } | |
83dc0f42 | 581 | |
e050c007 KW |
582 | /* Having excluded len==1, and knowing that FE and FF are both valid start |
583 | * bytes, we can call the function below to see if the sequence is | |
584 | * overlong. (We don't need the full generality of the called function, | |
585 | * but for these huge code points, speed shouldn't be a consideration, and | |
586 | * the compiler does have enough information, since it's static to this | |
587 | * file, to optimize to just the needed parts.) */ | |
588 | is_overlong = is_utf8_overlong_given_start_byte_ok(s, len); | |
83dc0f42 | 589 | |
e050c007 KW |
590 | /* If it isn't overlong, more than 31 bits are required. */ |
591 | if (is_overlong == 0) { | |
592 | return 1; | |
593 | } | |
83dc0f42 | 594 | |
e050c007 KW |
595 | /* If it is indeterminate if it is overlong, return that */ |
596 | if (is_overlong < 0) { | |
597 | return -1; | |
598 | } | |
599 | ||
600 | /* Here is overlong. Such a sequence starting with FE is below 31 bits, as | |
601 | * the max it can be is 2**31 - 1 */ | |
602 | if (*s == 0xFE) { | |
603 | return 0; | |
83dc0f42 KW |
604 | } |
605 | ||
e050c007 KW |
606 | #endif |
607 | ||
608 | /* Here, ASCII and EBCDIC rejoin: | |
609 | * On ASCII: We have an overlong sequence starting with FF | |
610 | * On EBCDIC: We have a sequence starting with FE. */ | |
611 | ||
612 | { /* For C89, use a block so the declaration can be close to its use */ | |
613 | ||
614 | #ifdef EBCDIC | |
615 | ||
5f995336 KW |
616 | /* U+7FFFFFFF (2 ** 31 - 1) |
617 | * [0] [1] [2] [3] [4] [5] [6] [7] [8] [9] 10 11 12 13 | |
618 | * IBM-1047: \xFE\x41\x41\x41\x41\x41\x41\x42\x73\x73\x73\x73\x73\x73 | |
619 | * IBM-037: \xFE\x41\x41\x41\x41\x41\x41\x42\x72\x72\x72\x72\x72\x72 | |
620 | * POSIX-BC: \xFE\x41\x41\x41\x41\x41\x41\x42\x75\x75\x75\x75\x75\x75 | |
621 | * I8: \xFF\xA0\xA0\xA0\xA0\xA0\xA0\xA1\xBF\xBF\xBF\xBF\xBF\xBF | |
622 | * U+80000000 (2 ** 31): | |
623 | * IBM-1047: \xFE\x41\x41\x41\x41\x41\x41\x43\x41\x41\x41\x41\x41\x41 | |
624 | * IBM-037: \xFE\x41\x41\x41\x41\x41\x41\x43\x41\x41\x41\x41\x41\x41 | |
625 | * POSIX-BC: \xFE\x41\x41\x41\x41\x41\x41\x43\x41\x41\x41\x41\x41\x41 | |
626 | * I8: \xFF\xA0\xA0\xA0\xA0\xA0\xA0\xA2\xA0\xA0\xA0\xA0\xA0\xA0 | |
e050c007 KW |
627 | * |
628 | * and since we know that *s = \xfe, any continuation sequcence | |
629 | * following it that is gt the below is above 31 bits | |
630 | [0] [1] [2] [3] [4] [5] [6] */ | |
631 | const U8 conts_for_highest_30_bit[] = "\x41\x41\x41\x41\x41\x41\x42"; | |
632 | ||
633 | #else | |
634 | ||
635 | /* FF overlong for U+7FFFFFFF (2 ** 31 - 1) | |
636 | * ASCII: \xFF\x80\x80\x80\x80\x80\x80\x81\xBF\xBF\xBF\xBF\xBF | |
637 | * FF overlong for U+80000000 (2 ** 31): | |
638 | * ASCII: \xFF\x80\x80\x80\x80\x80\x80\x82\x80\x80\x80\x80\x80 | |
639 | * and since we know that *s = \xff, any continuation sequcence | |
640 | * following it that is gt the below is above 30 bits | |
641 | [0] [1] [2] [3] [4] [5] [6] */ | |
642 | const U8 conts_for_highest_30_bit[] = "\x80\x80\x80\x80\x80\x80\x81"; | |
5f995336 | 643 | |
83dc0f42 KW |
644 | |
645 | #endif | |
e050c007 KW |
646 | const STRLEN conts_len = sizeof(conts_for_highest_30_bit) - 1; |
647 | const STRLEN cmp_len = MIN(conts_len, len - 1); | |
648 | ||
649 | /* Now compare the continuation bytes in s with the ones we have | |
650 | * compiled in that are for the largest 30 bit code point. If we have | |
651 | * enough bytes available to determine the answer, or the bytes we do | |
652 | * have differ from them, we can compare the two to get a definitive | |
653 | * answer (Note that in UTF-EBCDIC, the two lowest possible | |
654 | * continuation bytes are \x41 and \x42.) */ | |
655 | if (cmp_len >= conts_len || memNE(s + 1, | |
656 | conts_for_highest_30_bit, | |
657 | cmp_len)) | |
658 | { | |
659 | return cBOOL(memGT(s + 1, conts_for_highest_30_bit, cmp_len)); | |
660 | } | |
83dc0f42 | 661 | |
e050c007 KW |
662 | /* Here, all the bytes we have are the same as the highest 30-bit code |
663 | * point, but we are missing so many bytes that we can't make the | |
664 | * determination */ | |
665 | return -1; | |
666 | } | |
83dc0f42 KW |
667 | } |
668 | ||
57ff5f59 KW |
669 | #endif |
670 | ||
d6be65ae | 671 | PERL_STATIC_INLINE int |
12a4bed3 KW |
672 | S_is_utf8_overlong_given_start_byte_ok(const U8 * const s, const STRLEN len) |
673 | { | |
d6be65ae KW |
674 | /* Returns an int indicating whether or not the UTF-8 sequence from 's' to |
675 | * 's' + 'len' - 1 is an overlong. It returns 1 if it is an overlong; 0 if | |
676 | * it isn't, and -1 if there isn't enough information to tell. This last | |
677 | * return value can happen if the sequence is incomplete, missing some | |
678 | * trailing bytes that would form a complete character. If there are | |
679 | * enough bytes to make a definitive decision, this function does so. | |
680 | * Usually 2 bytes sufficient. | |
681 | * | |
682 | * Overlongs can occur whenever the number of continuation bytes changes. | |
683 | * That means whenever the number of leading 1 bits in a start byte | |
684 | * increases from the next lower start byte. That happens for start bytes | |
685 | * C0, E0, F0, F8, FC, FE, and FF. On modern perls, the following illegal | |
686 | * start bytes have already been excluded, so don't need to be tested here; | |
12a4bed3 KW |
687 | * ASCII platforms: C0, C1 |
688 | * EBCDIC platforms C0, C1, C2, C3, C4, E0 | |
d6be65ae | 689 | */ |
12a4bed3 KW |
690 | |
691 | const U8 s0 = NATIVE_UTF8_TO_I8(s[0]); | |
692 | const U8 s1 = NATIVE_UTF8_TO_I8(s[1]); | |
693 | ||
694 | PERL_ARGS_ASSERT_IS_UTF8_OVERLONG_GIVEN_START_BYTE_OK; | |
695 | assert(len > 1 && UTF8_IS_START(*s)); | |
696 | ||
697 | /* Each platform has overlongs after the start bytes given above (expressed | |
698 | * in I8 for EBCDIC). What constitutes an overlong varies by platform, but | |
699 | * the logic is the same, except the E0 overlong has already been excluded | |
700 | * on EBCDIC platforms. The values below were found by manually | |
701 | * inspecting the UTF-8 patterns. See the tables in utf8.h and | |
702 | * utfebcdic.h. */ | |
703 | ||
704 | # ifdef EBCDIC | |
705 | # define F0_ABOVE_OVERLONG 0xB0 | |
706 | # define F8_ABOVE_OVERLONG 0xA8 | |
707 | # define FC_ABOVE_OVERLONG 0xA4 | |
708 | # define FE_ABOVE_OVERLONG 0xA2 | |
709 | # define FF_OVERLONG_PREFIX "\xfe\x41\x41\x41\x41\x41\x41\x41" | |
710 | /* I8(0xfe) is FF */ | |
711 | # else | |
712 | ||
713 | if (s0 == 0xE0 && UNLIKELY(s1 < 0xA0)) { | |
d6be65ae | 714 | return 1; |
12a4bed3 KW |
715 | } |
716 | ||
717 | # define F0_ABOVE_OVERLONG 0x90 | |
718 | # define F8_ABOVE_OVERLONG 0x88 | |
719 | # define FC_ABOVE_OVERLONG 0x84 | |
720 | # define FE_ABOVE_OVERLONG 0x82 | |
721 | # define FF_OVERLONG_PREFIX "\xff\x80\x80\x80\x80\x80\x80" | |
722 | # endif | |
723 | ||
724 | ||
725 | if ( (s0 == 0xF0 && UNLIKELY(s1 < F0_ABOVE_OVERLONG)) | |
726 | || (s0 == 0xF8 && UNLIKELY(s1 < F8_ABOVE_OVERLONG)) | |
727 | || (s0 == 0xFC && UNLIKELY(s1 < FC_ABOVE_OVERLONG)) | |
728 | || (s0 == 0xFE && UNLIKELY(s1 < FE_ABOVE_OVERLONG))) | |
729 | { | |
d6be65ae | 730 | return 1; |
12a4bed3 KW |
731 | } |
732 | ||
b0b342d4 | 733 | /* Check for the FF overlong */ |
d6be65ae | 734 | return isFF_OVERLONG(s, len); |
b0b342d4 KW |
735 | } |
736 | ||
8d6204cc | 737 | PERL_STATIC_INLINE int |
b0b342d4 KW |
738 | S_isFF_OVERLONG(const U8 * const s, const STRLEN len) |
739 | { | |
8d6204cc KW |
740 | /* Returns an int indicating whether or not the UTF-8 sequence from 's' to |
741 | * 'e' - 1 is an overlong beginning with \xFF. It returns 1 if it is; 0 if | |
742 | * it isn't, and -1 if there isn't enough information to tell. This last | |
743 | * return value can happen if the sequence is incomplete, missing some | |
744 | * trailing bytes that would form a complete character. If there are | |
745 | * enough bytes to make a definitive decision, this function does so. */ | |
746 | ||
b0b342d4 | 747 | PERL_ARGS_ASSERT_ISFF_OVERLONG; |
12a4bed3 | 748 | |
8d6204cc KW |
749 | /* To be an FF overlong, all the available bytes must match */ |
750 | if (LIKELY(memNE(s, FF_OVERLONG_PREFIX, | |
751 | MIN(len, sizeof(FF_OVERLONG_PREFIX) - 1)))) | |
752 | { | |
753 | return 0; | |
754 | } | |
755 | ||
756 | /* To be an FF overlong sequence, all the bytes in FF_OVERLONG_PREFIX must | |
757 | * be there; what comes after them doesn't matter. See tables in utf8.h, | |
b0b342d4 | 758 | * utfebcdic.h. */ |
8d6204cc KW |
759 | if (len >= sizeof(FF_OVERLONG_PREFIX) - 1) { |
760 | return 1; | |
761 | } | |
12a4bed3 | 762 | |
8d6204cc KW |
763 | /* The missing bytes could cause the result to go one way or the other, so |
764 | * the result is indeterminate */ | |
765 | return -1; | |
12a4bed3 KW |
766 | } |
767 | ||
d22ec717 | 768 | #if defined(UV_IS_QUAD) /* These assume IV_MAX is 2**63-1 */ |
a77c906e KW |
769 | # ifdef EBCDIC /* Actually is I8 */ |
770 | # define HIGHEST_REPRESENTABLE_UTF8 \ | |
d22ec717 | 771 | "\xFF\xA7\xBF\xBF\xBF\xBF\xBF\xBF\xBF\xBF\xBF\xBF\xBF\xBF" |
a77c906e KW |
772 | # else |
773 | # define HIGHEST_REPRESENTABLE_UTF8 \ | |
d22ec717 | 774 | "\xFF\x80\x87\xBF\xBF\xBF\xBF\xBF\xBF\xBF\xBF\xBF\xBF" |
a77c906e KW |
775 | # endif |
776 | #endif | |
777 | ||
c285bbc4 | 778 | PERL_STATIC_INLINE int |
e050c007 KW |
779 | S_does_utf8_overflow(const U8 * const s, |
780 | const U8 * e, | |
781 | const bool consider_overlongs) | |
a77c906e | 782 | { |
c285bbc4 | 783 | /* Returns an int indicating whether or not the UTF-8 sequence from 's' to |
d22ec717 KW |
784 | * 'e' - 1 would overflow an IV on this platform; that is if it represents |
785 | * a code point larger than the highest representable code point. It | |
786 | * returns 1 if it does overflow; 0 if it doesn't, and -1 if there isn't | |
787 | * enough information to tell. This last return value can happen if the | |
788 | * sequence is incomplete, missing some trailing bytes that would form a | |
789 | * complete character. If there are enough bytes to make a definitive | |
790 | * decision, this function does so. | |
c285bbc4 | 791 | * |
e050c007 KW |
792 | * If 'consider_overlongs' is TRUE, the function checks for the possibility |
793 | * that the sequence is an overlong that doesn't overflow. Otherwise, it | |
794 | * assumes the sequence is not an overlong. This can give different | |
795 | * results only on ASCII 32-bit platforms. | |
796 | * | |
c285bbc4 KW |
797 | * (For ASCII platforms, we could use memcmp() because we don't have to |
798 | * convert each byte to I8, but it's very rare input indeed that would | |
799 | * approach overflow, so the loop below will likely only get executed once.) | |
800 | * | |
801 | * 'e' - 1 must not be beyond a full character. */ | |
a77c906e | 802 | |
a77c906e KW |
803 | |
804 | PERL_ARGS_ASSERT_DOES_UTF8_OVERFLOW; | |
805 | assert(s <= e && s + UTF8SKIP(s) >= e); | |
806 | ||
d22ec717 KW |
807 | #if ! defined(UV_IS_QUAD) |
808 | ||
809 | return is_utf8_cp_above_31_bits(s, e, consider_overlongs); | |
810 | ||
811 | #else | |
812 | ||
813 | PERL_UNUSED_ARG(consider_overlongs); | |
814 | ||
815 | { | |
816 | const STRLEN len = e - s; | |
817 | const U8 *x; | |
818 | const U8 * y = (const U8 *) HIGHEST_REPRESENTABLE_UTF8; | |
819 | ||
820 | for (x = s; x < e; x++, y++) { | |
821 | ||
822 | if (UNLIKELY(NATIVE_UTF8_TO_I8(*x) == *y)) { | |
823 | continue; | |
824 | } | |
825 | ||
826 | /* If this byte is larger than the corresponding highest UTF-8 | |
827 | * byte, the sequence overflow; otherwise the byte is less than, | |
828 | * and so the sequence doesn't overflow */ | |
829 | return NATIVE_UTF8_TO_I8(*x) > *y; | |
830 | ||
831 | } | |
832 | ||
833 | /* Got to the end and all bytes are the same. If the input is a whole | |
834 | * character, it doesn't overflow. And if it is a partial character, | |
835 | * there's not enough information to tell */ | |
836 | if (len < sizeof(HIGHEST_REPRESENTABLE_UTF8) - 1) { | |
837 | return -1; | |
838 | } | |
839 | ||
840 | return 0; | |
841 | } | |
842 | ||
843 | #endif | |
844 | ||
845 | } | |
846 | ||
847 | #if 0 | |
848 | ||
849 | /* This is the portions of the above function that deal with UV_MAX instead of | |
850 | * IV_MAX. They are left here in case we want to combine them so that internal | |
851 | * uses can have larger code points. The only logic difference is that the | |
852 | * 32-bit EBCDIC platform is treate like the 64-bit, and the 32-bit ASCII has | |
853 | * different logic. | |
854 | */ | |
855 | ||
856 | /* Anything larger than this will overflow the word if it were converted into a UV */ | |
857 | #if defined(UV_IS_QUAD) | |
858 | # ifdef EBCDIC /* Actually is I8 */ | |
859 | # define HIGHEST_REPRESENTABLE_UTF8 \ | |
860 | "\xFF\xAF\xBF\xBF\xBF\xBF\xBF\xBF\xBF\xBF\xBF\xBF\xBF\xBF" | |
861 | # else | |
862 | # define HIGHEST_REPRESENTABLE_UTF8 \ | |
863 | "\xFF\x80\x8F\xBF\xBF\xBF\xBF\xBF\xBF\xBF\xBF\xBF\xBF" | |
864 | # endif | |
865 | #else /* 32-bit */ | |
866 | # ifdef EBCDIC | |
867 | # define HIGHEST_REPRESENTABLE_UTF8 \ | |
868 | "\xFF\xA0\xA0\xA0\xA0\xA0\xA0\xA3\xBF\xBF\xBF\xBF\xBF\xBF" | |
869 | # else | |
870 | # define HIGHEST_REPRESENTABLE_UTF8 "\xFE\x83\xBF\xBF\xBF\xBF\xBF" | |
871 | # endif | |
872 | #endif | |
873 | ||
a77c906e KW |
874 | #if ! defined(UV_IS_QUAD) && ! defined(EBCDIC) |
875 | ||
876 | /* On 32 bit ASCII machines, many overlongs that start with FF don't | |
877 | * overflow */ | |
e050c007 | 878 | if (consider_overlongs && isFF_OVERLONG(s, len) > 0) { |
c285bbc4 KW |
879 | |
880 | /* To be such an overlong, the first bytes of 's' must match | |
881 | * FF_OVERLONG_PREFIX, which is "\xff\x80\x80\x80\x80\x80\x80". If we | |
882 | * don't have any additional bytes available, the sequence, when | |
883 | * completed might or might not fit in 32 bits. But if we have that | |
884 | * next byte, we can tell for sure. If it is <= 0x83, then it does | |
885 | * fit. */ | |
886 | if (len <= sizeof(FF_OVERLONG_PREFIX) - 1) { | |
887 | return -1; | |
888 | } | |
889 | ||
890 | return s[sizeof(FF_OVERLONG_PREFIX) - 1] > 0x83; | |
a77c906e KW |
891 | } |
892 | ||
d22ec717 KW |
893 | /* Starting with the #else, the rest of the function is identical except |
894 | * 1. we need to move the 'len' declaration to be global to the function | |
895 | * 2. the endif move to just after the UNUSED_ARG. | |
896 | * An empty endif is given just below to satisfy the preprocessor | |
897 | */ | |
a77c906e KW |
898 | #endif |
899 | ||
d22ec717 | 900 | #endif |
a77c906e | 901 | |
12a4bed3 KW |
902 | #undef F0_ABOVE_OVERLONG |
903 | #undef F8_ABOVE_OVERLONG | |
904 | #undef FC_ABOVE_OVERLONG | |
905 | #undef FE_ABOVE_OVERLONG | |
906 | #undef FF_OVERLONG_PREFIX | |
907 | ||
35f8c9bd | 908 | STRLEN |
edc2c47a | 909 | Perl__is_utf8_char_helper(const U8 * const s, const U8 * e, const U32 flags) |
35f8c9bd | 910 | { |
2b479609 | 911 | STRLEN len; |
12a4bed3 | 912 | const U8 *x; |
35f8c9bd | 913 | |
2b479609 KW |
914 | /* A helper function that should not be called directly. |
915 | * | |
916 | * This function returns non-zero if the string beginning at 's' and | |
917 | * looking no further than 'e - 1' is well-formed Perl-extended-UTF-8 for a | |
918 | * code point; otherwise it returns 0. The examination stops after the | |
919 | * first code point in 's' is validated, not looking at the rest of the | |
920 | * input. If 'e' is such that there are not enough bytes to represent a | |
921 | * complete code point, this function will return non-zero anyway, if the | |
922 | * bytes it does have are well-formed UTF-8 as far as they go, and aren't | |
923 | * excluded by 'flags'. | |
924 | * | |
925 | * A non-zero return gives the number of bytes required to represent the | |
926 | * code point. Be aware that if the input is for a partial character, the | |
927 | * return will be larger than 'e - s'. | |
928 | * | |
929 | * This function assumes that the code point represented is UTF-8 variant. | |
56576a04 KW |
930 | * The caller should have excluded the possibility of it being invariant |
931 | * before calling this function. | |
2b479609 KW |
932 | * |
933 | * 'flags' can be 0, or any combination of the UTF8_DISALLOW_foo flags | |
934 | * accepted by L</utf8n_to_uvchr>. If non-zero, this function will return | |
935 | * 0 if the code point represented is well-formed Perl-extended-UTF-8, but | |
936 | * disallowed by the flags. If the input is only for a partial character, | |
937 | * the function will return non-zero if there is any sequence of | |
938 | * well-formed UTF-8 that, when appended to the input sequence, could | |
939 | * result in an allowed code point; otherwise it returns 0. Non characters | |
940 | * cannot be determined based on partial character input. But many of the | |
941 | * other excluded types can be determined with just the first one or two | |
942 | * bytes. | |
943 | * | |
944 | */ | |
945 | ||
946 | PERL_ARGS_ASSERT__IS_UTF8_CHAR_HELPER; | |
947 | ||
948 | assert(0 == (flags & ~(UTF8_DISALLOW_ILLEGAL_INTERCHANGE | |
d044b7a7 | 949 | |UTF8_DISALLOW_PERL_EXTENDED))); |
2b479609 | 950 | assert(! UTF8_IS_INVARIANT(*s)); |
35f8c9bd | 951 | |
2b479609 | 952 | /* A variant char must begin with a start byte */ |
35f8c9bd KW |
953 | if (UNLIKELY(! UTF8_IS_START(*s))) { |
954 | return 0; | |
955 | } | |
956 | ||
edc2c47a KW |
957 | /* Examine a maximum of a single whole code point */ |
958 | if (e - s > UTF8SKIP(s)) { | |
959 | e = s + UTF8SKIP(s); | |
960 | } | |
961 | ||
2b479609 KW |
962 | len = e - s; |
963 | ||
964 | if (flags && isUTF8_POSSIBLY_PROBLEMATIC(*s)) { | |
965 | const U8 s0 = NATIVE_UTF8_TO_I8(s[0]); | |
35f8c9bd | 966 | |
56576a04 KW |
967 | /* Here, we are disallowing some set of largish code points, and the |
968 | * first byte indicates the sequence is for a code point that could be | |
969 | * in the excluded set. We generally don't have to look beyond this or | |
970 | * the second byte to see if the sequence is actually for one of the | |
971 | * excluded classes. The code below is derived from this table: | |
972 | * | |
2b479609 KW |
973 | * UTF-8 UTF-EBCDIC I8 |
974 | * U+D800: \xED\xA0\x80 \xF1\xB6\xA0\xA0 First surrogate | |
975 | * U+DFFF: \xED\xBF\xBF \xF1\xB7\xBF\xBF Final surrogate | |
976 | * U+110000: \xF4\x90\x80\x80 \xF9\xA2\xA0\xA0\xA0 First above Unicode | |
977 | * | |
56576a04 KW |
978 | * Keep in mind that legal continuation bytes range between \x80..\xBF |
979 | * for UTF-8, and \xA0..\xBF for I8. Anything above those aren't | |
980 | * continuation bytes. Hence, we don't have to test the upper edge | |
981 | * because if any of those is encountered, the sequence is malformed, | |
982 | * and would fail elsewhere in this function. | |
983 | * | |
984 | * The code here likewise assumes that there aren't other | |
985 | * malformations; again the function should fail elsewhere because of | |
986 | * these. For example, an overlong beginning with FC doesn't actually | |
987 | * have to be a super; it could actually represent a small code point, | |
988 | * even U+0000. But, since overlongs (and other malformations) are | |
989 | * illegal, the function should return FALSE in either case. | |
2b479609 KW |
990 | */ |
991 | ||
992 | #ifdef EBCDIC /* On EBCDIC, these are actually I8 bytes */ | |
993 | # define FIRST_START_BYTE_THAT_IS_DEFINITELY_SUPER 0xFA | |
19794540 | 994 | # define IS_UTF8_2_BYTE_SUPER(s0, s1) ((s0) == 0xF9 && (s1) >= 0xA2) |
2b479609 | 995 | |
19794540 KW |
996 | # define IS_UTF8_2_BYTE_SURROGATE(s0, s1) ((s0) == 0xF1 \ |
997 | /* B6 and B7 */ \ | |
998 | && ((s1) & 0xFE ) == 0xB6) | |
57ff5f59 | 999 | # define isUTF8_PERL_EXTENDED(s) (*s == I8_TO_NATIVE_UTF8(0xFF)) |
2b479609 KW |
1000 | #else |
1001 | # define FIRST_START_BYTE_THAT_IS_DEFINITELY_SUPER 0xF5 | |
19794540 KW |
1002 | # define IS_UTF8_2_BYTE_SUPER(s0, s1) ((s0) == 0xF4 && (s1) >= 0x90) |
1003 | # define IS_UTF8_2_BYTE_SURROGATE(s0, s1) ((s0) == 0xED && (s1) >= 0xA0) | |
57ff5f59 | 1004 | # define isUTF8_PERL_EXTENDED(s) (*s >= 0xFE) |
2b479609 KW |
1005 | #endif |
1006 | ||
1007 | if ( (flags & UTF8_DISALLOW_SUPER) | |
ddb65933 KW |
1008 | && UNLIKELY(s0 >= FIRST_START_BYTE_THAT_IS_DEFINITELY_SUPER)) |
1009 | { | |
2b479609 KW |
1010 | return 0; /* Above Unicode */ |
1011 | } | |
1012 | ||
d044b7a7 | 1013 | if ( (flags & UTF8_DISALLOW_PERL_EXTENDED) |
57ff5f59 | 1014 | && UNLIKELY(isUTF8_PERL_EXTENDED(s))) |
2b479609 | 1015 | { |
57ff5f59 | 1016 | return 0; |
2b479609 KW |
1017 | } |
1018 | ||
1019 | if (len > 1) { | |
1020 | const U8 s1 = NATIVE_UTF8_TO_I8(s[1]); | |
1021 | ||
1022 | if ( (flags & UTF8_DISALLOW_SUPER) | |
19794540 | 1023 | && UNLIKELY(IS_UTF8_2_BYTE_SUPER(s0, s1))) |
2b479609 KW |
1024 | { |
1025 | return 0; /* Above Unicode */ | |
1026 | } | |
1027 | ||
1028 | if ( (flags & UTF8_DISALLOW_SURROGATE) | |
19794540 | 1029 | && UNLIKELY(IS_UTF8_2_BYTE_SURROGATE(s0, s1))) |
2b479609 KW |
1030 | { |
1031 | return 0; /* Surrogate */ | |
1032 | } | |
1033 | ||
1034 | if ( (flags & UTF8_DISALLOW_NONCHAR) | |
1035 | && UNLIKELY(UTF8_IS_NONCHAR(s, e))) | |
1036 | { | |
1037 | return 0; /* Noncharacter code point */ | |
1038 | } | |
1039 | } | |
1040 | } | |
1041 | ||
1042 | /* Make sure that all that follows are continuation bytes */ | |
35f8c9bd KW |
1043 | for (x = s + 1; x < e; x++) { |
1044 | if (UNLIKELY(! UTF8_IS_CONTINUATION(*x))) { | |
1045 | return 0; | |
1046 | } | |
1047 | } | |
1048 | ||
af13dd8a | 1049 | /* Here is syntactically valid. Next, make sure this isn't the start of an |
12a4bed3 | 1050 | * overlong. */ |
d6be65ae | 1051 | if (len > 1 && is_utf8_overlong_given_start_byte_ok(s, len) > 0) { |
12a4bed3 | 1052 | return 0; |
af13dd8a KW |
1053 | } |
1054 | ||
12a4bed3 KW |
1055 | /* And finally, that the code point represented fits in a word on this |
1056 | * platform */ | |
e050c007 KW |
1057 | if (0 < does_utf8_overflow(s, e, |
1058 | 0 /* Don't consider overlongs */ | |
1059 | )) | |
1060 | { | |
12a4bed3 | 1061 | return 0; |
35f8c9bd KW |
1062 | } |
1063 | ||
2b479609 | 1064 | return UTF8SKIP(s); |
35f8c9bd KW |
1065 | } |
1066 | ||
7e2f38b2 | 1067 | char * |
63ab03b3 | 1068 | Perl__byte_dump_string(pTHX_ const U8 * const start, const STRLEN len, const bool format) |
7cf8d05d KW |
1069 | { |
1070 | /* Returns a mortalized C string that is a displayable copy of the 'len' | |
63ab03b3 | 1071 | * bytes starting at 'start'. 'format' gives how to display each byte. |
7e2f38b2 KW |
1072 | * Currently, there are only two formats, so it is currently a bool: |
1073 | * 0 \xab | |
1074 | * 1 ab (that is a space between two hex digit bytes) | |
1075 | */ | |
7cf8d05d KW |
1076 | |
1077 | const STRLEN output_len = 4 * len + 1; /* 4 bytes per each input, plus a | |
1078 | trailing NUL */ | |
63ab03b3 KW |
1079 | const U8 * s = start; |
1080 | const U8 * const e = start + len; | |
7cf8d05d KW |
1081 | char * output; |
1082 | char * d; | |
1083 | ||
1084 | PERL_ARGS_ASSERT__BYTE_DUMP_STRING; | |
1085 | ||
1086 | Newx(output, output_len, char); | |
1087 | SAVEFREEPV(output); | |
1088 | ||
1089 | d = output; | |
63ab03b3 | 1090 | for (s = start; s < e; s++) { |
7cf8d05d KW |
1091 | const unsigned high_nibble = (*s & 0xF0) >> 4; |
1092 | const unsigned low_nibble = (*s & 0x0F); | |
1093 | ||
7e2f38b2 | 1094 | if (format) { |
63ab03b3 KW |
1095 | if (s > start) { |
1096 | *d++ = ' '; | |
1097 | } | |
7e2f38b2 KW |
1098 | } |
1099 | else { | |
1100 | *d++ = '\\'; | |
1101 | *d++ = 'x'; | |
1102 | } | |
7cf8d05d KW |
1103 | |
1104 | if (high_nibble < 10) { | |
1105 | *d++ = high_nibble + '0'; | |
1106 | } | |
1107 | else { | |
1108 | *d++ = high_nibble - 10 + 'a'; | |
1109 | } | |
1110 | ||
1111 | if (low_nibble < 10) { | |
1112 | *d++ = low_nibble + '0'; | |
1113 | } | |
1114 | else { | |
1115 | *d++ = low_nibble - 10 + 'a'; | |
1116 | } | |
1117 | } | |
1118 | ||
1119 | *d = '\0'; | |
1120 | return output; | |
1121 | } | |
1122 | ||
806547a7 | 1123 | PERL_STATIC_INLINE char * |
7cf8d05d KW |
1124 | S_unexpected_non_continuation_text(pTHX_ const U8 * const s, |
1125 | ||
1126 | /* How many bytes to print */ | |
3cc6a05e | 1127 | STRLEN print_len, |
7cf8d05d KW |
1128 | |
1129 | /* Which one is the non-continuation */ | |
1130 | const STRLEN non_cont_byte_pos, | |
1131 | ||
1132 | /* How many bytes should there be? */ | |
1133 | const STRLEN expect_len) | |
806547a7 KW |
1134 | { |
1135 | /* Return the malformation warning text for an unexpected continuation | |
1136 | * byte. */ | |
1137 | ||
7cf8d05d | 1138 | const char * const where = (non_cont_byte_pos == 1) |
806547a7 | 1139 | ? "immediately" |
7cf8d05d KW |
1140 | : Perl_form(aTHX_ "%d bytes", |
1141 | (int) non_cont_byte_pos); | |
806547a7 KW |
1142 | |
1143 | PERL_ARGS_ASSERT_UNEXPECTED_NON_CONTINUATION_TEXT; | |
1144 | ||
7cf8d05d KW |
1145 | /* We don't need to pass this parameter, but since it has already been |
1146 | * calculated, it's likely faster to pass it; verify under DEBUGGING */ | |
1147 | assert(expect_len == UTF8SKIP(s)); | |
1148 | ||
1149 | return Perl_form(aTHX_ "%s: %s (unexpected non-continuation byte 0x%02x," | |
1150 | " %s after start byte 0x%02x; need %d bytes, got %d)", | |
1151 | malformed_text, | |
7e2f38b2 | 1152 | _byte_dump_string(s, print_len, 0), |
7cf8d05d KW |
1153 | *(s + non_cont_byte_pos), |
1154 | where, | |
1155 | *s, | |
1156 | (int) expect_len, | |
1157 | (int) non_cont_byte_pos); | |
806547a7 KW |
1158 | } |
1159 | ||
35f8c9bd KW |
1160 | /* |
1161 | ||
de69f3af | 1162 | =for apidoc utf8n_to_uvchr |
378516de KW |
1163 | |
1164 | THIS FUNCTION SHOULD BE USED IN ONLY VERY SPECIALIZED CIRCUMSTANCES. | |
de69f3af | 1165 | Most code should use L</utf8_to_uvchr_buf>() rather than call this directly. |
67e989fb | 1166 | |
9041c2e3 | 1167 | Bottom level UTF-8 decode routine. |
de69f3af | 1168 | Returns the native code point value of the first character in the string C<s>, |
746afd53 KW |
1169 | which is assumed to be in UTF-8 (or UTF-EBCDIC) encoding, and no longer than |
1170 | C<curlen> bytes; C<*retlen> (if C<retlen> isn't NULL) will be set to | |
1171 | the length, in bytes, of that character. | |
949cf498 KW |
1172 | |
1173 | The value of C<flags> determines the behavior when C<s> does not point to a | |
2b5e7bc2 KW |
1174 | well-formed UTF-8 character. If C<flags> is 0, encountering a malformation |
1175 | causes zero to be returned and C<*retlen> is set so that (S<C<s> + C<*retlen>>) | |
1176 | is the next possible position in C<s> that could begin a non-malformed | |
1177 | character. Also, if UTF-8 warnings haven't been lexically disabled, a warning | |
1178 | is raised. Some UTF-8 input sequences may contain multiple malformations. | |
1179 | This function tries to find every possible one in each call, so multiple | |
56576a04 | 1180 | warnings can be raised for the same sequence. |
949cf498 KW |
1181 | |
1182 | Various ALLOW flags can be set in C<flags> to allow (and not warn on) | |
1183 | individual types of malformations, such as the sequence being overlong (that | |
1184 | is, when there is a shorter sequence that can express the same code point; | |
1185 | overlong sequences are expressly forbidden in the UTF-8 standard due to | |
1186 | potential security issues). Another malformation example is the first byte of | |
1187 | a character not being a legal first byte. See F<utf8.h> for the list of such | |
94953955 KW |
1188 | flags. Even if allowed, this function generally returns the Unicode |
1189 | REPLACEMENT CHARACTER when it encounters a malformation. There are flags in | |
1190 | F<utf8.h> to override this behavior for the overlong malformations, but don't | |
1191 | do that except for very specialized purposes. | |
949cf498 | 1192 | |
796b6530 | 1193 | The C<UTF8_CHECK_ONLY> flag overrides the behavior when a non-allowed (by other |
949cf498 KW |
1194 | flags) malformation is found. If this flag is set, the routine assumes that |
1195 | the caller will raise a warning, and this function will silently just set | |
d088425d KW |
1196 | C<retlen> to C<-1> (cast to C<STRLEN>) and return zero. |
1197 | ||
75200dff | 1198 | Note that this API requires disambiguation between successful decoding a C<NUL> |
796b6530 | 1199 | character, and an error return (unless the C<UTF8_CHECK_ONLY> flag is set), as |
111fa700 KW |
1200 | in both cases, 0 is returned, and, depending on the malformation, C<retlen> may |
1201 | be set to 1. To disambiguate, upon a zero return, see if the first byte of | |
1202 | C<s> is 0 as well. If so, the input was a C<NUL>; if not, the input had an | |
f9380377 | 1203 | error. Or you can use C<L</utf8n_to_uvchr_error>>. |
949cf498 KW |
1204 | |
1205 | Certain code points are considered problematic. These are Unicode surrogates, | |
746afd53 | 1206 | Unicode non-characters, and code points above the Unicode maximum of 0x10FFFF. |
949cf498 | 1207 | By default these are considered regular code points, but certain situations |
ecc1615f KW |
1208 | warrant special handling for them, which can be specified using the C<flags> |
1209 | parameter. If C<flags> contains C<UTF8_DISALLOW_ILLEGAL_INTERCHANGE>, all | |
1210 | three classes are treated as malformations and handled as such. The flags | |
1211 | C<UTF8_DISALLOW_SURROGATE>, C<UTF8_DISALLOW_NONCHAR>, and | |
1212 | C<UTF8_DISALLOW_SUPER> (meaning above the legal Unicode maximum) can be set to | |
1213 | disallow these categories individually. C<UTF8_DISALLOW_ILLEGAL_INTERCHANGE> | |
1214 | restricts the allowed inputs to the strict UTF-8 traditionally defined by | |
1215 | Unicode. Use C<UTF8_DISALLOW_ILLEGAL_C9_INTERCHANGE> to use the strictness | |
1216 | definition given by | |
1217 | L<Unicode Corrigendum #9|http://www.unicode.org/versions/corrigendum9.html>. | |
1218 | The difference between traditional strictness and C9 strictness is that the | |
1219 | latter does not forbid non-character code points. (They are still discouraged, | |
1220 | however.) For more discussion see L<perlunicode/Noncharacter code points>. | |
1221 | ||
1222 | The flags C<UTF8_WARN_ILLEGAL_INTERCHANGE>, | |
1223 | C<UTF8_WARN_ILLEGAL_C9_INTERCHANGE>, C<UTF8_WARN_SURROGATE>, | |
796b6530 KW |
1224 | C<UTF8_WARN_NONCHAR>, and C<UTF8_WARN_SUPER> will cause warning messages to be |
1225 | raised for their respective categories, but otherwise the code points are | |
1226 | considered valid (not malformations). To get a category to both be treated as | |
1227 | a malformation and raise a warning, specify both the WARN and DISALLOW flags. | |
949cf498 | 1228 | (But note that warnings are not raised if lexically disabled nor if |
796b6530 | 1229 | C<UTF8_CHECK_ONLY> is also specified.) |
949cf498 | 1230 | |
57ff5f59 KW |
1231 | Extremely high code points were never specified in any standard, and require an |
1232 | extension to UTF-8 to express, which Perl does. It is likely that programs | |
1233 | written in something other than Perl would not be able to read files that | |
1234 | contain these; nor would Perl understand files written by something that uses a | |
1235 | different extension. For these reasons, there is a separate set of flags that | |
1236 | can warn and/or disallow these extremely high code points, even if other | |
1237 | above-Unicode ones are accepted. They are the C<UTF8_WARN_PERL_EXTENDED> and | |
1238 | C<UTF8_DISALLOW_PERL_EXTENDED> flags. For more information see | |
1239 | L</C<UTF8_GOT_PERL_EXTENDED>>. Of course C<UTF8_DISALLOW_SUPER> will treat all | |
1240 | above-Unicode code points, including these, as malformations. | |
1241 | (Note that the Unicode standard considers anything above 0x10FFFF to be | |
1242 | illegal, but there are standards predating it that allow up to 0x7FFF_FFFF | |
1243 | (2**31 -1)) | |
1244 | ||
1245 | A somewhat misleadingly named synonym for C<UTF8_WARN_PERL_EXTENDED> is | |
1246 | retained for backward compatibility: C<UTF8_WARN_ABOVE_31_BIT>. Similarly, | |
1247 | C<UTF8_DISALLOW_ABOVE_31_BIT> is usable instead of the more accurately named | |
1248 | C<UTF8_DISALLOW_PERL_EXTENDED>. The names are misleading because these flags | |
1249 | can apply to code points that actually do fit in 31 bits. This happens on | |
1250 | EBCDIC platforms, and sometimes when the L<overlong | |
1251 | malformation|/C<UTF8_GOT_LONG>> is also present. The new names accurately | |
1252 | describe the situation in all cases. | |
1253 | ||
ab8e6d41 | 1254 | |
949cf498 KW |
1255 | All other code points corresponding to Unicode characters, including private |
1256 | use and those yet to be assigned, are never considered malformed and never | |
1257 | warn. | |
67e989fb | 1258 | |
37607a96 | 1259 | =cut |
f9380377 KW |
1260 | |
1261 | Also implemented as a macro in utf8.h | |
1262 | */ | |
1263 | ||
1264 | UV | |
1265 | Perl_utf8n_to_uvchr(pTHX_ const U8 *s, | |
1266 | STRLEN curlen, | |
1267 | STRLEN *retlen, | |
1268 | const U32 flags) | |
1269 | { | |
1270 | PERL_ARGS_ASSERT_UTF8N_TO_UVCHR; | |
1271 | ||
1272 | return utf8n_to_uvchr_error(s, curlen, retlen, flags, NULL); | |
1273 | } | |
1274 | ||
2b9519f0 KW |
1275 | /* The tables below come from http://bjoern.hoehrmann.de/utf-8/decoder/dfa/, |
1276 | * which requires this copyright notice */ | |
1277 | ||
1278 | /* Copyright (c) 2008-2009 Bjoern Hoehrmann <bjoern@hoehrmann.de> | |
1279 | ||
1280 | Permission is hereby granted, free of charge, to any person obtaining a copy of | |
1281 | this software and associated documentation files (the "Software"), to deal in | |
1282 | the Software without restriction, including without limitation the rights to | |
1283 | use, copy, modify, merge, publish, distribute, sublicense, and/or sell copies | |
1284 | of the Software, and to permit persons to whom the Software is furnished to do | |
1285 | so, subject to the following conditions: | |
1286 | ||
1287 | The above copyright notice and this permission notice shall be included in all | |
1288 | copies or substantial portions of the Software. | |
1289 | ||
1290 | THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR | |
1291 | IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, | |
1292 | FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE | |
1293 | AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER | |
1294 | LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, | |
1295 | OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE | |
1296 | SOFTWARE. | |
1297 | ||
1298 | */ | |
1299 | ||
1300 | #if 0 | |
1301 | static U8 utf8d_C9[] = { | |
1302 | /* The first part of the table maps bytes to character classes that | |
1303 | * to reduce the size of the transition table and create bitmasks. */ | |
1304 | 0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0, 0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0, /*-1F*/ | |
1305 | 0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0, 0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0, /*-3F*/ | |
1306 | 0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0, 0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0, /*-5F*/ | |
1307 | 0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0, 0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0, /*-7F*/ | |
1308 | 1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1, 9,9,9,9,9,9,9,9,9,9,9,9,9,9,9,9, /*-9F*/ | |
1309 | 7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7, 7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7, /*-BF*/ | |
1310 | 8,8,2,2,2,2,2,2,2,2,2,2,2,2,2,2, 2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2, /*-DF*/ | |
1311 | 10,3,3,3,3,3,3,3,3,3,3,3,3,4,3,3, 11,6,6,6,5,8,8,8,8,8,8,8,8,8,8,8, /*-FF*/ | |
1312 | ||
1313 | /* The second part is a transition table that maps a combination | |
1314 | * of a state of the automaton and a character class to a state. */ | |
1315 | 0,12,24,36,60,96,84,12,12,12,48,72, 12,12,12,12,12,12,12,12,12,12,12,12, | |
1316 | 12, 0,12,12,12,12,12, 0,12, 0,12,12, 12,24,12,12,12,12,12,24,12,24,12,12, | |
1317 | 12,12,12,12,12,12,12,24,12,12,12,12, 12,24,12,12,12,12,12,12,12,24,12,12, | |
1318 | 12,12,12,12,12,12,12,36,12,36,12,12, 12,36,12,12,12,12,12,36,12,36,12,12, | |
1319 | 12,36,12,12,12,12,12,12,12,12,12,12 | |
1320 | }; | |
1321 | ||
1322 | #endif | |
1323 | ||
1324 | #ifndef EBCDIC | |
1325 | ||
1326 | /* This is a version of the above table customized for Perl that doesn't | |
1327 | * exclude surrogates and accepts start bytes up through F7 (representing | |
1328 | * 2**21 - 1). */ | |
1329 | static U8 dfa_tab_for_perl[] = { | |
1330 | /* The first part of the table maps bytes to character classes to reduce | |
1331 | * the size of the transition table and create bitmasks. */ | |
1332 | 0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0, 0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0, /*-1F*/ | |
1333 | 0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0, 0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0, /*-3F*/ | |
1334 | 0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0, 0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0, /*-5F*/ | |
1335 | 0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0, 0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0, /*-7F*/ | |
1336 | 1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1, 9,9,9,9,9,9,9,9,9,9,9,9,9,9,9,9, /*-9F*/ | |
1337 | 7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7, 7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7, /*-BF*/ | |
1338 | 8,8,2,2,2,2,2,2,2,2,2,2,2,2,2,2, 2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2, /*-DF*/ | |
1339 | 10,3,3,3,3,3,3,3,3,3,3,3,3,3,3,3, 11,4,4,4,4,4,4,4,8,8,8,8,8,8,8,8, /*-FF*/ | |
1340 | ||
1341 | /* The second part is a transition table that maps a combination | |
1342 | * of a state of the automaton and a character class to a state. */ | |
1343 | 0,12,24,36,96,12,12,12,12,12,48,72, 12,12,12,12,12,12,12,12,12,12,12,12,/*23*/ | |
1344 | 12, 0,12,12,12,12,12, 0,12, 0,12,12, 12,24,12,12,12,12,12,24,12,24,12,12,/*47*/ | |
1345 | 12,12,12,12,12,12,12,24,12,12,12,12, 12,24,12,12,12,12,12,12,12,24,12,12,/*71*/ | |
1346 | 12,12,12,12,12,12,12,36,12,36,12,12, 12,36,12,12,12,12,12,36,12,36,12,12,/*95*/ | |
1347 | 12,36,12,12,12,12,12,36,12,36,12,12 /* 96- 107 */ | |
1348 | ||
1349 | /* The customization was to repurpose the surrogates type '4' to instead be | |
1350 | * for start bytes F1-F7. Types 5 and 6 are now unused, and their entries in | |
1351 | * the transition part of the table are set to 12, so are illegal. | |
1352 | * | |
1353 | * To do higher code points would require expansion and some rearrangement of | |
1354 | * the table. The type '1' entries for continuation bytes 80-8f would have to | |
1355 | * be split into several types, because they aren't treated uniformly for | |
1356 | * higher start bytes, since overlongs for F8 are 80-87; FC: 80-83; and FE: | |
1357 | * 80-81. We start needing to worry about overflow if FE is included. | |
1358 | * Ignoring, FE and FF, we could use type 5 for F9-FB, and 6 for FD (remember | |
1359 | * from the web site that these are used to right shift). FE would | |
1360 | * necessarily be type 7; and FF, type 8. And new states would have to be | |
1361 | * created for F8 and FC (and FE and FF if used), so quite a bit of work would | |
1362 | * be involved. | |
1363 | * | |
1364 | * XXX Better would be to customize the table so that the noncharacters are | |
1365 | * excluded. This again is non trivial, but doing so would simplify the code | |
1366 | * that uses this, and might make it small enough to make it inlinable */ | |
1367 | }; | |
1368 | ||
1369 | #endif | |
1370 | ||
f9380377 KW |
1371 | /* |
1372 | ||
1373 | =for apidoc utf8n_to_uvchr_error | |
1374 | ||
1375 | THIS FUNCTION SHOULD BE USED IN ONLY VERY SPECIALIZED CIRCUMSTANCES. | |
1376 | Most code should use L</utf8_to_uvchr_buf>() rather than call this directly. | |
1377 | ||
1378 | This function is for code that needs to know what the precise malformation(s) | |
37657a5b KW |
1379 | are when an error is found. If you also need to know the generated warning |
1380 | messages, use L</utf8n_to_uvchr_msgs>() instead. | |
f9380377 KW |
1381 | |
1382 | It is like C<L</utf8n_to_uvchr>> but it takes an extra parameter placed after | |
1383 | all the others, C<errors>. If this parameter is 0, this function behaves | |
1384 | identically to C<L</utf8n_to_uvchr>>. Otherwise, C<errors> should be a pointer | |
1385 | to a C<U32> variable, which this function sets to indicate any errors found. | |
1386 | Upon return, if C<*errors> is 0, there were no errors found. Otherwise, | |
1387 | C<*errors> is the bit-wise C<OR> of the bits described in the list below. Some | |
1388 | of these bits will be set if a malformation is found, even if the input | |
7a65503b | 1389 | C<flags> parameter indicates that the given malformation is allowed; those |
f9380377 KW |
1390 | exceptions are noted: |
1391 | ||
1392 | =over 4 | |
1393 | ||
57ff5f59 | 1394 | =item C<UTF8_GOT_PERL_EXTENDED> |
f9380377 | 1395 | |
57ff5f59 KW |
1396 | The input sequence is not standard UTF-8, but a Perl extension. This bit is |
1397 | set only if the input C<flags> parameter contains either the | |
1398 | C<UTF8_DISALLOW_PERL_EXTENDED> or the C<UTF8_WARN_PERL_EXTENDED> flags. | |
1399 | ||
1400 | Code points above 0x7FFF_FFFF (2**31 - 1) were never specified in any standard, | |
1401 | and so some extension must be used to express them. Perl uses a natural | |
1402 | extension to UTF-8 to represent the ones up to 2**36-1, and invented a further | |
1403 | extension to represent even higher ones, so that any code point that fits in a | |
1404 | 64-bit word can be represented. Text using these extensions is not likely to | |
1405 | be portable to non-Perl code. We lump both of these extensions together and | |
1406 | refer to them as Perl extended UTF-8. There exist other extensions that people | |
1407 | have invented, incompatible with Perl's. | |
1408 | ||
1409 | On EBCDIC platforms starting in Perl v5.24, the Perl extension for representing | |
1410 | extremely high code points kicks in at 0x3FFF_FFFF (2**30 -1), which is lower | |
1411 | than on ASCII. Prior to that, code points 2**31 and higher were simply | |
1412 | unrepresentable, and a different, incompatible method was used to represent | |
1413 | code points between 2**30 and 2**31 - 1. | |
1414 | ||
1415 | On both platforms, ASCII and EBCDIC, C<UTF8_GOT_PERL_EXTENDED> is set if | |
1416 | Perl extended UTF-8 is used. | |
1417 | ||
1418 | In earlier Perls, this bit was named C<UTF8_GOT_ABOVE_31_BIT>, which you still | |
1419 | may use for backward compatibility. That name is misleading, as this flag may | |
1420 | be set when the code point actually does fit in 31 bits. This happens on | |
1421 | EBCDIC platforms, and sometimes when the L<overlong | |
1422 | malformation|/C<UTF8_GOT_LONG>> is also present. The new name accurately | |
1423 | describes the situation in all cases. | |
f9380377 KW |
1424 | |
1425 | =item C<UTF8_GOT_CONTINUATION> | |
1426 | ||
1427 | The input sequence was malformed in that the first byte was a a UTF-8 | |
1428 | continuation byte. | |
1429 | ||
1430 | =item C<UTF8_GOT_EMPTY> | |
1431 | ||
1432 | The input C<curlen> parameter was 0. | |
1433 | ||
1434 | =item C<UTF8_GOT_LONG> | |
1435 | ||
1436 | The input sequence was malformed in that there is some other sequence that | |
1437 | evaluates to the same code point, but that sequence is shorter than this one. | |
1438 | ||
fecaf136 KW |
1439 | Until Unicode 3.1, it was legal for programs to accept this malformation, but |
1440 | it was discovered that this created security issues. | |
1441 | ||
f9380377 KW |
1442 | =item C<UTF8_GOT_NONCHAR> |
1443 | ||
1444 | The code point represented by the input UTF-8 sequence is for a Unicode | |
1445 | non-character code point. | |
1446 | This bit is set only if the input C<flags> parameter contains either the | |
1447 | C<UTF8_DISALLOW_NONCHAR> or the C<UTF8_WARN_NONCHAR> flags. | |
1448 | ||
1449 | =item C<UTF8_GOT_NON_CONTINUATION> | |
1450 | ||
1451 | The input sequence was malformed in that a non-continuation type byte was found | |
1452 | in a position where only a continuation type one should be. | |
1453 | ||
1454 | =item C<UTF8_GOT_OVERFLOW> | |
1455 | ||
1456 | The input sequence was malformed in that it is for a code point that is not | |
d22ec717 | 1457 | representable in the number of bits available in an IV on the current platform. |
f9380377 KW |
1458 | |
1459 | =item C<UTF8_GOT_SHORT> | |
1460 | ||
1461 | The input sequence was malformed in that C<curlen> is smaller than required for | |
1462 | a complete sequence. In other words, the input is for a partial character | |
1463 | sequence. | |
1464 | ||
1465 | =item C<UTF8_GOT_SUPER> | |
1466 | ||
1467 | The input sequence was malformed in that it is for a non-Unicode code point; | |
1468 | that is, one above the legal Unicode maximum. | |
1469 | This bit is set only if the input C<flags> parameter contains either the | |
1470 | C<UTF8_DISALLOW_SUPER> or the C<UTF8_WARN_SUPER> flags. | |
1471 | ||
1472 | =item C<UTF8_GOT_SURROGATE> | |
1473 | ||
1474 | The input sequence was malformed in that it is for a -Unicode UTF-16 surrogate | |
1475 | code point. | |
1476 | This bit is set only if the input C<flags> parameter contains either the | |
1477 | C<UTF8_DISALLOW_SURROGATE> or the C<UTF8_WARN_SURROGATE> flags. | |
1478 | ||
1479 | =back | |
1480 | ||
133551d8 KW |
1481 | To do your own error handling, call this function with the C<UTF8_CHECK_ONLY> |
1482 | flag to suppress any warnings, and then examine the C<*errors> return. | |
1483 | ||
f9380377 | 1484 | =cut |
37657a5b KW |
1485 | |
1486 | Also implemented as a macro in utf8.h | |
37607a96 | 1487 | */ |
67e989fb | 1488 | |
a0ed51b3 | 1489 | UV |
f9380377 | 1490 | Perl_utf8n_to_uvchr_error(pTHX_ const U8 *s, |
37657a5b KW |
1491 | STRLEN curlen, |
1492 | STRLEN *retlen, | |
1493 | const U32 flags, | |
1494 | U32 * errors) | |
1495 | { | |
1496 | PERL_ARGS_ASSERT_UTF8N_TO_UVCHR_ERROR; | |
1497 | ||
1498 | return utf8n_to_uvchr_msgs(s, curlen, retlen, flags, errors, NULL); | |
1499 | } | |
1500 | ||
1501 | /* | |
1502 | ||
1503 | =for apidoc utf8n_to_uvchr_msgs | |
1504 | ||
1505 | THIS FUNCTION SHOULD BE USED IN ONLY VERY SPECIALIZED CIRCUMSTANCES. | |
1506 | Most code should use L</utf8_to_uvchr_buf>() rather than call this directly. | |
1507 | ||
1508 | This function is for code that needs to know what the precise malformation(s) | |
1509 | are when an error is found, and wants the corresponding warning and/or error | |
1510 | messages to be returned to the caller rather than be displayed. All messages | |
1511 | that would have been displayed if all lexcial warnings are enabled will be | |
1512 | returned. | |
1513 | ||
1514 | It is just like C<L</utf8n_to_uvchr_error>> but it takes an extra parameter | |
1515 | placed after all the others, C<msgs>. If this parameter is 0, this function | |
1516 | behaves identically to C<L</utf8n_to_uvchr_error>>. Otherwise, C<msgs> should | |
1517 | be a pointer to an C<AV *> variable, in which this function creates a new AV to | |
1518 | contain any appropriate messages. The elements of the array are ordered so | |
1519 | that the first message that would have been displayed is in the 0th element, | |
1520 | and so on. Each element is a hash with three key-value pairs, as follows: | |
1521 | ||
1522 | =over 4 | |
1523 | ||
1524 | =item C<text> | |
1525 | ||
1526 | The text of the message as a C<SVpv>. | |
1527 | ||
1528 | =item C<warn_categories> | |
1529 | ||
1530 | The warning category (or categories) packed into a C<SVuv>. | |
1531 | ||
1532 | =item C<flag> | |
1533 | ||
1534 | A single flag bit associated with this message, in a C<SVuv>. | |
1535 | The bit corresponds to some bit in the C<*errors> return value, | |
1536 | such as C<UTF8_GOT_LONG>. | |
1537 | ||
1538 | =back | |
1539 | ||
1540 | It's important to note that specifying this parameter as non-null will cause | |
1541 | any warnings this function would otherwise generate to be suppressed, and | |
1542 | instead be placed in C<*msgs>. The caller can check the lexical warnings state | |
1543 | (or not) when choosing what to do with the returned messages. | |
1544 | ||
1545 | If the flag C<UTF8_CHECK_ONLY> is passed, no warnings are generated, and hence | |
1546 | no AV is created. | |
1547 | ||
1548 | The caller, of course, is responsible for freeing any returned AV. | |
1549 | ||
1550 | =cut | |
1551 | */ | |
1552 | ||
1553 | UV | |
1554 | Perl_utf8n_to_uvchr_msgs(pTHX_ const U8 *s, | |
1555 | STRLEN curlen, | |
1556 | STRLEN *retlen, | |
1557 | const U32 flags, | |
1558 | U32 * errors, | |
1559 | AV ** msgs) | |
a0ed51b3 | 1560 | { |
d4c19fe8 | 1561 | const U8 * const s0 = s; |
2b9519f0 | 1562 | const U8 * send = s0 + curlen; |
2b5e7bc2 KW |
1563 | U32 possible_problems = 0; /* A bit is set here for each potential problem |
1564 | found as we go along */ | |
2b672cf5 | 1565 | UV uv = (UV) -1; |
2b5e7bc2 KW |
1566 | STRLEN expectlen = 0; /* How long should this sequence be? |
1567 | (initialized to silence compilers' wrong | |
1568 | warning) */ | |
e308b348 | 1569 | STRLEN avail_len = 0; /* When input is too short, gives what that is */ |
f9380377 KW |
1570 | U32 discard_errors = 0; /* Used to save branches when 'errors' is NULL; |
1571 | this gets set and discarded */ | |
a0dbb045 | 1572 | |
2b5e7bc2 KW |
1573 | /* The below are used only if there is both an overlong malformation and a |
1574 | * too short one. Otherwise the first two are set to 's0' and 'send', and | |
1575 | * the third not used at all */ | |
1576 | U8 * adjusted_s0 = (U8 *) s0; | |
e9f2c446 KW |
1577 | U8 temp_char_buf[UTF8_MAXBYTES + 1]; /* Used to avoid a Newx in this |
1578 | routine; see [perl #130921] */ | |
2b5e7bc2 | 1579 | UV uv_so_far = 0; /* (Initialized to silence compilers' wrong warning) */ |
7918f24d | 1580 | |
2b9519f0 KW |
1581 | UV state = 0; |
1582 | ||
37657a5b | 1583 | PERL_ARGS_ASSERT_UTF8N_TO_UVCHR_MSGS; |
f9380377 KW |
1584 | |
1585 | if (errors) { | |
1586 | *errors = 0; | |
1587 | } | |
1588 | else { | |
1589 | errors = &discard_errors; | |
1590 | } | |
a0dbb045 | 1591 | |
eb83ed87 KW |
1592 | /* The order of malformation tests here is important. We should consume as |
1593 | * few bytes as possible in order to not skip any valid character. This is | |
1594 | * required by the Unicode Standard (section 3.9 of Unicode 6.0); see also | |
1595 | * http://unicode.org/reports/tr36 for more discussion as to why. For | |
1596 | * example, once we've done a UTF8SKIP, we can tell the expected number of | |
1597 | * bytes, and could fail right off the bat if the input parameters indicate | |
1598 | * that there are too few available. But it could be that just that first | |
1599 | * byte is garbled, and the intended character occupies fewer bytes. If we | |
1600 | * blindly assumed that the first byte is correct, and skipped based on | |
1601 | * that number, we could skip over a valid input character. So instead, we | |
1602 | * always examine the sequence byte-by-byte. | |
1603 | * | |
1604 | * We also should not consume too few bytes, otherwise someone could inject | |
1605 | * things. For example, an input could be deliberately designed to | |
1606 | * overflow, and if this code bailed out immediately upon discovering that, | |
e2660c54 | 1607 | * returning to the caller C<*retlen> pointing to the very next byte (one |
eb83ed87 KW |
1608 | * which is actually part of of the overflowing sequence), that could look |
1609 | * legitimate to the caller, which could discard the initial partial | |
2b5e7bc2 KW |
1610 | * sequence and process the rest, inappropriately. |
1611 | * | |
1612 | * Some possible input sequences are malformed in more than one way. This | |
1613 | * function goes to lengths to try to find all of them. This is necessary | |
1614 | * for correctness, as the inputs may allow one malformation but not | |
1615 | * another, and if we abandon searching for others after finding the | |
1616 | * allowed one, we could allow in something that shouldn't have been. | |
1617 | */ | |
eb83ed87 | 1618 | |
b5b9af04 | 1619 | if (UNLIKELY(curlen == 0)) { |
2b5e7bc2 KW |
1620 | possible_problems |= UTF8_GOT_EMPTY; |
1621 | curlen = 0; | |
5a48568d | 1622 | uv = UNICODE_REPLACEMENT; |
2b5e7bc2 | 1623 | goto ready_to_handle_errors; |
0c443dc2 JH |
1624 | } |
1625 | ||
eb83ed87 KW |
1626 | expectlen = UTF8SKIP(s); |
1627 | ||
1628 | /* A well-formed UTF-8 character, as the vast majority of calls to this | |
1629 | * function will be for, has this expected length. For efficiency, set | |
1630 | * things up here to return it. It will be overriden only in those rare | |
1631 | * cases where a malformation is found */ | |
1632 | if (retlen) { | |
1633 | *retlen = expectlen; | |
1634 | } | |
1635 | ||
1636 | /* An invariant is trivially well-formed */ | |
2b9519f0 KW |
1637 | if (UTF8_IS_INVARIANT(*s0)) { |
1638 | return *s0; | |
a0ed51b3 | 1639 | } |
67e989fb | 1640 | |
2b9519f0 KW |
1641 | #ifndef EBCDIC |
1642 | ||
1643 | /* Measurements show that this dfa is somewhat faster than the regular code | |
1644 | * below, so use it first, dropping down for the non-normal cases. */ | |
1645 | ||
1646 | # define PERL_UTF8_DECODE_REJECT 12 | |
1647 | ||
1648 | while (s < send && LIKELY(state != PERL_UTF8_DECODE_REJECT)) { | |
1649 | UV type = dfa_tab_for_perl[*s]; | |
1650 | ||
1651 | if (state != 0) { | |
1652 | uv = (*s & 0x3fu) | (uv << UTF_ACCUMULATION_SHIFT); | |
1653 | state = dfa_tab_for_perl[256 + state + type]; | |
1654 | } | |
1655 | else { | |
1656 | uv = (0xff >> type) & (*s); | |
1657 | state = dfa_tab_for_perl[256 + type]; | |
1658 | } | |
1659 | ||
1660 | if (state == 0) { | |
1661 | ||
1662 | /* If this could be a code point that the flags don't allow (the first | |
1663 | * surrogate is the first such possible one), delve further, but we already | |
1664 | * have calculated 'uv' */ | |
1665 | if ( (flags & (UTF8_DISALLOW_ILLEGAL_INTERCHANGE | |
1666 | |UTF8_WARN_ILLEGAL_INTERCHANGE)) | |
1667 | && uv >= UNICODE_SURROGATE_FIRST) | |
1668 | { | |
1669 | curlen = s + 1 - s0; | |
1670 | goto got_uv; | |
1671 | } | |
1672 | ||
1673 | return uv; | |
1674 | } | |
1675 | ||
1676 | s++; | |
1677 | } | |
1678 | ||
1679 | /* Here, is some sort of failure. Use the full mechanism */ | |
1680 | ||
1681 | uv = *s0; | |
1682 | ||
1683 | #endif | |
1684 | ||
eb83ed87 | 1685 | /* A continuation character can't start a valid sequence */ |
b5b9af04 | 1686 | if (UNLIKELY(UTF8_IS_CONTINUATION(uv))) { |
2b5e7bc2 KW |
1687 | possible_problems |= UTF8_GOT_CONTINUATION; |
1688 | curlen = 1; | |
1689 | uv = UNICODE_REPLACEMENT; | |
1690 | goto ready_to_handle_errors; | |
ba210ebe | 1691 | } |
9041c2e3 | 1692 | |
dcd27b3c | 1693 | /* Here is not a continuation byte, nor an invariant. The only thing left |
ddb65933 KW |
1694 | * is a start byte (possibly for an overlong). (We can't use UTF8_IS_START |
1695 | * because it excludes start bytes like \xC0 that always lead to | |
1696 | * overlongs.) */ | |
dcd27b3c | 1697 | |
534752c1 KW |
1698 | /* Convert to I8 on EBCDIC (no-op on ASCII), then remove the leading bits |
1699 | * that indicate the number of bytes in the character's whole UTF-8 | |
1700 | * sequence, leaving just the bits that are part of the value. */ | |
1701 | uv = NATIVE_UTF8_TO_I8(uv) & UTF_START_MASK(expectlen); | |
ba210ebe | 1702 | |
e308b348 KW |
1703 | /* Setup the loop end point, making sure to not look past the end of the |
1704 | * input string, and flag it as too short if the size isn't big enough. */ | |
e308b348 KW |
1705 | if (UNLIKELY(curlen < expectlen)) { |
1706 | possible_problems |= UTF8_GOT_SHORT; | |
1707 | avail_len = curlen; | |
e308b348 KW |
1708 | } |
1709 | else { | |
2b9519f0 | 1710 | send = (U8*) s0 + expectlen; |
e308b348 | 1711 | } |
e308b348 | 1712 | |
eb83ed87 | 1713 | /* Now, loop through the remaining bytes in the character's sequence, |
e308b348 | 1714 | * accumulating each into the working value as we go. */ |
eb83ed87 | 1715 | for (s = s0 + 1; s < send; s++) { |
b5b9af04 | 1716 | if (LIKELY(UTF8_IS_CONTINUATION(*s))) { |
8850bf83 | 1717 | uv = UTF8_ACCUMULATE(uv, *s); |
2b5e7bc2 KW |
1718 | continue; |
1719 | } | |
1720 | ||
1721 | /* Here, found a non-continuation before processing all expected bytes. | |
1722 | * This byte indicates the beginning of a new character, so quit, even | |
1723 | * if allowing this malformation. */ | |
2b5e7bc2 | 1724 | possible_problems |= UTF8_GOT_NON_CONTINUATION; |
e308b348 | 1725 | break; |
eb83ed87 KW |
1726 | } /* End of loop through the character's bytes */ |
1727 | ||
1728 | /* Save how many bytes were actually in the character */ | |
1729 | curlen = s - s0; | |
1730 | ||
2b5e7bc2 KW |
1731 | /* Note that there are two types of too-short malformation. One is when |
1732 | * there is actual wrong data before the normal termination of the | |
1733 | * sequence. The other is that the sequence wasn't complete before the end | |
1734 | * of the data we are allowed to look at, based on the input 'curlen'. | |
1735 | * This means that we were passed data for a partial character, but it is | |
1736 | * valid as far as we saw. The other is definitely invalid. This | |
1737 | * distinction could be important to a caller, so the two types are kept | |
15b010f0 KW |
1738 | * separate. |
1739 | * | |
1740 | * A convenience macro that matches either of the too-short conditions. */ | |
1741 | # define UTF8_GOT_TOO_SHORT (UTF8_GOT_SHORT|UTF8_GOT_NON_CONTINUATION) | |
1742 | ||
1743 | if (UNLIKELY(possible_problems & UTF8_GOT_TOO_SHORT)) { | |
1744 | uv_so_far = uv; | |
1745 | uv = UNICODE_REPLACEMENT; | |
1746 | } | |
2b5e7bc2 | 1747 | |
08e73697 KW |
1748 | /* Check for overflow. The algorithm requires us to not look past the end |
1749 | * of the current character, even if partial, so the upper limit is 's' */ | |
e050c007 KW |
1750 | if (UNLIKELY(0 < does_utf8_overflow(s0, s, |
1751 | 1 /* Do consider overlongs */ | |
1752 | ))) | |
1753 | { | |
2b5e7bc2 KW |
1754 | possible_problems |= UTF8_GOT_OVERFLOW; |
1755 | uv = UNICODE_REPLACEMENT; | |
eb83ed87 | 1756 | } |
eb83ed87 | 1757 | |
2b5e7bc2 KW |
1758 | /* Check for overlong. If no problems so far, 'uv' is the correct code |
1759 | * point value. Simply see if it is expressible in fewer bytes. Otherwise | |
1760 | * we must look at the UTF-8 byte sequence itself to see if it is for an | |
1761 | * overlong */ | |
1762 | if ( ( LIKELY(! possible_problems) | |
1763 | && UNLIKELY(expectlen > (STRLEN) OFFUNISKIP(uv))) | |
56576a04 | 1764 | || ( UNLIKELY(possible_problems) |
2b5e7bc2 KW |
1765 | && ( UNLIKELY(! UTF8_IS_START(*s0)) |
1766 | || ( curlen > 1 | |
d6be65ae | 1767 | && UNLIKELY(0 < is_utf8_overlong_given_start_byte_ok(s0, |
08e73697 | 1768 | s - s0)))))) |
2f8f112e | 1769 | { |
2b5e7bc2 KW |
1770 | possible_problems |= UTF8_GOT_LONG; |
1771 | ||
abc28b54 | 1772 | if ( UNLIKELY( possible_problems & UTF8_GOT_TOO_SHORT) |
56576a04 | 1773 | |
abc28b54 KW |
1774 | /* The calculation in the 'true' branch of this 'if' |
1775 | * below won't work if overflows, and isn't needed | |
1776 | * anyway. Further below we handle all overflow | |
1777 | * cases */ | |
1778 | && LIKELY(! (possible_problems & UTF8_GOT_OVERFLOW))) | |
1779 | { | |
2b5e7bc2 KW |
1780 | UV min_uv = uv_so_far; |
1781 | STRLEN i; | |
1782 | ||
1783 | /* Here, the input is both overlong and is missing some trailing | |
1784 | * bytes. There is no single code point it could be for, but there | |
1785 | * may be enough information present to determine if what we have | |
1786 | * so far is for an unallowed code point, such as for a surrogate. | |
56576a04 KW |
1787 | * The code further below has the intelligence to determine this, |
1788 | * but just for non-overlong UTF-8 sequences. What we do here is | |
1789 | * calculate the smallest code point the input could represent if | |
1790 | * there were no too short malformation. Then we compute and save | |
1791 | * the UTF-8 for that, which is what the code below looks at | |
1792 | * instead of the raw input. It turns out that the smallest such | |
1793 | * code point is all we need. */ | |
2b5e7bc2 KW |
1794 | for (i = curlen; i < expectlen; i++) { |
1795 | min_uv = UTF8_ACCUMULATE(min_uv, | |
1796 | I8_TO_NATIVE_UTF8(UTF_CONTINUATION_MARK)); | |
1797 | } | |
1798 | ||
e9f2c446 | 1799 | adjusted_s0 = temp_char_buf; |
57ff5f59 | 1800 | (void) uvoffuni_to_utf8_flags(adjusted_s0, min_uv, 0); |
2b5e7bc2 | 1801 | } |
eb83ed87 KW |
1802 | } |
1803 | ||
2b9519f0 KW |
1804 | got_uv: |
1805 | ||
56576a04 KW |
1806 | /* Here, we have found all the possible problems, except for when the input |
1807 | * is for a problematic code point not allowed by the input parameters. */ | |
1808 | ||
06188866 KW |
1809 | /* uv is valid for overlongs */ |
1810 | if ( ( ( LIKELY(! (possible_problems & ~UTF8_GOT_LONG)) | |
1811 | ||
1812 | /* isn't problematic if < this */ | |
1813 | && uv >= UNICODE_SURROGATE_FIRST) | |
2b5e7bc2 | 1814 | || ( UNLIKELY(possible_problems) |
d60baaa7 KW |
1815 | |
1816 | /* if overflow, we know without looking further | |
1817 | * precisely which of the problematic types it is, | |
1818 | * and we deal with those in the overflow handling | |
1819 | * code */ | |
1820 | && LIKELY(! (possible_problems & UTF8_GOT_OVERFLOW)) | |
57ff5f59 KW |
1821 | && ( isUTF8_POSSIBLY_PROBLEMATIC(*adjusted_s0) |
1822 | || UNLIKELY(isUTF8_PERL_EXTENDED(s0))))) | |
760c7c2f KW |
1823 | && ((flags & ( UTF8_DISALLOW_NONCHAR |
1824 | |UTF8_DISALLOW_SURROGATE | |
1825 | |UTF8_DISALLOW_SUPER | |
d044b7a7 | 1826 | |UTF8_DISALLOW_PERL_EXTENDED |
760c7c2f KW |
1827 | |UTF8_WARN_NONCHAR |
1828 | |UTF8_WARN_SURROGATE | |
1829 | |UTF8_WARN_SUPER | |
d22ec717 | 1830 | |UTF8_WARN_PERL_EXTENDED)))) |
eb83ed87 | 1831 | { |
2b5e7bc2 KW |
1832 | /* If there were no malformations, or the only malformation is an |
1833 | * overlong, 'uv' is valid */ | |
1834 | if (LIKELY(! (possible_problems & ~UTF8_GOT_LONG))) { | |
1835 | if (UNLIKELY(UNICODE_IS_SURROGATE(uv))) { | |
1836 | possible_problems |= UTF8_GOT_SURROGATE; | |
1837 | } | |
1838 | else if (UNLIKELY(uv > PERL_UNICODE_MAX)) { | |
1839 | possible_problems |= UTF8_GOT_SUPER; | |
1840 | } | |
1841 | else if (UNLIKELY(UNICODE_IS_NONCHAR(uv))) { | |
1842 | possible_problems |= UTF8_GOT_NONCHAR; | |
1843 | } | |
1844 | } | |
1845 | else { /* Otherwise, need to look at the source UTF-8, possibly | |
1846 | adjusted to be non-overlong */ | |
1847 | ||
1848 | if (UNLIKELY(NATIVE_UTF8_TO_I8(*adjusted_s0) | |
1849 | >= FIRST_START_BYTE_THAT_IS_DEFINITELY_SUPER)) | |
ea5ced44 | 1850 | { |
2b5e7bc2 KW |
1851 | possible_problems |= UTF8_GOT_SUPER; |
1852 | } | |
1853 | else if (curlen > 1) { | |
1854 | if (UNLIKELY(IS_UTF8_2_BYTE_SUPER( | |
1855 | NATIVE_UTF8_TO_I8(*adjusted_s0), | |
1856 | NATIVE_UTF8_TO_I8(*(adjusted_s0 + 1))))) | |
ea5ced44 | 1857 | { |
2b5e7bc2 | 1858 | possible_problems |= UTF8_GOT_SUPER; |
ea5ced44 | 1859 | } |
2b5e7bc2 KW |
1860 | else if (UNLIKELY(IS_UTF8_2_BYTE_SURROGATE( |
1861 | NATIVE_UTF8_TO_I8(*adjusted_s0), | |
1862 | NATIVE_UTF8_TO_I8(*(adjusted_s0 + 1))))) | |
1863 | { | |
1864 | possible_problems |= UTF8_GOT_SURROGATE; | |
ea5ced44 KW |
1865 | } |
1866 | } | |
c0236afe | 1867 | |
2b5e7bc2 KW |
1868 | /* We need a complete well-formed UTF-8 character to discern |
1869 | * non-characters, so can't look for them here */ | |
1870 | } | |
1871 | } | |
949cf498 | 1872 | |
2b5e7bc2 KW |
1873 | ready_to_handle_errors: |
1874 | ||
1875 | /* At this point: | |
1876 | * curlen contains the number of bytes in the sequence that | |
1877 | * this call should advance the input by. | |
e308b348 KW |
1878 | * avail_len gives the available number of bytes passed in, but |
1879 | * only if this is less than the expected number of | |
1880 | * bytes, based on the code point's start byte. | |
2b5e7bc2 KW |
1881 | * possible_problems' is 0 if there weren't any problems; otherwise a bit |
1882 | * is set in it for each potential problem found. | |
1883 | * uv contains the code point the input sequence | |
1884 | * represents; or if there is a problem that prevents | |
1885 | * a well-defined value from being computed, it is | |
1886 | * some subsitute value, typically the REPLACEMENT | |
1887 | * CHARACTER. | |
1888 | * s0 points to the first byte of the character | |
56576a04 KW |
1889 | * s points to just after were we left off processing |
1890 | * the character | |
1891 | * send points to just after where that character should | |
1892 | * end, based on how many bytes the start byte tells | |
1893 | * us should be in it, but no further than s0 + | |
1894 | * avail_len | |
2b5e7bc2 | 1895 | */ |
eb83ed87 | 1896 | |
2b5e7bc2 KW |
1897 | if (UNLIKELY(possible_problems)) { |
1898 | bool disallowed = FALSE; | |
1899 | const U32 orig_problems = possible_problems; | |
1900 | ||
37657a5b KW |
1901 | if (msgs) { |
1902 | *msgs = NULL; | |
1903 | } | |
1904 | ||
2b5e7bc2 KW |
1905 | while (possible_problems) { /* Handle each possible problem */ |
1906 | UV pack_warn = 0; | |
1907 | char * message = NULL; | |
37657a5b | 1908 | U32 this_flag_bit = 0; |
2b5e7bc2 KW |
1909 | |
1910 | /* Each 'if' clause handles one problem. They are ordered so that | |
1911 | * the first ones' messages will be displayed before the later | |
6c64cd9d KW |
1912 | * ones; this is kinda in decreasing severity order. But the |
1913 | * overlong must come last, as it changes 'uv' looked at by the | |
1914 | * others */ | |
2b5e7bc2 KW |
1915 | if (possible_problems & UTF8_GOT_OVERFLOW) { |
1916 | ||
56576a04 KW |
1917 | /* Overflow means also got a super and are using Perl's |
1918 | * extended UTF-8, but we handle all three cases here */ | |
2b5e7bc2 | 1919 | possible_problems |
d044b7a7 | 1920 | &= ~(UTF8_GOT_OVERFLOW|UTF8_GOT_SUPER|UTF8_GOT_PERL_EXTENDED); |
f9380377 KW |
1921 | *errors |= UTF8_GOT_OVERFLOW; |
1922 | ||
1923 | /* But the API says we flag all errors found */ | |
1924 | if (flags & (UTF8_WARN_SUPER|UTF8_DISALLOW_SUPER)) { | |
1925 | *errors |= UTF8_GOT_SUPER; | |
1926 | } | |
ddb65933 | 1927 | if (flags |
d044b7a7 | 1928 | & (UTF8_WARN_PERL_EXTENDED|UTF8_DISALLOW_PERL_EXTENDED)) |
ddb65933 | 1929 | { |
d044b7a7 | 1930 | *errors |= UTF8_GOT_PERL_EXTENDED; |
f9380377 | 1931 | } |
2b5e7bc2 | 1932 | |
d60baaa7 | 1933 | /* Disallow if any of the three categories say to */ |
56576a04 | 1934 | if ( ! (flags & UTF8_ALLOW_OVERFLOW) |
d60baaa7 | 1935 | || (flags & ( UTF8_DISALLOW_SUPER |
d044b7a7 | 1936 | |UTF8_DISALLOW_PERL_EXTENDED))) |
d60baaa7 KW |
1937 | { |
1938 | disallowed = TRUE; | |
1939 | } | |
1940 | ||
d22ec717 KW |
1941 | /* Likewise, warn if any say to */ |
1942 | if ( ! (flags & UTF8_ALLOW_OVERFLOW) | |
1943 | || (flags & (UTF8_WARN_SUPER|UTF8_WARN_PERL_EXTENDED))) | |
d60baaa7 | 1944 | { |
2b5e7bc2 | 1945 | |
ddb65933 KW |
1946 | /* The warnings code explicitly says it doesn't handle the |
1947 | * case of packWARN2 and two categories which have | |
1948 | * parent-child relationship. Even if it works now to | |
1949 | * raise the warning if either is enabled, it wouldn't | |
1950 | * necessarily do so in the future. We output (only) the | |
56576a04 | 1951 | * most dire warning */ |
ddb65933 | 1952 | if (! (flags & UTF8_CHECK_ONLY)) { |
37657a5b | 1953 | if (msgs || ckWARN_d(WARN_UTF8)) { |
ddb65933 KW |
1954 | pack_warn = packWARN(WARN_UTF8); |
1955 | } | |
37657a5b | 1956 | else if (msgs || ckWARN_d(WARN_NON_UNICODE)) { |
ddb65933 KW |
1957 | pack_warn = packWARN(WARN_NON_UNICODE); |
1958 | } | |
1959 | if (pack_warn) { | |
1960 | message = Perl_form(aTHX_ "%s: %s (overflows)", | |
1961 | malformed_text, | |
05b9033b | 1962 | _byte_dump_string(s0, curlen, 0)); |
37657a5b | 1963 | this_flag_bit = UTF8_GOT_OVERFLOW; |
ddb65933 | 1964 | } |
2b5e7bc2 KW |
1965 | } |
1966 | } | |
1967 | } | |
1968 | else if (possible_problems & UTF8_GOT_EMPTY) { | |
1969 | possible_problems &= ~UTF8_GOT_EMPTY; | |
f9380377 | 1970 | *errors |= UTF8_GOT_EMPTY; |
2b5e7bc2 KW |
1971 | |
1972 | if (! (flags & UTF8_ALLOW_EMPTY)) { | |
d1f8d421 KW |
1973 | |
1974 | /* This so-called malformation is now treated as a bug in | |
1975 | * the caller. If you have nothing to decode, skip calling | |
1976 | * this function */ | |
1977 | assert(0); | |
1978 | ||
2b5e7bc2 | 1979 | disallowed = TRUE; |
37657a5b KW |
1980 | if ( (msgs |
1981 | || ckWARN_d(WARN_UTF8)) && ! (flags & UTF8_CHECK_ONLY)) | |
1982 | { | |
2b5e7bc2 KW |
1983 | pack_warn = packWARN(WARN_UTF8); |
1984 | message = Perl_form(aTHX_ "%s (empty string)", | |
1985 | malformed_text); | |
37657a5b | 1986 | this_flag_bit = UTF8_GOT_EMPTY; |
2b5e7bc2 KW |
1987 | } |
1988 | } | |
1989 | } | |
1990 | else if (possible_problems & UTF8_GOT_CONTINUATION) { | |
1991 | possible_problems &= ~UTF8_GOT_CONTINUATION; | |
f9380377 | 1992 | *errors |= UTF8_GOT_CONTINUATION; |
2b5e7bc2 KW |
1993 | |
1994 | if (! (flags & UTF8_ALLOW_CONTINUATION)) { | |
1995 | disallowed = TRUE; | |
37657a5b KW |
1996 | if (( msgs |
1997 | || ckWARN_d(WARN_UTF8)) && ! (flags & UTF8_CHECK_ONLY)) | |
1998 | { | |
2b5e7bc2 KW |
1999 | pack_warn = packWARN(WARN_UTF8); |
2000 | message = Perl_form(aTHX_ | |
2001 | "%s: %s (unexpected continuation byte 0x%02x," | |
2002 | " with no preceding start byte)", | |
2003 | malformed_text, | |
7e2f38b2 | 2004 | _byte_dump_string(s0, 1, 0), *s0); |
37657a5b | 2005 | this_flag_bit = UTF8_GOT_CONTINUATION; |
2b5e7bc2 KW |
2006 | } |
2007 | } | |
2008 | } | |
2b5e7bc2 KW |
2009 | else if (possible_problems & UTF8_GOT_SHORT) { |
2010 | possible_problems &= ~UTF8_GOT_SHORT; | |
f9380377 | 2011 | *errors |= UTF8_GOT_SHORT; |
2b5e7bc2 KW |
2012 | |
2013 | if (! (flags & UTF8_ALLOW_SHORT)) { | |
2014 | disallowed = TRUE; | |
37657a5b KW |
2015 | if (( msgs |
2016 | || ckWARN_d(WARN_UTF8)) && ! (flags & UTF8_CHECK_ONLY)) | |
2017 | { | |
2b5e7bc2 KW |
2018 | pack_warn = packWARN(WARN_UTF8); |
2019 | message = Perl_form(aTHX_ | |
56576a04 KW |
2020 | "%s: %s (too short; %d byte%s available, need %d)", |
2021 | malformed_text, | |
2022 | _byte_dump_string(s0, send - s0, 0), | |
2023 | (int)avail_len, | |
2024 | avail_len == 1 ? "" : "s", | |
2025 | (int)expectlen); | |
37657a5b | 2026 | this_flag_bit = UTF8_GOT_SHORT; |
2b5e7bc2 KW |
2027 | } |
2028 | } | |
ba210ebe | 2029 | |
2b5e7bc2 | 2030 | } |
e308b348 KW |
2031 | else if (possible_problems & UTF8_GOT_NON_CONTINUATION) { |
2032 | possible_problems &= ~UTF8_GOT_NON_CONTINUATION; | |
2033 | *errors |= UTF8_GOT_NON_CONTINUATION; | |
2034 | ||
2035 | if (! (flags & UTF8_ALLOW_NON_CONTINUATION)) { | |
2036 | disallowed = TRUE; | |
37657a5b KW |
2037 | if (( msgs |
2038 | || ckWARN_d(WARN_UTF8)) && ! (flags & UTF8_CHECK_ONLY)) | |
2039 | { | |
99a765e9 KW |
2040 | |
2041 | /* If we don't know for sure that the input length is | |
2042 | * valid, avoid as much as possible reading past the | |
2043 | * end of the buffer */ | |
2044 | int printlen = (flags & _UTF8_NO_CONFIDENCE_IN_CURLEN) | |
2045 | ? s - s0 | |
2046 | : send - s0; | |
e308b348 KW |
2047 | pack_warn = packWARN(WARN_UTF8); |
2048 | message = Perl_form(aTHX_ "%s", | |
2049 | unexpected_non_continuation_text(s0, | |
99a765e9 | 2050 | printlen, |
e308b348 KW |
2051 | s - s0, |
2052 | (int) expectlen)); | |
37657a5b | 2053 | this_flag_bit = UTF8_GOT_NON_CONTINUATION; |
e308b348 KW |
2054 | } |
2055 | } | |
2056 | } | |
2b5e7bc2 KW |
2057 | else if (possible_problems & UTF8_GOT_SURROGATE) { |
2058 | possible_problems &= ~UTF8_GOT_SURROGATE; | |
2059 | ||
f9380377 KW |
2060 | if (flags & UTF8_WARN_SURROGATE) { |
2061 | *errors |= UTF8_GOT_SURROGATE; | |
2062 | ||
2063 | if ( ! (flags & UTF8_CHECK_ONLY) | |
37657a5b | 2064 | && (msgs || ckWARN_d(WARN_SURROGATE))) |
f9380377 | 2065 | { |
2b5e7bc2 KW |
2066 | pack_warn = packWARN(WARN_SURROGATE); |
2067 | ||
2068 | /* These are the only errors that can occur with a | |
2069 | * surrogate when the 'uv' isn't valid */ | |
2070 | if (orig_problems & UTF8_GOT_TOO_SHORT) { | |
2071 | message = Perl_form(aTHX_ | |
2072 | "UTF-16 surrogate (any UTF-8 sequence that" | |
2073 | " starts with \"%s\" is for a surrogate)", | |
7e2f38b2 | 2074 | _byte_dump_string(s0, curlen, 0)); |
2b5e7bc2 KW |
2075 | } |
2076 | else { | |
c94c2f39 | 2077 | message = Perl_form(aTHX_ surrogate_cp_format, uv); |
2b5e7bc2 | 2078 | } |
37657a5b | 2079 | this_flag_bit = UTF8_GOT_SURROGATE; |
f9380377 | 2080 | } |
2b5e7bc2 | 2081 | } |
ba210ebe | 2082 | |
2b5e7bc2 KW |
2083 | if (flags & UTF8_DISALLOW_SURROGATE) { |
2084 | disallowed = TRUE; | |
f9380377 | 2085 | *errors |= UTF8_GOT_SURROGATE; |
2b5e7bc2 KW |
2086 | } |
2087 | } | |
2088 | else if (possible_problems & UTF8_GOT_SUPER) { | |
2089 | possible_problems &= ~UTF8_GOT_SUPER; | |
949cf498 | 2090 | |
f9380377 KW |
2091 | if (flags & UTF8_WARN_SUPER) { |
2092 | *errors |= UTF8_GOT_SUPER; | |
2093 | ||
2094 | if ( ! (flags & UTF8_CHECK_ONLY) | |
37657a5b | 2095 | && (msgs || ckWARN_d(WARN_NON_UNICODE))) |
f9380377 | 2096 | { |
2b5e7bc2 KW |
2097 | pack_warn = packWARN(WARN_NON_UNICODE); |
2098 | ||
2099 | if (orig_problems & UTF8_GOT_TOO_SHORT) { | |
2100 | message = Perl_form(aTHX_ | |
2101 | "Any UTF-8 sequence that starts with" | |
2102 | " \"%s\" is for a non-Unicode code point," | |
2103 | " may not be portable", | |
7e2f38b2 | 2104 | _byte_dump_string(s0, curlen, 0)); |
2b5e7bc2 KW |
2105 | } |
2106 | else { | |
c94c2f39 | 2107 | message = Perl_form(aTHX_ super_cp_format, uv); |
2b5e7bc2 | 2108 | } |
37657a5b | 2109 | this_flag_bit = UTF8_GOT_SUPER; |
f9380377 | 2110 | } |
2b5e7bc2 | 2111 | } |
ba210ebe | 2112 | |
57ff5f59 KW |
2113 | /* Test for Perl's extended UTF-8 after the regular SUPER ones, |
2114 | * and before possibly bailing out, so that the more dire | |
2115 | * warning will override the regular one. */ | |
2116 | if (UNLIKELY(isUTF8_PERL_EXTENDED(s0))) { | |
2b5e7bc2 | 2117 | if ( ! (flags & UTF8_CHECK_ONLY) |
d044b7a7 | 2118 | && (flags & (UTF8_WARN_PERL_EXTENDED|UTF8_WARN_SUPER)) |
37657a5b | 2119 | && (msgs || ckWARN_d(WARN_NON_UNICODE))) |
2b5e7bc2 | 2120 | { |
db0f09e6 | 2121 | pack_warn = packWARN(WARN_NON_UNICODE); |
2b5e7bc2 | 2122 | |
57ff5f59 KW |
2123 | /* If it is an overlong that evaluates to a code point |
2124 | * that doesn't have to use the Perl extended UTF-8, it | |
2125 | * still used it, and so we output a message that | |
2126 | * doesn't refer to the code point. The same is true | |
2127 | * if there was a SHORT malformation where the code | |
2128 | * point is not valid. In that case, 'uv' will have | |
2129 | * been set to the REPLACEMENT CHAR, and the message | |
2130 | * below without the code point in it will be selected | |
2131 | * */ | |
2132 | if (UNICODE_IS_PERL_EXTENDED(uv)) { | |
2b5e7bc2 | 2133 | message = Perl_form(aTHX_ |
57ff5f59 | 2134 | perl_extended_cp_format, uv); |
2b5e7bc2 KW |
2135 | } |
2136 | else { | |
2137 | message = Perl_form(aTHX_ | |
57ff5f59 KW |
2138 | "Any UTF-8 sequence that starts with" |
2139 | " \"%s\" is a Perl extension, and" | |
2140 | " so is not portable", | |
2141 | _byte_dump_string(s0, curlen, 0)); | |
2b5e7bc2 | 2142 | } |
37657a5b | 2143 | this_flag_bit = UTF8_GOT_PERL_EXTENDED; |
2b5e7bc2 KW |
2144 | } |
2145 | ||
d044b7a7 KW |
2146 | if (flags & ( UTF8_WARN_PERL_EXTENDED |
2147 | |UTF8_DISALLOW_PERL_EXTENDED)) | |
ddb65933 | 2148 | { |
d044b7a7 | 2149 | *errors |= UTF8_GOT_PERL_EXTENDED; |
f9380377 | 2150 | |
d044b7a7 | 2151 | if (flags & UTF8_DISALLOW_PERL_EXTENDED) { |
f9380377 KW |
2152 | disallowed = TRUE; |
2153 | } | |
2b5e7bc2 KW |
2154 | } |
2155 | } | |
eb83ed87 | 2156 | |
2b5e7bc2 | 2157 | if (flags & UTF8_DISALLOW_SUPER) { |
f9380377 | 2158 | *errors |= UTF8_GOT_SUPER; |
2b5e7bc2 KW |
2159 | disallowed = TRUE; |
2160 | } | |
2b5e7bc2 KW |
2161 | } |
2162 | else if (possible_problems & UTF8_GOT_NONCHAR) { | |
2163 | possible_problems &= ~UTF8_GOT_NONCHAR; | |
ba210ebe | 2164 | |
f9380377 KW |
2165 | if (flags & UTF8_WARN_NONCHAR) { |
2166 | *errors |= UTF8_GOT_NONCHAR; | |
2167 | ||
2168 | if ( ! (flags & UTF8_CHECK_ONLY) | |
37657a5b | 2169 | && (msgs || ckWARN_d(WARN_NONCHAR))) |
f9380377 | 2170 | { |
2b5e7bc2 KW |
2171 | /* The code above should have guaranteed that we don't |
2172 | * get here with errors other than overlong */ | |
2173 | assert (! (orig_problems | |
2174 | & ~(UTF8_GOT_LONG|UTF8_GOT_NONCHAR))); | |
2175 | ||
2176 | pack_warn = packWARN(WARN_NONCHAR); | |
c94c2f39 | 2177 | message = Perl_form(aTHX_ nonchar_cp_format, uv); |
37657a5b | 2178 | this_flag_bit = UTF8_GOT_NONCHAR; |
f9380377 | 2179 | } |
2b5e7bc2 | 2180 | } |
5b311467 | 2181 | |
2b5e7bc2 KW |
2182 | if (flags & UTF8_DISALLOW_NONCHAR) { |
2183 | disallowed = TRUE; | |
f9380377 | 2184 | *errors |= UTF8_GOT_NONCHAR; |
2b5e7bc2 | 2185 | } |
6c64cd9d KW |
2186 | } |
2187 | else if (possible_problems & UTF8_GOT_LONG) { | |
2188 | possible_problems &= ~UTF8_GOT_LONG; | |
2189 | *errors |= UTF8_GOT_LONG; | |
2190 | ||
2191 | if (flags & UTF8_ALLOW_LONG) { | |
2192 | ||
2193 | /* We don't allow the actual overlong value, unless the | |
2194 | * special extra bit is also set */ | |
2195 | if (! (flags & ( UTF8_ALLOW_LONG_AND_ITS_VALUE | |
2196 | & ~UTF8_ALLOW_LONG))) | |
2197 | { | |
2198 | uv = UNICODE_REPLACEMENT; | |
2199 | } | |
2200 | } | |
2201 | else { | |
2202 | disallowed = TRUE; | |
2203 | ||
37657a5b KW |
2204 | if (( msgs |
2205 | || ckWARN_d(WARN_UTF8)) && ! (flags & UTF8_CHECK_ONLY)) | |
2206 | { | |
6c64cd9d KW |
2207 | pack_warn = packWARN(WARN_UTF8); |
2208 | ||
2209 | /* These error types cause 'uv' to be something that | |
2210 | * isn't what was intended, so can't use it in the | |
2211 | * message. The other error types either can't | |
2212 | * generate an overlong, or else the 'uv' is valid */ | |
2213 | if (orig_problems & | |
2214 | (UTF8_GOT_TOO_SHORT|UTF8_GOT_OVERFLOW)) | |
2215 | { | |
2216 | message = Perl_form(aTHX_ | |
2217 | "%s: %s (any UTF-8 sequence that starts" | |
2218 | " with \"%s\" is overlong which can and" | |
2219 | " should be represented with a" | |
2220 | " different, shorter sequence)", | |
2221 | malformed_text, | |
2222 | _byte_dump_string(s0, send - s0, 0), | |
2223 | _byte_dump_string(s0, curlen, 0)); | |
2224 | } | |
2225 | else { | |
2226 | U8 tmpbuf[UTF8_MAXBYTES+1]; | |
1be62ab9 KW |
2227 | const U8 * const e = uvoffuni_to_utf8_flags(tmpbuf, |
2228 | uv, 0); | |
d819dc50 KW |
2229 | /* Don't use U+ for non-Unicode code points, which |
2230 | * includes those in the Latin1 range */ | |
2231 | const char * preface = ( uv > PERL_UNICODE_MAX | |
2232 | #ifdef EBCDIC | |
2233 | || uv <= 0xFF | |
2234 | #endif | |
2235 | ) | |
2236 | ? "0x" | |
2237 | : "U+"; | |
6c64cd9d KW |
2238 | message = Perl_form(aTHX_ |
2239 | "%s: %s (overlong; instead use %s to represent" | |
2240 | " %s%0*" UVXf ")", | |
2241 | malformed_text, | |
2242 | _byte_dump_string(s0, send - s0, 0), | |
2243 | _byte_dump_string(tmpbuf, e - tmpbuf, 0), | |
2244 | preface, | |
2245 | ((uv < 256) ? 2 : 4), /* Field width of 2 for | |
2246 | small code points */ | |
1be62ab9 | 2247 | UNI_TO_NATIVE(uv)); |
6c64cd9d | 2248 | } |
37657a5b | 2249 | this_flag_bit = UTF8_GOT_LONG; |
6c64cd9d KW |
2250 | } |
2251 | } | |
2b5e7bc2 KW |
2252 | } /* End of looking through the possible flags */ |
2253 | ||
2254 | /* Display the message (if any) for the problem being handled in | |
2255 | * this iteration of the loop */ | |
2256 | if (message) { | |
37657a5b | 2257 | if (msgs) { |
37657a5b KW |
2258 | assert(this_flag_bit); |
2259 | ||
2260 | if (*msgs == NULL) { | |
2261 | *msgs = newAV(); | |
2262 | } | |
2263 | ||
bb07812e KW |
2264 | av_push(*msgs, newRV_noinc((SV*) new_msg_hv(message, |
2265 | pack_warn, | |
2266 | this_flag_bit))); | |
37657a5b KW |
2267 | } |
2268 | else if (PL_op) | |
2b5e7bc2 KW |
2269 | Perl_warner(aTHX_ pack_warn, "%s in %s", message, |
2270 | OP_DESC(PL_op)); | |
2271 | else | |
2272 | Perl_warner(aTHX_ pack_warn, "%s", message); | |
2273 | } | |
ddb65933 | 2274 | } /* End of 'while (possible_problems)' */ |
a0dbb045 | 2275 | |
2b5e7bc2 KW |
2276 | /* Since there was a possible problem, the returned length may need to |
2277 | * be changed from the one stored at the beginning of this function. | |
2278 | * Instead of trying to figure out if that's needed, just do it. */ | |
2279 | if (retlen) { | |
2280 | *retlen = curlen; | |
2281 | } | |
a0dbb045 | 2282 | |
2b5e7bc2 KW |
2283 | if (disallowed) { |
2284 | if (flags & UTF8_CHECK_ONLY && retlen) { | |
2285 | *retlen = ((STRLEN) -1); | |
2286 | } | |
2287 | return 0; | |
2288 | } | |
eb83ed87 | 2289 | } |
ba210ebe | 2290 | |
2b5e7bc2 | 2291 | return UNI_TO_NATIVE(uv); |
a0ed51b3 LW |
2292 | } |
2293 | ||
8e84507e | 2294 | /* |
ec5f19d0 KW |
2295 | =for apidoc utf8_to_uvchr_buf |
2296 | ||
2297 | Returns the native code point of the first character in the string C<s> which | |
2298 | is assumed to be in UTF-8 encoding; C<send> points to 1 beyond the end of C<s>. | |
524080c4 | 2299 | C<*retlen> will be set to the length, in bytes, of that character. |
ec5f19d0 | 2300 | |
524080c4 KW |
2301 | If C<s> does not point to a well-formed UTF-8 character and UTF8 warnings are |
2302 | enabled, zero is returned and C<*retlen> is set (if C<retlen> isn't | |
796b6530 | 2303 | C<NULL>) to -1. If those warnings are off, the computed value, if well-defined |
173db420 | 2304 | (or the Unicode REPLACEMENT CHARACTER if not), is silently returned, and |
796b6530 | 2305 | C<*retlen> is set (if C<retlen> isn't C<NULL>) so that (S<C<s> + C<*retlen>>) is |
173db420 | 2306 | the next possible position in C<s> that could begin a non-malformed character. |
de69f3af | 2307 | See L</utf8n_to_uvchr> for details on when the REPLACEMENT CHARACTER is |
173db420 | 2308 | returned. |
ec5f19d0 KW |
2309 | |
2310 | =cut | |
52be2536 KW |
2311 | |
2312 | Also implemented as a macro in utf8.h | |
2313 | ||
ec5f19d0 KW |
2314 | */ |
2315 | ||
2316 | ||
2317 | UV | |
2318 | Perl_utf8_to_uvchr_buf(pTHX_ const U8 *s, const U8 *send, STRLEN *retlen) | |
2319 | { | |
7f974d7e KW |
2320 | PERL_ARGS_ASSERT_UTF8_TO_UVCHR_BUF; |
2321 | ||
ec5f19d0 KW |
2322 | assert(s < send); |
2323 | ||
2324 | return utf8n_to_uvchr(s, send - s, retlen, | |
ddb65933 | 2325 | ckWARN_d(WARN_UTF8) ? 0 : UTF8_ALLOW_ANY); |
ec5f19d0 KW |
2326 | } |
2327 | ||
52be2536 KW |
2328 | /* This is marked as deprecated |
2329 | * | |
ec5f19d0 KW |
2330 | =for apidoc utf8_to_uvuni_buf |
2331 | ||
de69f3af KW |
2332 | Only in very rare circumstances should code need to be dealing in Unicode |
2333 | (as opposed to native) code points. In those few cases, use | |
07dfe0a4 KW |
2334 | C<L<NATIVE_TO_UNI(utf8_to_uvchr_buf(...))|/utf8_to_uvchr_buf>> instead. If you |
2335 | are not absolutely sure this is one of those cases, then assume it isn't and | |
2336 | use plain C<utf8_to_uvchr_buf> instead. | |
4f83cdcd KW |
2337 | |
2338 | Returns the Unicode (not-native) code point of the first character in the | |
2339 | string C<s> which | |
ec5f19d0 KW |
2340 | is assumed to be in UTF-8 encoding; C<send> points to 1 beyond the end of C<s>. |
2341 | C<retlen> will be set to the length, in bytes, of that character. | |
2342 | ||
524080c4 KW |
2343 | If C<s> does not point to a well-formed UTF-8 character and UTF8 warnings are |
2344 | enabled, zero is returned and C<*retlen> is set (if C<retlen> isn't | |
2345 | NULL) to -1. If those warnings are off, the computed value if well-defined (or | |
2346 | the Unicode REPLACEMENT CHARACTER, if not) is silently returned, and C<*retlen> | |
2347 | is set (if C<retlen> isn't NULL) so that (S<C<s> + C<*retlen>>) is the | |
2348 | next possible position in C<s> that could begin a non-malformed character. | |
de69f3af | 2349 | See L</utf8n_to_uvchr> for details on when the REPLACEMENT CHARACTER is returned. |
ec5f19d0 KW |
2350 | |
2351 | =cut | |
2352 | */ | |
2353 | ||
2354 | UV | |
2355 | Perl_utf8_to_uvuni_buf(pTHX_ const U8 *s, const U8 *send, STRLEN *retlen) | |
2356 | { | |
2357 | PERL_ARGS_ASSERT_UTF8_TO_UVUNI_BUF; | |
2358 | ||
2359 | assert(send > s); | |
2360 | ||
5962d97e | 2361 | return NATIVE_TO_UNI(utf8_to_uvchr_buf(s, send, retlen)); |
ec5f19d0 KW |
2362 | } |
2363 | ||
b76347f2 | 2364 | /* |
87cea99e | 2365 | =for apidoc utf8_length |
b76347f2 | 2366 | |
b2e7ed74 KW |
2367 | Returns the number of characters in the sequence of UTF-8-encoded bytes starting |
2368 | at C<s> and ending at the byte just before C<e>. If <s> and <e> point to the | |
2369 | same place, it returns 0 with no warning raised. | |
2370 | ||
2371 | If C<e E<lt> s> or if the scan would end up past C<e>, it raises a UTF8 warning | |
2372 | and returns the number of valid characters. | |
b76347f2 JH |
2373 | |
2374 | =cut | |
2375 | */ | |
2376 | ||
2377 | STRLEN | |
35a4481c | 2378 | Perl_utf8_length(pTHX_ const U8 *s, const U8 *e) |
b76347f2 JH |
2379 | { |
2380 | STRLEN len = 0; | |
2381 | ||
7918f24d NC |
2382 | PERL_ARGS_ASSERT_UTF8_LENGTH; |
2383 | ||
8850bf83 JH |
2384 | /* Note: cannot use UTF8_IS_...() too eagerly here since e.g. |
2385 | * the bitops (especially ~) can create illegal UTF-8. | |
2386 | * In other words: in Perl UTF-8 is not just for Unicode. */ | |
2387 | ||
a3b680e6 AL |
2388 | if (e < s) |
2389 | goto warn_and_return; | |
b76347f2 | 2390 | while (s < e) { |
4cbf4130 | 2391 | s += UTF8SKIP(s); |
8e91ec7f AV |
2392 | len++; |
2393 | } | |
2394 | ||
2395 | if (e != s) { | |
2396 | len--; | |
2397 | warn_and_return: | |
9b387841 NC |
2398 | if (PL_op) |
2399 | Perl_ck_warner_d(aTHX_ packWARN(WARN_UTF8), | |
2400 | "%s in %s", unees, OP_DESC(PL_op)); | |
2401 | else | |
61a12c31 | 2402 | Perl_ck_warner_d(aTHX_ packWARN(WARN_UTF8), "%s", unees); |
b76347f2 JH |
2403 | } |
2404 | ||
2405 | return len; | |
2406 | } | |
2407 | ||
b06226ff | 2408 | /* |
fed3ba5d NC |
2409 | =for apidoc bytes_cmp_utf8 |
2410 | ||
a1433954 | 2411 | Compares the sequence of characters (stored as octets) in C<b>, C<blen> with the |
72d33970 FC |
2412 | sequence of characters (stored as UTF-8) |
2413 | in C<u>, C<ulen>. Returns 0 if they are | |
fed3ba5d NC |
2414 | equal, -1 or -2 if the first string is less than the second string, +1 or +2 |
2415 | if the first string is greater than the second string. | |
2416 | ||
2417 | -1 or +1 is returned if the shorter string was identical to the start of the | |
72d33970 FC |
2418 | longer string. -2 or +2 is returned if |
2419 | there was a difference between characters | |
fed3ba5d NC |
2420 | within the strings. |
2421 | ||
2422 | =cut | |
2423 | */ | |
2424 | ||
2425 | int | |
2426 | Perl_bytes_cmp_utf8(pTHX_ const U8 *b, STRLEN blen, const U8 *u, STRLEN ulen) | |
2427 | { | |
2428 | const U8 *const bend = b + blen; | |
2429 | const U8 *const uend = u + ulen; | |
2430 | ||
2431 | PERL_ARGS_ASSERT_BYTES_CMP_UTF8; | |
fed3ba5d NC |
2432 | |
2433 | while (b < bend && u < uend) { | |
2434 | U8 c = *u++; | |
2435 | if (!UTF8_IS_INVARIANT(c)) { | |
2436 | if (UTF8_IS_DOWNGRADEABLE_START(c)) { | |
2437 | if (u < uend) { | |
2438 | U8 c1 = *u++; | |
2439 | if (UTF8_IS_CONTINUATION(c1)) { | |
a62b247b | 2440 | c = EIGHT_BIT_UTF8_TO_NATIVE(c, c1); |
fed3ba5d | 2441 | } else { |
2b5e7bc2 | 2442 | /* diag_listed_as: Malformed UTF-8 character%s */ |
fed3ba5d | 2443 | Perl_ck_warner_d(aTHX_ packWARN(WARN_UTF8), |
56576a04 KW |
2444 | "%s %s%s", |
2445 | unexpected_non_continuation_text(u - 2, 2, 1, 2), | |
2446 | PL_op ? " in " : "", | |
2447 | PL_op ? OP_DESC(PL_op) : ""); | |
fed3ba5d NC |
2448 | return -2; |
2449 | } | |
2450 | } else { | |
2451 | if (PL_op) | |
2452 | Perl_ck_warner_d(aTHX_ packWARN(WARN_UTF8), | |
2453 | "%s in %s", unees, OP_DESC(PL_op)); | |
2454 | else | |
61a12c31 | 2455 | Perl_ck_warner_d(aTHX_ packWARN(WARN_UTF8), "%s", unees); |
fed3ba5d NC |
2456 | return -2; /* Really want to return undef :-) */ |
2457 | } | |
2458 | } else { | |
2459 | return -2; | |
2460 | } | |
2461 | } | |
2462 | if (*b != c) { | |
2463 | return *b < c ? -2 : +2; | |
2464 | } | |
2465 | ++b; | |
2466 | } | |
2467 | ||
2468 | if (b == bend && u == uend) | |
2469 | return 0; | |
2470 | ||
2471 | return b < bend ? +1 : -1; | |
2472 | } | |
2473 | ||
2474 | /* | |
87cea99e | 2475 | =for apidoc utf8_to_bytes |
6940069f | 2476 | |
3bc0c78c | 2477 | Converts a string C<"s"> of length C<*lenp> from UTF-8 into native byte encoding. |
a1433954 | 2478 | Unlike L</bytes_to_utf8>, this over-writes the original string, and |
09af0336 | 2479 | updates C<*lenp> to contain the new length. |
3bc0c78c KW |
2480 | Returns zero on failure (leaving C<"s"> unchanged) setting C<*lenp> to -1. |
2481 | ||
2482 | Upon successful return, the number of variants in the string can be computed by | |
23b37b12 KW |
2483 | having saved the value of C<*lenp> before the call, and subtracting the |
2484 | after-call value of C<*lenp> from it. | |
6940069f | 2485 | |
a1433954 | 2486 | If you need a copy of the string, see L</bytes_from_utf8>. |
95be277c | 2487 | |
6940069f GS |
2488 | =cut |
2489 | */ | |
2490 | ||
2491 | U8 * | |
09af0336 | 2492 | Perl_utf8_to_bytes(pTHX_ U8 *s, STRLEN *lenp) |
6940069f | 2493 | { |
9fe0d3c2 | 2494 | U8 * first_variant; |
246fae53 | 2495 | |
7918f24d | 2496 | PERL_ARGS_ASSERT_UTF8_TO_BYTES; |
81611534 | 2497 | PERL_UNUSED_CONTEXT; |
7918f24d | 2498 | |
9fe0d3c2 | 2499 | /* This is a no-op if no variants at all in the input */ |
09af0336 | 2500 | if (is_utf8_invariant_string_loc(s, *lenp, (const U8 **) &first_variant)) { |
9fe0d3c2 KW |
2501 | return s; |
2502 | } | |
2503 | ||
2504 | { | |
3c5aa262 | 2505 | U8 * const save = s; |
09af0336 | 2506 | U8 * const send = s + *lenp; |
3c5aa262 KW |
2507 | U8 * d; |
2508 | ||
2509 | /* Nothing before the first variant needs to be changed, so start the real | |
2510 | * work there */ | |
2511 | s = first_variant; | |
2512 | while (s < send) { | |
2513 | if (! UTF8_IS_INVARIANT(*s)) { | |
2514 | if (! UTF8_IS_NEXT_CHAR_DOWNGRADEABLE(s, send)) { | |
09af0336 | 2515 | *lenp = ((STRLEN) -1); |
3c5aa262 KW |
2516 | return 0; |
2517 | } | |
2518 | s++; | |
d59937ca KW |
2519 | } |
2520 | s++; | |
dcad2880 | 2521 | } |
dcad2880 | 2522 | |
3c5aa262 KW |
2523 | /* Is downgradable, so do it */ |
2524 | d = s = first_variant; | |
2525 | while (s < send) { | |
2526 | U8 c = *s++; | |
2527 | if (! UVCHR_IS_INVARIANT(c)) { | |
2528 | /* Then it is two-byte encoded */ | |
2529 | c = EIGHT_BIT_UTF8_TO_NATIVE(c, *s); | |
2530 | s++; | |
2531 | } | |
2532 | *d++ = c; | |
2533 | } | |
2534 | *d = '\0'; | |
09af0336 | 2535 | *lenp = d - save; |
3c5aa262 KW |
2536 | |
2537 | return save; | |
9fe0d3c2 | 2538 | } |
6940069f GS |
2539 | } |
2540 | ||
2541 | /* | |
87cea99e | 2542 | =for apidoc bytes_from_utf8 |
f9a63242 | 2543 | |
09af0336 | 2544 | Converts a potentially UTF-8 encoded string C<s> of length C<*lenp> into native |
41ae6089 | 2545 | byte encoding. On input, the boolean C<*is_utf8p> gives whether or not C<s> is |
4f3d592d KW |
2546 | actually encoded in UTF-8. |
2547 | ||
2548 | Unlike L</utf8_to_bytes> but like L</bytes_to_utf8>, this is non-destructive of | |
2549 | the input string. | |
2550 | ||
41ae6089 KW |
2551 | Do nothing if C<*is_utf8p> is 0, or if there are code points in the string |
2552 | not expressible in native byte encoding. In these cases, C<*is_utf8p> and | |
09af0336 | 2553 | C<*lenp> are unchanged, and the return value is the original C<s>. |
4f3d592d | 2554 | |
41ae6089 | 2555 | Otherwise, C<*is_utf8p> is set to 0, and the return value is a pointer to a |
4f3d592d | 2556 | newly created string containing a downgraded copy of C<s>, and whose length is |
23b37b12 | 2557 | returned in C<*lenp>, updated. The new string is C<NUL>-terminated. |
f9a63242 | 2558 | |
3bc0c78c | 2559 | Upon successful return, the number of variants in the string can be computed by |
23b37b12 KW |
2560 | having saved the value of C<*lenp> before the call, and subtracting the |
2561 | after-call value of C<*lenp> from it. | |
3bc0c78c | 2562 | |
37607a96 | 2563 | =cut |
976c1b08 KW |
2564 | |
2565 | There is a macro that avoids this function call, but this is retained for | |
2566 | anyone who calls it with the Perl_ prefix */ | |
f9a63242 JH |
2567 | |
2568 | U8 * | |
41ae6089 | 2569 | Perl_bytes_from_utf8(pTHX_ const U8 *s, STRLEN *lenp, bool *is_utf8p) |
f9a63242 | 2570 | { |
7918f24d | 2571 | PERL_ARGS_ASSERT_BYTES_FROM_UTF8; |
96a5add6 | 2572 | PERL_UNUSED_CONTEXT; |
f9a63242 | 2573 | |
976c1b08 KW |
2574 | return bytes_from_utf8_loc(s, lenp, is_utf8p, NULL); |
2575 | } | |
2576 | ||
2577 | /* | |
2578 | No = here because currently externally undocumented | |
2579 | for apidoc bytes_from_utf8_loc | |
2580 | ||
2581 | Like C<L</bytes_from_utf8>()>, but takes an extra parameter, a pointer to where | |
2582 | to store the location of the first character in C<"s"> that cannot be | |
2583 | converted to non-UTF8. | |
2584 | ||
2585 | If that parameter is C<NULL>, this function behaves identically to | |
2586 | C<bytes_from_utf8>. | |
2587 | ||
2588 | Otherwise if C<*is_utf8p> is 0 on input, the function behaves identically to | |
2589 | C<bytes_from_utf8>, except it also sets C<*first_non_downgradable> to C<NULL>. | |
2590 | ||
2591 | Otherwise, the function returns a newly created C<NUL>-terminated string | |
2592 | containing the non-UTF8 equivalent of the convertible first portion of | |
2593 | C<"s">. C<*lenp> is set to its length, not including the terminating C<NUL>. | |
2594 | If the entire input string was converted, C<*is_utf8p> is set to a FALSE value, | |
2595 | and C<*first_non_downgradable> is set to C<NULL>. | |
2596 | ||
2597 | Otherwise, C<*first_non_downgradable> set to point to the first byte of the | |
2598 | first character in the original string that wasn't converted. C<*is_utf8p> is | |
2599 | unchanged. Note that the new string may have length 0. | |
2600 | ||
2601 | Another way to look at it is, if C<*first_non_downgradable> is non-C<NULL> and | |
2602 | C<*is_utf8p> is TRUE, this function starts at the beginning of C<"s"> and | |
2603 | converts as many characters in it as possible stopping at the first one it | |
385b74be | 2604 | finds that can't be converted to non-UTF-8. C<*first_non_downgradable> is |
976c1b08 KW |
2605 | set to point to that. The function returns the portion that could be converted |
2606 | in a newly created C<NUL>-terminated string, and C<*lenp> is set to its length, | |
2607 | not including the terminating C<NUL>. If the very first character in the | |
2608 | original could not be converted, C<*lenp> will be 0, and the new string will | |
2609 | contain just a single C<NUL>. If the entire input string was converted, | |
2610 | C<*is_utf8p> is set to FALSE and C<*first_non_downgradable> is set to C<NULL>. | |
2611 | ||
2612 | Upon successful return, the number of variants in the converted portion of the | |
2613 | string can be computed by having saved the value of C<*lenp> before the call, | |
2614 | and subtracting the after-call value of C<*lenp> from it. | |
2615 | ||
2616 | =cut | |
2617 | ||
2618 | ||
2619 | */ | |
2620 | ||
2621 | U8 * | |
2622 | Perl_bytes_from_utf8_loc(const U8 *s, STRLEN *lenp, bool *is_utf8p, const U8** first_unconverted) | |
2623 | { | |
2624 | U8 *d; | |
2625 | const U8 *original = s; | |
2626 | U8 *converted_start; | |
2627 | const U8 *send = s + *lenp; | |
f9a63242 | 2628 | |
976c1b08 | 2629 | PERL_ARGS_ASSERT_BYTES_FROM_UTF8_LOC; |
170a1c22 | 2630 | |
976c1b08 KW |
2631 | if (! *is_utf8p) { |
2632 | if (first_unconverted) { | |
2633 | *first_unconverted = NULL; | |
2634 | } | |
2635 | ||
2636 | return (U8 *) original; | |
2637 | } | |
2638 | ||
2639 | Newx(d, (*lenp) + 1, U8); | |
2640 | ||
2641 | converted_start = d; | |
7299a045 KW |
2642 | while (s < send) { |
2643 | U8 c = *s++; | |
2644 | if (! UTF8_IS_INVARIANT(c)) { | |
976c1b08 KW |
2645 | |
2646 | /* Then it is multi-byte encoded. If the code point is above 0xFF, | |
2647 | * have to stop now */ | |
2648 | if (UNLIKELY (! UTF8_IS_NEXT_CHAR_DOWNGRADEABLE(s - 1, send))) { | |
2649 | if (first_unconverted) { | |
2650 | *first_unconverted = s - 1; | |
2651 | goto finish_and_return; | |
2652 | } | |
2653 | else { | |
2654 | Safefree(converted_start); | |
2655 | return (U8 *) original; | |
2656 | } | |
2657 | } | |
2658 | ||
7299a045 KW |
2659 | c = EIGHT_BIT_UTF8_TO_NATIVE(c, *s); |
2660 | s++; | |
38af28cf | 2661 | } |
7299a045 KW |
2662 | *d++ = c; |
2663 | } | |
170a1c22 | 2664 | |
976c1b08 KW |
2665 | /* Here, converted the whole of the input */ |
2666 | *is_utf8p = FALSE; | |
2667 | if (first_unconverted) { | |
2668 | *first_unconverted = NULL; | |
170a1c22 | 2669 | } |
976c1b08 KW |
2670 | |
2671 | finish_and_return: | |
46a08a6f KW |
2672 | *d = '\0'; |
2673 | *lenp = d - converted_start; | |
976c1b08 KW |
2674 | |
2675 | /* Trim unused space */ | |
2676 | Renew(converted_start, *lenp + 1, U8); | |
2677 | ||
2678 | return converted_start; | |
f9a63242 JH |
2679 | } |
2680 | ||
2681 | /* | |
87cea99e | 2682 | =for apidoc bytes_to_utf8 |
6940069f | 2683 | |
09af0336 | 2684 | Converts a string C<s> of length C<*lenp> bytes from the native encoding into |
ff97e5cf | 2685 | UTF-8. |
09af0336 | 2686 | Returns a pointer to the newly-created string, and sets C<*lenp> to |
ff97e5cf | 2687 | reflect the new length in bytes. |
6940069f | 2688 | |
3bc0c78c | 2689 | Upon successful return, the number of variants in the string can be computed by |
23b37b12 | 2690 | having saved the value of C<*lenp> before the call, and subtracting it from the |
3bc0c78c KW |
2691 | after-call value of C<*lenp>. |
2692 | ||
75200dff | 2693 | A C<NUL> character will be written after the end of the string. |
2bbc8d55 SP |
2694 | |
2695 | If you want to convert to UTF-8 from encodings other than | |
2696 | the native (Latin1 or EBCDIC), | |
a1433954 | 2697 | see L</sv_recode_to_utf8>(). |
c9ada85f | 2698 | |
497711e7 | 2699 | =cut |
6940069f GS |
2700 | */ |
2701 | ||
2702 | U8* | |
09af0336 | 2703 | Perl_bytes_to_utf8(pTHX_ const U8 *s, STRLEN *lenp) |
6940069f | 2704 | { |
09af0336 | 2705 | const U8 * const send = s + (*lenp); |
6940069f GS |
2706 | U8 *d; |
2707 | U8 *dst; | |
7918f24d NC |
2708 | |
2709 | PERL_ARGS_ASSERT_BYTES_TO_UTF8; | |
96a5add6 | 2710 | PERL_UNUSED_CONTEXT; |
6940069f | 2711 | |
09af0336 | 2712 | Newx(d, (*lenp) * 2 + 1, U8); |
6940069f GS |
2713 | dst = d; |
2714 | ||
2715 | while (s < send) { | |
55d09dc8 KW |
2716 | append_utf8_from_native_byte(*s, &d); |
2717 | s++; | |
6940069f | 2718 | } |
2e11cf67 | 2719 | |
6940069f | 2720 | *d = '\0'; |
09af0336 | 2721 | *lenp = d-dst; |
2e11cf67 KW |
2722 | |
2723 | /* Trim unused space */ | |
2724 | Renew(dst, *lenp + 1, U8); | |
2725 | ||
6940069f GS |
2726 | return dst; |
2727 | } | |
2728 | ||
a0ed51b3 | 2729 | /* |
624504c5 KW |
2730 | * Convert native (big-endian) UTF-16 to UTF-8. For reversed (little-endian), |
2731 | * use utf16_to_utf8_reversed(). | |
a0ed51b3 | 2732 | * |
624504c5 KW |
2733 | * UTF-16 requires 2 bytes for every code point below 0x10000; otherwise 4 bytes. |
2734 | * UTF-8 requires 1-3 bytes for every code point below 0x1000; otherwise 4 bytes. | |
2735 | * UTF-EBCDIC requires 1-4 bytes for every code point below 0x1000; otherwise 4-5 bytes. | |
2736 | * | |
2737 | * These functions don't check for overflow. The worst case is every code | |
2738 | * point in the input is 2 bytes, and requires 4 bytes on output. (If the code | |
2739 | * is never going to run in EBCDIC, it is 2 bytes requiring 3 on output.) Therefore the | |
2740 | * destination must be pre-extended to 2 times the source length. | |
2741 | * | |
2742 | * Do not use in-place. We optimize for native, for obvious reasons. */ | |
a0ed51b3 LW |
2743 | |
2744 | U8* | |
dea0fc0b | 2745 | Perl_utf16_to_utf8(pTHX_ U8* p, U8* d, I32 bytelen, I32 *newlen) |
a0ed51b3 | 2746 | { |
dea0fc0b JH |
2747 | U8* pend; |
2748 | U8* dstart = d; | |
2749 | ||
7918f24d NC |
2750 | PERL_ARGS_ASSERT_UTF16_TO_UTF8; |
2751 | ||
dea0fc0b | 2752 | if (bytelen & 1) |
56576a04 KW |
2753 | Perl_croak(aTHX_ "panic: utf16_to_utf8: odd bytelen %" UVuf, |
2754 | (UV)bytelen); | |
dea0fc0b JH |
2755 | |
2756 | pend = p + bytelen; | |
2757 | ||
a0ed51b3 | 2758 | while (p < pend) { |
dea0fc0b JH |
2759 | UV uv = (p[0] << 8) + p[1]; /* UTF-16BE */ |
2760 | p += 2; | |
2d1545e5 | 2761 | if (OFFUNI_IS_INVARIANT(uv)) { |
56d37426 | 2762 | *d++ = LATIN1_TO_NATIVE((U8) uv); |
a0ed51b3 LW |
2763 | continue; |
2764 | } | |
56d37426 KW |
2765 | if (uv <= MAX_UTF8_TWO_BYTE) { |
2766 | *d++ = UTF8_TWO_BYTE_HI(UNI_TO_NATIVE(uv)); | |
2767 | *d++ = UTF8_TWO_BYTE_LO(UNI_TO_NATIVE(uv)); | |
a0ed51b3 LW |
2768 | continue; |
2769 | } | |
ffd0a9d3 | 2770 | |
46956fad KW |
2771 | #define FIRST_HIGH_SURROGATE UNICODE_SURROGATE_FIRST |
2772 | #define LAST_HIGH_SURROGATE 0xDBFF | |
2773 | #define FIRST_LOW_SURROGATE 0xDC00 | |
2774 | #define LAST_LOW_SURROGATE UNICODE_SURROGATE_LAST | |
ffd0a9d3 | 2775 | #define FIRST_IN_PLANE1 0x10000 |
e23c50db KW |
2776 | |
2777 | /* This assumes that most uses will be in the first Unicode plane, not | |
2778 | * needing surrogates */ | |
2779 | if (UNLIKELY(uv >= UNICODE_SURROGATE_FIRST | |
2780 | && uv <= UNICODE_SURROGATE_LAST)) | |
2781 | { | |
2782 | if (UNLIKELY(p >= pend) || UNLIKELY(uv > LAST_HIGH_SURROGATE)) { | |
2783 | Perl_croak(aTHX_ "Malformed UTF-16 surrogate"); | |
2784 | } | |
2785 | else { | |
01ea242b | 2786 | UV low = (p[0] << 8) + p[1]; |
e23c50db KW |
2787 | if ( UNLIKELY(low < FIRST_LOW_SURROGATE) |
2788 | || UNLIKELY(low > LAST_LOW_SURROGATE)) | |
2789 | { | |
01ea242b | 2790 | Perl_croak(aTHX_ "Malformed UTF-16 surrogate"); |
e23c50db KW |
2791 | } |
2792 | p += 2; | |
46956fad | 2793 | uv = ((uv - FIRST_HIGH_SURROGATE) << 10) |
ffd0a9d3 | 2794 | + (low - FIRST_LOW_SURROGATE) + FIRST_IN_PLANE1; |
01ea242b | 2795 | } |
a0ed51b3 | 2796 | } |
56d37426 KW |
2797 | #ifdef EBCDIC |
2798 | d = uvoffuni_to_utf8_flags(d, uv, 0); | |
2799 | #else | |
ffd0a9d3 | 2800 | if (uv < FIRST_IN_PLANE1) { |
eb160463 GS |
2801 | *d++ = (U8)(( uv >> 12) | 0xe0); |
2802 | *d++ = (U8)(((uv >> 6) & 0x3f) | 0x80); | |
2803 | *d++ = (U8)(( uv & 0x3f) | 0x80); | |
a0ed51b3 LW |
2804 | continue; |
2805 | } | |
2806 | else { | |
eb160463 GS |
2807 | *d++ = (U8)(( uv >> 18) | 0xf0); |
2808 | *d++ = (U8)(((uv >> 12) & 0x3f) | 0x80); | |
2809 | *d++ = (U8)(((uv >> 6) & 0x3f) | 0x80); | |
2810 | *d++ = (U8)(( uv & 0x3f) | 0x80); | |
a0ed51b3 LW |
2811 | continue; |
2812 | } | |
56d37426 | 2813 | #endif |
a0ed51b3 | 2814 | } |
dea0fc0b | 2815 | *newlen = d - dstart; |
a0ed51b3 LW |
2816 | return d; |
2817 | } | |
2818 | ||
2819 | /* Note: this one is slightly destructive of the source. */ | |
2820 | ||
2821 | U8* | |
dea0fc0b | 2822 | Perl_utf16_to_utf8_reversed(pTHX_ U8* p, U8* d, I32 bytelen, I32 *newlen) |
a0ed51b3 LW |
2823 | { |
2824 | U8* s = (U8*)p; | |
d4c19fe8 | 2825 | U8* const send = s + bytelen; |
7918f24d NC |
2826 | |
2827 | PERL_ARGS_ASSERT_UTF16_TO_UTF8_REVERSED; | |
2828 | ||
e0ea5e2d | 2829 | if (bytelen & 1) |
147e3846 | 2830 | Perl_croak(aTHX_ "panic: utf16_to_utf8_reversed: odd bytelen %" UVuf, |
e0ea5e2d NC |
2831 | (UV)bytelen); |
2832 | ||
a0ed51b3 | 2833 | while (s < send) { |
d4c19fe8 | 2834 | const U8 tmp = s[0]; |
a0ed51b3 LW |
2835 | s[0] = s[1]; |
2836 | s[1] = tmp; | |
2837 | s += 2; | |
2838 | } | |
dea0fc0b | 2839 | return utf16_to_utf8(p, d, bytelen, newlen); |
a0ed51b3 LW |
2840 | } |
2841 | ||
922e8cb4 KW |
2842 | bool |
2843 | Perl__is_uni_FOO(pTHX_ const U8 classnum, const UV c) | |
2844 | { | |
2845 | U8 tmpbuf[UTF8_MAXBYTES+1]; | |
2846 | uvchr_to_utf8(tmpbuf, c); | |
da8c1a98 | 2847 | return _is_utf8_FOO_with_len(classnum, tmpbuf, tmpbuf + sizeof(tmpbuf)); |
922e8cb4 KW |
2848 | } |
2849 | ||
f9ae8fb6 JD |
2850 | /* Internal function so we can deprecate the external one, and call |
2851 | this one from other deprecated functions in this file */ | |
2852 | ||
f2645549 KW |
2853 | bool |
2854 | Perl__is_utf8_idstart(pTHX_ const U8 *p) | |
61b19385 | 2855 | { |
f2645549 | 2856 | PERL_ARGS_ASSERT__IS_UTF8_IDSTART; |
61b19385 KW |
2857 | |
2858 | if (*p == '_') | |
2859 | return TRUE; | |
f25ce844 | 2860 | return is_utf8_common(p, &PL_utf8_idstart, "IdStart", NULL); |
61b19385 KW |
2861 | } |
2862 | ||
5092f92a | 2863 | bool |
eba68aa0 KW |
2864 | Perl__is_uni_perl_idcont(pTHX_ UV c) |
2865 | { | |
2866 | U8 tmpbuf[UTF8_MAXBYTES+1]; | |
2867 | uvchr_to_utf8(tmpbuf, c); | |
da8c1a98 | 2868 | return _is_utf8_perl_idcont_with_len(tmpbuf, tmpbuf + sizeof(tmpbuf)); |
eba68aa0 KW |
2869 | } |
2870 | ||
2871 | bool | |
f91dcd13 KW |
2872 | Perl__is_uni_perl_idstart(pTHX_ UV c) |
2873 | { | |
2874 | U8 tmpbuf[UTF8_MAXBYTES+1]; | |
2875 | uvchr_to_utf8(tmpbuf, c); | |
da8c1a98 | 2876 | return _is_utf8_perl_idstart_with_len(tmpbuf, tmpbuf + sizeof(tmpbuf)); |
f91dcd13 KW |
2877 | } |
2878 | ||
3a4c58c9 | 2879 | UV |
56576a04 KW |
2880 | Perl__to_upper_title_latin1(pTHX_ const U8 c, U8* p, STRLEN *lenp, |
2881 | const char S_or_s) | |
3a4c58c9 KW |
2882 | { |
2883 | /* We have the latin1-range values compiled into the core, so just use | |
4a4088c4 | 2884 | * those, converting the result to UTF-8. The only difference between upper |
3a4c58c9 KW |
2885 | * and title case in this range is that LATIN_SMALL_LETTER_SHARP_S is |
2886 | * either "SS" or "Ss". Which one to use is passed into the routine in | |
2887 | * 'S_or_s' to avoid a test */ | |
2888 | ||
2889 | UV converted = toUPPER_LATIN1_MOD(c); | |
2890 | ||
2891 | PERL_ARGS_ASSERT__TO_UPPER_TITLE_LATIN1; | |
2892 | ||
2893 | assert(S_or_s == 'S' || S_or_s == 's'); | |
2894 | ||
6f2d5cbc | 2895 | if (UVCHR_IS_INVARIANT(converted)) { /* No difference between the two for |
f4cd282c | 2896 | characters in this range */ |
3a4c58c9 KW |
2897 | *p = (U8) converted; |
2898 | *lenp = 1; | |
2899 | return converted; | |
2900 | } | |
2901 | ||
2902 | /* toUPPER_LATIN1_MOD gives the correct results except for three outliers, | |
2903 | * which it maps to one of them, so as to only have to have one check for | |
2904 | * it in the main case */ | |
2905 | if (UNLIKELY(converted == LATIN_SMALL_LETTER_Y_WITH_DIAERESIS)) { | |
2906 | switch (c) { | |
2907 | case LATIN_SMALL_LETTER_Y_WITH_DIAERESIS: | |
2908 | converted = LATIN_CAPITAL_LETTER_Y_WITH_DIAERESIS; | |
2909 | break; | |
2910 | case MICRO_SIGN: | |
2911 | converted = GREEK_CAPITAL_LETTER_MU; | |
2912 | break; | |
79e064b9 KW |
2913 | #if UNICODE_MAJOR_VERSION > 2 \ |
2914 | || (UNICODE_MAJOR_VERSION == 2 && UNICODE_DOT_VERSION >= 1 \ | |
2915 | && UNICODE_DOT_DOT_VERSION >= 8) | |
3a4c58c9 KW |
2916 | case LATIN_SMALL_LETTER_SHARP_S: |
2917 | *(p)++ = 'S'; | |
2918 | *p = S_or_s; | |
2919 | *lenp = 2; | |
2920 | return 'S'; | |
79e064b9 | 2921 | #endif |
3a4c58c9 | 2922 | default: |
56576a04 KW |
2923 | Perl_croak(aTHX_ "panic: to_upper_title_latin1 did not expect" |
2924 | " '%c' to map to '%c'", | |
2925 | c, LATIN_SMALL_LETTER_Y_WITH_DIAERESIS); | |
e5964223 | 2926 | NOT_REACHED; /* NOTREACHED */ |
3a4c58c9 KW |
2927 | } |
2928 | } | |
2929 | ||
2930 | *(p)++ = UTF8_TWO_BYTE_HI(converted); | |
2931 | *p = UTF8_TWO_BYTE_LO(converted); | |
2932 | *lenp = 2; | |
2933 | ||
2934 | return converted; | |
2935 | } | |
2936 | ||
50bda2c3 KW |
2937 | /* Call the function to convert a UTF-8 encoded character to the specified case. |
2938 | * Note that there may be more than one character in the result. | |
2939 | * INP is a pointer to the first byte of the input character | |
2940 | * OUTP will be set to the first byte of the string of changed characters. It | |
2941 | * needs to have space for UTF8_MAXBYTES_CASE+1 bytes | |
2942 | * LENP will be set to the length in bytes of the string of changed characters | |
2943 | * | |
56576a04 KW |
2944 | * The functions return the ordinal of the first character in the string of |
2945 | * OUTP */ | |
2946 | #define CALL_UPPER_CASE(uv, s, d, lenp) \ | |
2947 | _to_utf8_case(uv, s, d, lenp, &PL_utf8_toupper, "ToUc", "") | |
2948 | #define CALL_TITLE_CASE(uv, s, d, lenp) \ | |
2949 | _to_utf8_case(uv, s, d, lenp, &PL_utf8_totitle, "ToTc", "") | |
2950 | #define CALL_LOWER_CASE(uv, s, d, lenp) \ | |
2951 | _to_utf8_case(uv, s, d, lenp, &PL_utf8_tolower, "ToLc", "") | |
50bda2c3 | 2952 | |
b9992569 KW |
2953 | /* This additionally has the input parameter 'specials', which if non-zero will |
2954 | * cause this to use the specials hash for folding (meaning get full case | |
50bda2c3 | 2955 | * folding); otherwise, when zero, this implies a simple case fold */ |
56576a04 KW |
2956 | #define CALL_FOLD_CASE(uv, s, d, lenp, specials) \ |
2957 | _to_utf8_case(uv, s, d, lenp, &PL_utf8_tofold, "ToCf", (specials) ? "" : NULL) | |
c3fd2246 | 2958 | |
84afefe6 JH |
2959 | UV |
2960 | Perl_to_uni_upper(pTHX_ UV c, U8* p, STRLEN *lenp) | |
a0ed51b3 | 2961 | { |
a1433954 KW |
2962 | /* Convert the Unicode character whose ordinal is <c> to its uppercase |
2963 | * version and store that in UTF-8 in <p> and its length in bytes in <lenp>. | |
2964 | * Note that the <p> needs to be at least UTF8_MAXBYTES_CASE+1 bytes since | |
c3fd2246 KW |
2965 | * the changed version may be longer than the original character. |
2966 | * | |
2967 | * The ordinal of the first character of the changed version is returned | |
2968 | * (but note, as explained above, that there may be more.) */ | |
2969 | ||
7918f24d NC |
2970 | PERL_ARGS_ASSERT_TO_UNI_UPPER; |
2971 | ||
3a4c58c9 KW |
2972 | if (c < 256) { |
2973 | return _to_upper_title_latin1((U8) c, p, lenp, 'S'); | |
2974 | } | |
2975 | ||
0ebc6274 | 2976 | uvchr_to_utf8(p, c); |
b9992569 | 2977 | return CALL_UPPER_CASE(c, p, p, lenp); |
a0ed51b3 LW |
2978 | } |
2979 | ||
84afefe6 JH |
2980 | UV |
2981 | Perl_to_uni_title(pTHX_ UV c, U8* p, STRLEN *lenp) | |
a0ed51b3 | 2982 | { |
7918f24d NC |
2983 | PERL_ARGS_ASSERT_TO_UNI_TITLE; |
2984 | ||
3a4c58c9 KW |
2985 | if (c < 256) { |
2986 | return _to_upper_title_latin1((U8) c, p, lenp, 's'); | |
2987 | } | |
2988 | ||
0ebc6274 | 2989 | uvchr_to_utf8(p, c); |
b9992569 | 2990 | return CALL_TITLE_CASE(c, p, p, lenp); |
a0ed51b3 LW |
2991 | } |
2992 | ||
afc16117 | 2993 | STATIC U8 |
eaf412bf | 2994 | S_to_lower_latin1(const U8 c, U8* p, STRLEN *lenp, const char dummy) |
afc16117 KW |
2995 | { |
2996 | /* We have the latin1-range values compiled into the core, so just use | |
4a4088c4 | 2997 | * those, converting the result to UTF-8. Since the result is always just |
a1433954 | 2998 | * one character, we allow <p> to be NULL */ |
afc16117 KW |
2999 | |
3000 | U8 converted = toLOWER_LATIN1(c); | |
3001 | ||
eaf412bf KW |
3002 | PERL_UNUSED_ARG(dummy); |
3003 | ||
afc16117 | 3004 | if (p != NULL) { |
6f2d5cbc | 3005 | if (NATIVE_BYTE_IS_INVARIANT(converted)) { |
afc16117 KW |
3006 | *p = converted; |
3007 | *lenp = 1; | |
3008 | } | |
3009 | else { | |
430c9760 KW |
3010 | /* Result is known to always be < 256, so can use the EIGHT_BIT |
3011 | * macros */ | |
3012 | *p = UTF8_EIGHT_BIT_HI(converted); | |
3013 | *(p+1) = UTF8_EIGHT_BIT_LO(converted); | |
afc16117 KW |
3014 | *lenp = 2; |
3015 | } | |
3016 | } | |
3017 | return converted; | |
3018 | } | |
3019 | ||
84afefe6 JH |
3020 | UV |
3021 | Perl_to_uni_lower(pTHX_ UV c, U8* p, STRLEN *lenp) | |
a0ed51b3 | 3022 | { |
7918f24d NC |
3023 | PERL_ARGS_ASSERT_TO_UNI_LOWER; |
3024 | ||
afc16117 | 3025 | if (c < 256) { |
eaf412bf | 3026 | return to_lower_latin1((U8) c, p, lenp, 0 /* 0 is a dummy arg */ ); |
bca00c02 KW |
3027 | } |
3028 | ||
afc16117 | 3029 | uvchr_to_utf8(p, c); |
b9992569 | 3030 | return CALL_LOWER_CASE(c, p, p, lenp); |
a0ed51b3 LW |
3031 | } |
3032 | ||
84afefe6 | 3033 | UV |
56576a04 KW |
3034 | Perl__to_fold_latin1(pTHX_ const U8 c, U8* p, STRLEN *lenp, |
3035 | const unsigned int flags) | |
a1dde8de | 3036 | { |
51910141 | 3037 | /* Corresponds to to_lower_latin1(); <flags> bits meanings: |
1ca267a5 | 3038 | * FOLD_FLAGS_NOMIX_ASCII iff non-ASCII to ASCII folds are prohibited |
51910141 | 3039 | * FOLD_FLAGS_FULL iff full folding is to be used; |
1ca267a5 KW |
3040 | * |
3041 | * Not to be used for locale folds | |
51910141 | 3042 | */ |
f673fad4 | 3043 | |
a1dde8de KW |
3044 | UV converted; |
3045 | ||
3046 | PERL_ARGS_ASSERT__TO_FOLD_LATIN1; | |
81611534 | 3047 | PERL_UNUSED_CONTEXT; |
a1dde8de | 3048 | |
1ca267a5 KW |
3049 | assert (! (flags & FOLD_FLAGS_LOCALE)); |
3050 | ||
659a7c2d | 3051 | if (UNLIKELY(c == MICRO_SIGN)) { |
a1dde8de KW |
3052 | converted = GREEK_SMALL_LETTER_MU; |
3053 | } | |
9b63e895 KW |
3054 | #if UNICODE_MAJOR_VERSION > 3 /* no multifolds in early Unicode */ \ |
3055 | || (UNICODE_MAJOR_VERSION == 3 && ( UNICODE_DOT_VERSION > 0) \ | |
3056 | || UNICODE_DOT_DOT_VERSION > 0) | |
659a7c2d KW |
3057 | else if ( (flags & FOLD_FLAGS_FULL) |
3058 | && UNLIKELY(c == LATIN_SMALL_LETTER_SHARP_S)) | |
3059 | { | |
1ca267a5 KW |
3060 | /* If can't cross 127/128 boundary, can't return "ss"; instead return |
3061 | * two U+017F characters, as fc("\df") should eq fc("\x{17f}\x{17f}") | |
3062 | * under those circumstances. */ | |
3063 | if (flags & FOLD_FLAGS_NOMIX_ASCII) { | |
3064 | *lenp = 2 * sizeof(LATIN_SMALL_LETTER_LONG_S_UTF8) - 2; | |
3065 | Copy(LATIN_SMALL_LETTER_LONG_S_UTF8 LATIN_SMALL_LETTER_LONG_S_UTF8, | |
3066 | p, *lenp, U8); | |
3067 | return LATIN_SMALL_LETTER_LONG_S; | |
3068 | } | |
3069 | else { | |
4f489194 KW |
3070 | *(p)++ = 's'; |
3071 | *p = 's'; | |
3072 | *lenp = 2; | |
3073 | return 's'; | |
1ca267a5 | 3074 | } |
a1dde8de | 3075 | } |
9b63e895 | 3076 | #endif |
a1dde8de KW |
3077 | else { /* In this range the fold of all other characters is their lower |
3078 | case */ | |
3079 | converted = toLOWER_LATIN1(c); | |
3080 | } | |
3081 | ||
6f2d5cbc | 3082 | if (UVCHR_IS_INVARIANT(converted)) { |
a1dde8de KW |
3083 | *p = (U8) converted; |
3084 | *lenp = 1; | |
3085 | } | |
3086 | else { | |
3087 | *(p)++ = UTF8_TWO_BYTE_HI(converted); | |
3088 | *p = UTF8_TWO_BYTE_LO(converted); | |
3089 | *lenp = 2; | |
3090 | } | |
3091 | ||
3092 | return converted; | |
3093 | } | |
3094 | ||
3095 | UV | |
31f05a37 | 3096 | Perl__to_uni_fold_flags(pTHX_ UV c, U8* p, STRLEN *lenp, U8 flags) |
84afefe6 | 3097 | { |
4b593389 | 3098 | |
a0270393 KW |
3099 | /* Not currently externally documented, and subject to change |
3100 | * <flags> bits meanings: | |
3101 | * FOLD_FLAGS_FULL iff full folding is to be used; | |
31f05a37 KW |
3102 | * FOLD_FLAGS_LOCALE is set iff the rules from the current underlying |
3103 | * locale are to be used. | |
a0270393 KW |
3104 | * FOLD_FLAGS_NOMIX_ASCII iff non-ASCII to ASCII folds are prohibited |
3105 | */ | |
4b593389 | 3106 | |
36bb2ab6 | 3107 | PERL_ARGS_ASSERT__TO_UNI_FOLD_FLAGS; |
7918f24d | 3108 | |
780fcc9f | 3109 | if (flags & FOLD_FLAGS_LOCALE) { |
8b7358b9 KW |
3110 | /* Treat a UTF-8 locale as not being in locale at all, except for |
3111 | * potentially warning */ | |
3112 | _CHECK_AND_WARN_PROBLEMATIC_LOCALE; | |
780fcc9f KW |
3113 | if (IN_UTF8_CTYPE_LOCALE) { |
3114 | flags &= ~FOLD_FLAGS_LOCALE; | |
3115 | } | |
3116 | else { | |
e7b7ac46 | 3117 | goto needs_full_generality; |
780fcc9f | 3118 | } |
31f05a37 KW |
3119 | } |
3120 | ||
a1dde8de | 3121 | if (c < 256) { |
e7b7ac46 | 3122 | return _to_fold_latin1((U8) c, p, lenp, |
31f05a37 | 3123 | flags & (FOLD_FLAGS_FULL | FOLD_FLAGS_NOMIX_ASCII)); |
a1dde8de KW |
3124 | } |
3125 | ||
2f306ab9 | 3126 | /* Here, above 255. If no special needs, just use the macro */ |
a0270393 KW |
3127 | if ( ! (flags & (FOLD_FLAGS_LOCALE|FOLD_FLAGS_NOMIX_ASCII))) { |
3128 | uvchr_to_utf8(p, c); | |
b9992569 | 3129 | return CALL_FOLD_CASE(c, p, p, lenp, flags & FOLD_FLAGS_FULL); |
a0270393 | 3130 | } |
567b353c | 3131 | else { /* Otherwise, _toFOLD_utf8_flags has the intelligence to deal with |
a0270393 KW |
3132 | the special flags. */ |
3133 | U8 utf8_c[UTF8_MAXBYTES + 1]; | |
e7b7ac46 KW |
3134 | |
3135 | needs_full_generality: | |
a0270393 | 3136 | uvchr_to_utf8(utf8_c, c); |
56576a04 KW |
3137 | return _toFOLD_utf8_flags(utf8_c, utf8_c + sizeof(utf8_c), |
3138 | p, lenp, flags); | |
a0270393 | 3139 | } |
84afefe6 JH |
3140 | } |
3141 | ||
26483009 | 3142 | PERL_STATIC_INLINE bool |
5141f98e | 3143 | S_is_utf8_common(pTHX_ const U8 *const p, SV **swash, |
f25ce844 | 3144 | const char *const swashname, SV* const invlist) |
bde6a22d | 3145 | { |
ea317ccb KW |
3146 | /* returns a boolean giving whether or not the UTF8-encoded character that |
3147 | * starts at <p> is in the swash indicated by <swashname>. <swash> | |
3148 | * contains a pointer to where the swash indicated by <swashname> | |
3149 | * is to be stored; which this routine will do, so that future calls will | |
f25ce844 KW |
3150 | * look at <*swash> and only generate a swash if it is not null. <invlist> |
3151 | * is NULL or an inversion list that defines the swash. If not null, it | |
3152 | * saves time during initialization of the swash. | |
ea317ccb KW |
3153 | * |
3154 | * Note that it is assumed that the buffer length of <p> is enough to | |
3155 | * contain all the bytes that comprise the character. Thus, <*p> should | |
3156 | * have been checked before this call for mal-formedness enough to assure | |
3157 | * that. */ | |
3158 | ||
7918f24d NC |
3159 | PERL_ARGS_ASSERT_IS_UTF8_COMMON; |
3160 | ||
492a624f | 3161 | /* The API should have included a length for the UTF-8 character in <p>, |
28123549 | 3162 | * but it doesn't. We therefore assume that p has been validated at least |
492a624f KW |
3163 | * as far as there being enough bytes available in it to accommodate the |
3164 | * character without reading beyond the end, and pass that number on to the | |
3165 | * validating routine */ | |
6302f837 | 3166 | if (! isUTF8_CHAR(p, p + UTF8SKIP(p))) { |
86ae6e94 | 3167 | _force_out_malformed_utf8_message(p, p + UTF8SKIP(p), |
99a765e9 | 3168 | _UTF8_NO_CONFIDENCE_IN_CURLEN, |
86ae6e94 KW |
3169 | 1 /* Die */ ); |
3170 | NOT_REACHED; /* NOTREACHED */ | |
28123549 | 3171 | } |
86ae6e94 | 3172 | |
87367d5f KW |
3173 | if (!*swash) { |
3174 | U8 flags = _CORE_SWASH_INIT_ACCEPT_INVLIST; | |
f25ce844 KW |
3175 | *swash = _core_swash_init("utf8", |
3176 | ||
3177 | /* Only use the name if there is no inversion | |
3178 | * list; otherwise will go out to disk */ | |
3179 | (invlist) ? "" : swashname, | |
3180 | ||
3181 | &PL_sv_undef, 1, 0, invlist, &flags); | |
87367d5f | 3182 | } |
28123549 | 3183 | |
bde6a22d NC |
3184 | return swash_fetch(*swash, p, TRUE) != 0; |
3185 | } | |
3186 | ||
da8c1a98 | 3187 | PERL_STATIC_INLINE bool |
56576a04 KW |
3188 | S_is_utf8_common_with_len(pTHX_ const U8 *const p, const U8 * const e, |
3189 | SV **swash, const char *const swashname, | |
3190 | SV* const invlist) | |
da8c1a98 KW |
3191 | { |
3192 | /* returns a boolean giving whether or not the UTF8-encoded character that | |
3193 | * starts at <p>, and extending no further than <e - 1> is in the swash | |
3194 | * indicated by <swashname>. <swash> contains a pointer to where the swash | |
3195 | * indicated by <swashname> is to be stored; which this routine will do, so | |
3196 | * that future calls will look at <*swash> and only generate a swash if it | |
3197 | * is not null. <invlist> is NULL or an inversion list that defines the | |
3198 | * swash. If not null, it saves time during initialization of the swash. | |
3199 | */ | |
3200 | ||
3201 | PERL_ARGS_ASSERT_IS_UTF8_COMMON_WITH_LEN; | |
3202 | ||
3203 | if (! isUTF8_CHAR(p, e)) { | |
3204 | _force_out_malformed_utf8_message(p, e, 0, 1); | |
3205 | NOT_REACHED; /* NOTREACHED */ | |
3206 | } | |
3207 | ||
3208 | if (!*swash) { | |
3209 | U8 flags = _CORE_SWASH_INIT_ACCEPT_INVLIST; | |
3210 | *swash = _core_swash_init("utf8", | |
3211 | ||
3212 | /* Only use the name if there is no inversion | |
3213 | * list; otherwise will go out to disk */ | |
3214 | (invlist) ? "" : swashname, | |
3215 | ||
3216 | &PL_sv_undef, 1, 0, invlist, &flags); | |
3217 | } | |
3218 | ||
3219 | return swash_fetch(*swash, p, TRUE) != 0; | |
3220 | } | |
3221 | ||
34aeb2e9 KW |
3222 | STATIC void |
3223 | S_warn_on_first_deprecated_use(pTHX_ const char * const name, | |
3224 | const char * const alternative, | |
3225 | const bool use_locale, | |
3226 | const char * const file, | |
3227 | const unsigned line) | |
3228 | { | |
3229 | const char * key; | |
3230 | ||
3231 | PERL_ARGS_ASSERT_WARN_ON_FIRST_DEPRECATED_USE; | |
3232 | ||
3233 | if (ckWARN_d(WARN_DEPRECATED)) { | |
3234 | ||
3235 | key = Perl_form(aTHX_ "%s;%d;%s;%d", name, use_locale, file, line); | |
3236 | if (! hv_fetch(PL_seen_deprecated_macro, key, strlen(key), 0)) { | |
3237 | if (! PL_seen_deprecated_macro) { | |
3238 | PL_seen_deprecated_macro = newHV(); | |
3239 | } | |
3240 | if (! hv_store(PL_seen_deprecated_macro, key, | |
3241 | strlen(key), &PL_sv_undef, 0)) | |
3242 | { | |
3243 | Perl_croak(aTHX_ "panic: hv_store() unexpectedly failed"); | |
3244 | } | |
3245 | ||
c44e9413 | 3246 | if (instr(file, "mathoms.c")) { |
607313a1 KW |
3247 | Perl_warner(aTHX_ WARN_DEPRECATED, |
3248 | "In %s, line %d, starting in Perl v5.30, %s()" | |
3249 | " will be removed. Avoid this message by" | |
3250 | " converting to use %s().\n", | |
3251 | file, line, name, alternative); | |
3252 | } | |
3253 | else { | |
34aeb2e9 KW |
3254 | Perl_warner(aTHX_ WARN_DEPRECATED, |
3255 | "In %s, line %d, starting in Perl v5.30, %s() will" | |
3256 | " require an additional parameter. Avoid this" | |
3257 | " message by converting to use %s().\n", | |
3258 | file, line, name, alternative); | |
607313a1 | 3259 | } |
34aeb2e9 KW |
3260 | } |
3261 | } | |
3262 | } | |
3263 | ||
bde6a22d | 3264 | bool |
34aeb2e9 | 3265 | Perl__is_utf8_FOO(pTHX_ U8 classnum, |
be99e2c2 | 3266 | const U8 * const p, |
34aeb2e9 KW |
3267 | const char * const name, |
3268 | const char * const alternative, | |
3269 | const bool use_utf8, | |
3270 | const bool use_locale, | |
3271 | const char * const file, | |
3272 | const unsigned line) | |
922e8cb4 | 3273 | { |
922e8cb4 KW |
3274 | PERL_ARGS_ASSERT__IS_UTF8_FOO; |
3275 | ||
34aeb2e9 KW |
3276 | warn_on_first_deprecated_use(name, alternative, use_locale, file, line); |
3277 | ||
3278 | if (use_utf8 && UTF8_IS_ABOVE_LATIN1(*p)) { | |
34aeb2e9 KW |
3279 | |
3280 | switch (classnum) { | |
3281 | case _CC_WORDCHAR: | |
3282 | case _CC_DIGIT: | |
3283 | case _CC_ALPHA: | |
3284 | case _CC_LOWER: | |
3285 | case _CC_UPPER: | |
3286 | case _CC_PUNCT: | |
3287 | case _CC_PRINT: | |
3288 | case _CC_ALPHANUMERIC: | |
3289 | case _CC_GRAPH: | |
3290 | case _CC_CASED: | |
3291 | ||
3292 | return is_utf8_common(p, | |
3293 | &PL_utf8_swash_ptrs[classnum], | |
3294 | swash_property_names[classnum], | |
3295 | PL_XPosix_ptrs[classnum]); | |
3296 | ||
3297 | case _CC_SPACE: | |
3298 | return is_XPERLSPACE_high(p); | |
3299 | case _CC_BLANK: | |
3300 | return is_HORIZWS_high(p); | |
3301 | case _CC_XDIGIT: | |
3302 | return is_XDIGIT_high(p); | |
3303 | case _CC_CNTRL: | |
3304 | return 0; | |
3305 | case _CC_ASCII: | |
3306 | return 0; | |
3307 | case _CC_VERTSPACE: | |
3308 | return is_VERTWS_high(p); | |
3309 | case _CC_IDFIRST: | |
3310 | if (! PL_utf8_perl_idstart) { | |
22f0498f KW |
3311 | PL_utf8_perl_idstart |
3312 | = _new_invlist_C_array(_Perl_IDStart_invlist); | |
34aeb2e9 | 3313 | } |
22f0498f KW |
3314 | return is_utf8_common(p, &PL_utf8_perl_idstart, |
3315 | "_Perl_IDStart", NULL); | |
34aeb2e9 KW |
3316 | case _CC_IDCONT: |
3317 | if (! PL_utf8_perl_idcont) { | |
22f0498f KW |
3318 | PL_utf8_perl_idcont |
3319 | = _new_invlist_C_array(_Perl_IDCont_invlist); | |
34aeb2e9 | 3320 | } |
22f0498f KW |
3321 | return is_utf8_common(p, &PL_utf8_perl_idcont, |
3322 | "_Perl_IDCont", NULL); | |
34aeb2e9 KW |
3323 | } |
3324 | } | |
3325 | ||
3326 | /* idcont is the same as wordchar below 256 */ | |
3327 | if (classnum == _CC_IDCONT) { | |
3328 | classnum = _CC_WORDCHAR; | |
3329 | } | |
3330 | else if (classnum == _CC_IDFIRST) { | |
3331 | if (*p == '_') { | |
3332 | return TRUE; | |
3333 | } | |
3334 | classnum = _CC_ALPHA; | |
3335 | } | |
3336 | ||
3337 | if (! use_locale) { | |
3338 | if (! use_utf8 || UTF8_IS_INVARIANT(*p)) { | |
3339 | return _generic_isCC(*p, classnum); | |
3340 | } | |
922e8cb4 | 3341 | |
34aeb2e9 KW |
3342 | return _generic_isCC(EIGHT_BIT_UTF8_TO_NATIVE(*p, *(p + 1 )), classnum); |
3343 | } | |
3344 | else { | |
3345 | if (! use_utf8 || UTF8_IS_INVARIANT(*p)) { | |
3346 | return isFOO_lc(classnum, *p); | |
3347 | } | |
3348 | ||
3349 | return isFOO_lc(classnum, EIGHT_BIT_UTF8_TO_NATIVE(*p, *(p + 1 ))); | |
3350 | } | |
3351 | ||
3352 | NOT_REACHED; /* NOTREACHED */ | |
922e8cb4 KW |
3353 | } |
3354 | ||
3355 | bool | |
da8c1a98 KW |
3356 | Perl__is_utf8_FOO_with_len(pTHX_ const U8 classnum, const U8 *p, |
3357 | const U8 * const e) | |
3358 | { | |
3359 | PERL_ARGS_ASSERT__IS_UTF8_FOO_WITH_LEN; | |
3360 | ||
3361 | assert(classnum < _FIRST_NON_SWASH_CC); | |
3362 | ||
3363 | return is_utf8_common_with_len(p, | |
3364 | e, | |
3365 | &PL_utf8_swash_ptrs[classnum], | |
3366 | swash_property_names[classnum], | |
3367 | PL_XPosix_ptrs[classnum]); | |
3368 | } | |
3369 | ||
3370 | bool | |
da8c1a98 KW |
3371 | Perl__is_utf8_perl_idstart_with_len(pTHX_ const U8 *p, const U8 * const e) |
3372 | { | |
3373 | SV* invlist = NULL; | |
3374 | ||
3375 | PERL_ARGS_ASSERT__IS_UTF8_PERL_IDSTART_WITH_LEN; | |
3376 | ||
3377 | if (! PL_utf8_perl_idstart) { | |
3378 | invlist = _new_invlist_C_array(_Perl_IDStart_invlist); | |
3379 | } | |
3380 | return is_utf8_common_with_len(p, e, &PL_utf8_perl_idstart, | |
3381 | "_Perl_IDStart", invlist); | |
3382 | } | |
3383 | ||
3384 | bool | |
f2645549 | 3385 | Perl__is_utf8_xidstart(pTHX_ const U8 *p) |
c11ff943 | 3386 | { |
f2645549 | 3387 | PERL_ARGS_ASSERT__IS_UTF8_XIDSTART; |
c11ff943 KW |
3388 | |
3389 | if (*p == '_') | |
3390 | return TRUE; | |
f25ce844 | 3391 | return is_utf8_common(p, &PL_utf8_xidstart, "XIdStart", NULL); |
c11ff943 KW |
3392 | } |
3393 | ||
3394 | bool | |
da8c1a98 KW |
3395 | Perl__is_utf8_perl_idcont_with_len(pTHX_ const U8 *p, const U8 * const e) |
3396 | { | |
3397 | SV* invlist = NULL; | |
3398 | ||
3399 | PERL_ARGS_ASSERT__IS_UTF8_PERL_IDCONT_WITH_LEN; | |
3400 | ||
3401 | if (! PL_utf8_perl_idcont) { | |
3402 | invlist = _new_invlist_C_array(_Perl_IDCont_invlist); | |
3403 | } | |
3404 | return is_utf8_common_with_len(p, e, &PL_utf8_perl_idcont, | |
3405 | "_Perl_IDCont", invlist); | |
3406 | } | |
3407 | ||
3408 | bool | |
f2645549 | 3409 | Perl__is_utf8_idcont(pTHX_ const U8 *p) |
82686b01 | 3410 | { |
f2645549 | 3411 | PERL_ARGS_ASSERT__IS_UTF8_IDCONT; |
7918f24d | 3412 | |
f25ce844 | 3413 | return is_utf8_common(p, &PL_utf8_idcont, "IdContinue", NULL); |
a0ed51b3 LW |
3414 | } |
3415 | ||
3416 | bool | |
f2645549 | 3417 | Perl__is_utf8_xidcont(pTHX_ const U8 *p) |
c11ff943 | 3418 | { |
f2645549 | 3419 | PERL_ARGS_ASSERT__IS_UTF8_XIDCONT; |
c11ff943 | 3420 | |
f25ce844 | 3421 | return is_utf8_common(p, &PL_utf8_idcont, "XIdContinue", NULL); |
c11ff943 KW |
3422 | } |
3423 | ||
3424 | bool | |
7dbf68d2 KW |
3425 | Perl__is_utf8_mark(pTHX_ const U8 *p) |
3426 | { | |
7dbf68d2 KW |
3427 | PERL_ARGS_ASSERT__IS_UTF8_MARK; |
3428 | ||
f25ce844 | 3429 | return is_utf8_common(p, &PL_utf8_mark, "IsM", NULL); |
7dbf68d2 KW |
3430 | } |
3431 | ||
b9992569 | 3432 | /* change namve uv1 to 'from' */ |
6a4a25f4 | 3433 | STATIC UV |
b9992569 KW |
3434 | S__to_utf8_case(pTHX_ const UV uv1, const U8 *p, U8* ustrp, STRLEN *lenp, |
3435 | SV **swashp, const char *normal, const char *special) | |
3436 | { | |
0134edef | 3437 | STRLEN len = 0; |
7918f24d | 3438 | |
b9992569 | 3439 | PERL_ARGS_ASSERT__TO_UTF8_CASE; |
7918f24d | 3440 | |
36eaa811 KW |
3441 | /* For code points that don't change case, we already know that the output |
3442 | * of this function is the unchanged input, so we can skip doing look-ups | |
3443 | * for them. Unfortunately the case-changing code points are scattered | |
3444 | * around. But there are some long consecutive ranges where there are no | |
3445 | * case changing code points. By adding tests, we can eliminate the lookup | |
3446 | * for all the ones in such ranges. This is currently done here only for | |
3447 | * just a few cases where the scripts are in common use in modern commerce | |
3448 | * (and scripts adjacent to those which can be included without additional | |
3449 | * tests). */ | |
3450 | ||
3451 | if (uv1 >= 0x0590) { | |
3452 | /* This keeps from needing further processing the code points most | |
3453 | * likely to be used in the following non-cased scripts: Hebrew, | |
3454 | * Arabic, Syriac, Thaana, NKo, Samaritan, Mandaic, Devanagari, | |
3455 | * Bengali, Gurmukhi, Gujarati, Oriya, Tamil, Telugu, Kannada, | |
3456 | * Malayalam, Sinhala, Thai, Lao, Tibetan, Myanmar */ | |
3457 | if (uv1 < 0x10A0) { | |
3458 | goto cases_to_self; | |
3459 | } | |
3460 | ||
3461 | /* The following largish code point ranges also don't have case | |
3462 | * changes, but khw didn't think they warranted extra tests to speed | |
3463 | * them up (which would slightly slow down everything else above them): | |
3464 | * 1100..139F Hangul Jamo, Ethiopic | |
3465 | * 1400..1CFF Unified Canadian Aboriginal Syllabics, Ogham, Runic, | |
3466 | * Tagalog, Hanunoo, Buhid, Tagbanwa, Khmer, Mongolian, | |
3467 | * Limbu, Tai Le, New Tai Lue, Buginese, Tai Tham, | |
3468 | * Combining Diacritical Marks Extended, Balinese, | |
3469 | * Sundanese, Batak, Lepcha, Ol Chiki | |
3470 | * 2000..206F General Punctuation | |
3471 | */ | |
3472 | ||
3473 | if (uv1 >= 0x2D30) { | |
3474 | ||
3475 | /* This keeps the from needing further processing the code points | |
3476 | * most likely to be used in the following non-cased major scripts: | |
3477 | * CJK, Katakana, Hiragana, plus some less-likely scripts. | |
3478 | * | |
3479 | * (0x2D30 above might have to be changed to 2F00 in the unlikely | |
3480 | * event that Unicode eventually allocates the unused block as of | |
3481 | * v8.0 2FE0..2FEF to code points that are cased. khw has verified | |
3482 | * that the test suite will start having failures to alert you | |
3483 | * should that happen) */ | |
3484 | if (uv1 < 0xA640) { | |
3485 | goto cases_to_self; | |
3486 | } | |
3487 | ||
3488 | if (uv1 >= 0xAC00) { | |
3489 | if (UNLIKELY(UNICODE_IS_SURROGATE(uv1))) { | |
5af9bc97 KW |
3490 | if (ckWARN_d(WARN_SURROGATE)) { |
3491 | const char* desc = (PL_op) ? OP_DESC(PL_op) : normal; | |
3492 | Perl_warner(aTHX_ packWARN(WARN_SURROGATE), | |
56576a04 KW |
3493 | "Operation \"%s\" returns its argument for" |
3494 | " UTF-16 surrogate U+%04" UVXf, desc, uv1); | |
5af9bc97 KW |
3495 | } |
3496 | goto cases_to_self; | |
3497 | } | |
36eaa811 KW |
3498 | |
3499 | /* AC00..FAFF Catches Hangul syllables and private use, plus | |
3500 | * some others */ | |
3501 | if (uv1 < 0xFB00) { | |
3502 | goto cases_to_self; | |
3503 | ||
3504 | } | |
3505 | ||
5af9bc97 | 3506 | if (UNLIKELY(UNICODE_IS_SUPER(uv1))) { |
d22ec717 KW |
3507 | if (UNLIKELY(uv1 > MAX_EXTERNALLY_LEGAL_CP)) { |
3508 | Perl_croak(aTHX_ cp_above_legal_max, uv1, | |
3509 | MAX_EXTERNALLY_LEGAL_CP); | |
5af9bc97 KW |
3510 | } |
3511 | if (ckWARN_d(WARN_NON_UNICODE)) { | |
3512 | const char* desc = (PL_op) ? OP_DESC(PL_op) : normal; | |
3513 | Perl_warner(aTHX_ packWARN(WARN_NON_UNICODE), | |
56576a04 KW |
3514 | "Operation \"%s\" returns its argument for" |
3515 | " non-Unicode code point 0x%04" UVXf, desc, uv1); | |
5af9bc97 KW |
3516 | } |
3517 | goto cases_to_self; | |
3518 | } | |
3bfc1e70 KW |
3519 | #ifdef HIGHEST_CASE_CHANGING_CP_FOR_USE_ONLY_BY_UTF8_DOT_C |
3520 | if (UNLIKELY(uv1 | |
3521 | > HIGHEST_CASE_CHANGING_CP_FOR_USE_ONLY_BY_UTF8_DOT_C)) | |
3522 | { | |
3523 | ||
56576a04 KW |
3524 | /* As of Unicode 10.0, this means we avoid swash creation |
3525 | * for anything beyond high Plane 1 (below emojis) */ | |
3bfc1e70 KW |
3526 | goto cases_to_self; |
3527 | } | |
3528 | #endif | |
36eaa811 KW |
3529 | } |
3530 | } | |
9ae3ac1a | 3531 | |
36eaa811 KW |
3532 | /* Note that non-characters are perfectly legal, so no warning should |
3533 | * be given. There are so few of them, that it isn't worth the extra | |
3534 | * tests to avoid swash creation */ | |
9ae3ac1a KW |
3535 | } |
3536 | ||
0134edef | 3537 | if (!*swashp) /* load on-demand */ |
56576a04 KW |
3538 | *swashp = _core_swash_init("utf8", normal, &PL_sv_undef, |
3539 | 4, 0, NULL, NULL); | |
0134edef | 3540 | |
a6f87d8c | 3541 | if (special) { |
0134edef | 3542 | /* It might be "special" (sometimes, but not always, |
2a37f04d | 3543 | * a multicharacter mapping) */ |
4a8240a3 | 3544 | HV *hv = NULL; |
b08cf34e JH |
3545 | SV **svp; |
3546 | ||
4a8240a3 KW |
3547 | /* If passed in the specials name, use that; otherwise use any |
3548 | * given in the swash */ | |
3549 | if (*special != '\0') { | |
3550 | hv = get_hv(special, 0); | |
3551 | } | |
3552 | else { | |
3553 | svp = hv_fetchs(MUTABLE_HV(SvRV(*swashp)), "SPECIALS", 0); | |
3554 | if (svp) { | |
3555 | hv = MUTABLE_HV(SvRV(*svp)); | |
3556 | } | |
3557 | } | |
3558 | ||
176fe009 | 3559 | if (hv |
5f560d8a | 3560 | && (svp = hv_fetch(hv, (const char*)p, UVCHR_SKIP(uv1), FALSE)) |
176fe009 KW |
3561 | && (*svp)) |
3562 | { | |
cfd0369c | 3563 | const char *s; |
47654450 | 3564 | |
cfd0369c | 3565 | s = SvPV_const(*svp, len); |
47654450 | 3566 | if (len == 1) |
f4cd282c | 3567 | /* EIGHTBIT */ |
c80e42f3 | 3568 | len = uvchr_to_utf8(ustrp, *(U8*)s) - ustrp; |
2a37f04d | 3569 | else { |
d2dcd0fb | 3570 | Copy(s, ustrp, len, U8); |
29e98929 | 3571 | } |
983ffd37 | 3572 | } |
0134edef JH |
3573 | } |
3574 | ||
3575 | if (!len && *swashp) { | |
4a4088c4 | 3576 | const UV uv2 = swash_fetch(*swashp, p, TRUE /* => is UTF-8 */); |
d4c19fe8 | 3577 | |
0134edef JH |
3578 | if (uv2) { |
3579 | /* It was "normal" (a single character mapping). */ | |
f4cd282c | 3580 | len = uvchr_to_utf8(ustrp, uv2) - ustrp; |
2a37f04d JH |
3581 | } |
3582 | } | |
1feea2c7 | 3583 | |
cbe07460 KW |
3584 | if (len) { |
3585 | if (lenp) { | |
3586 | *lenp = len; | |
3587 | } | |
3588 | return valid_utf8_to_uvchr(ustrp, 0); | |
3589 | } | |
3590 | ||
3591 | /* Here, there was no mapping defined, which means that the code point maps | |
3592 | * to itself. Return the inputs */ | |
e24dfe9c | 3593 | cases_to_self: |
bfdf22ec | 3594 | len = UTF8SKIP(p); |
ca9fab46 KW |
3595 | if (p != ustrp) { /* Don't copy onto itself */ |
3596 | Copy(p, ustrp, len, U8); | |
3597 | } | |
0134edef | 3598 | |
2a37f04d JH |
3599 | if (lenp) |
3600 | *lenp = len; | |
3601 | ||
f4cd282c | 3602 | return uv1; |
cbe07460 | 3603 | |
a0ed51b3 LW |
3604 | } |
3605 | ||
051a06d4 | 3606 | STATIC UV |
56576a04 KW |
3607 | S_check_locale_boundary_crossing(pTHX_ const U8* const p, const UV result, |
3608 | U8* const ustrp, STRLEN *lenp) | |
051a06d4 | 3609 | { |
4a4088c4 | 3610 | /* This is called when changing the case of a UTF-8-encoded character above |
31f05a37 KW |
3611 | * the Latin1 range, and the operation is in a non-UTF-8 locale. If the |
3612 | * result contains a character that crosses the 255/256 boundary, disallow | |
3613 | * the change, and return the original code point. See L<perlfunc/lc> for | |
3614 | * why; | |
051a06d4 | 3615 | * |
a1433954 KW |
3616 | * p points to the original string whose case was changed; assumed |
3617 | * by this routine to be well-formed | |
051a06d4 | 3618 | * result the code point of the first character in the changed-case string |
56576a04 KW |
3619 | * ustrp points to the changed-case string (<result> represents its |
3620 | * first char) | |
051a06d4 KW |
3621 | * lenp points to the length of <ustrp> */ |
3622 | ||
3623 | UV original; /* To store the first code point of <p> */ | |
3624 | ||
3625 | PERL_ARGS_ASSERT_CHECK_LOCALE_BOUNDARY_CROSSING; | |
3626 | ||
a4f12ed7 | 3627 | assert(UTF8_IS_ABOVE_LATIN1(*p)); |
051a06d4 KW |
3628 | |
3629 | /* We know immediately if the first character in the string crosses the | |
3630 | * boundary, so can skip */ | |
3631 | if (result > 255) { | |
3632 | ||
3633 | /* Look at every character in the result; if any cross the | |
3634 | * boundary, the whole thing is disallowed */ | |
3635 | U8* s = ustrp + UTF8SKIP(ustrp); | |
3636 | U8* e = ustrp + *lenp; | |
3637 | while (s < e) { | |
a4f12ed7 | 3638 | if (! UTF8_IS_ABOVE_LATIN1(*s)) { |
051a06d4 KW |
3639 | goto bad_crossing; |
3640 | } | |
3641 | s += UTF8SKIP(s); | |
3642 | } | |
3643 | ||
613abc6d KW |
3644 | /* Here, no characters crossed, result is ok as-is, but we warn. */ |
3645 | _CHECK_AND_OUTPUT_WIDE_LOCALE_UTF8_MSG(p, p + UTF8SKIP(p)); | |
051a06d4 KW |
3646 | return result; |
3647 | } | |
3648 | ||
7b52d656 | 3649 | bad_crossing: |
051a06d4 KW |
3650 | |
3651 | /* Failed, have to return the original */ | |
4b88fb76 | 3652 | original = valid_utf8_to_uvchr(p, lenp); |
ab0b796c KW |
3653 | |
3654 | /* diag_listed_as: Can't do %s("%s") on non-UTF-8 locale; resolved to "%s". */ | |
3655 | Perl_ck_warner(aTHX_ packWARN(WARN_LOCALE), | |
56576a04 KW |
3656 | "Can't do %s(\"\\x{%" UVXf "}\") on non-UTF-8" |
3657 | " locale; resolved to \"\\x{%" UVXf "}\".", | |
357aadde | 3658 | OP_DESC(PL_op), |
ab0b796c KW |
3659 | original, |
3660 | original); | |
051a06d4 KW |
3661 | Copy(p, ustrp, *lenp, char); |
3662 | return original; | |
3663 | } | |
3664 | ||
607313a1 KW |
3665 | STATIC U32 |
3666 | S_check_and_deprecate(pTHX_ const U8 *p, | |
3667 | const U8 **e, | |
3668 | const unsigned int type, /* See below */ | |
3669 | const bool use_locale, /* Is this a 'LC_' | |
3670 | macro call? */ | |
3671 | const char * const file, | |
3672 | const unsigned line) | |
3673 | { | |
3674 | /* This is a temporary function to deprecate the unsafe calls to the case | |
3675 | * changing macros and functions. It keeps all the special stuff in just | |
3676 | * one place. | |
3677 | * | |
3678 | * It updates *e with the pointer to the end of the input string. If using | |
3679 | * the old-style macros, *e is NULL on input, and so this function assumes | |
3680 | * the input string is long enough to hold the entire UTF-8 sequence, and | |
3681 | * sets *e accordingly, but it then returns a flag to pass the | |
3682 | * utf8n_to_uvchr(), to tell it that this size is a guess, and to avoid | |
3683 | * using the full length if possible. | |
3684 | * | |
3685 | * It also does the assert that *e > p when *e is not NULL. This should be | |
3686 | * migrated to the callers when this function gets deleted. | |
3687 | * | |
3688 | * The 'type' parameter is used for the caller to specify which case | |
3689 | * changing function this is called from: */ | |
3690 | ||
3691 | # define DEPRECATE_TO_UPPER 0 | |
3692 | # define DEPRECATE_TO_TITLE 1 | |
3693 | # define DEPRECATE_TO_LOWER 2 | |
3694 | # define DEPRECATE_TO_FOLD 3 | |
3695 | ||
3696 | U32 utf8n_flags = 0; | |
3697 | const char * name; | |
3698 | const char * alternative; | |
3699 | ||
3700 | PERL_ARGS_ASSERT_CHECK_AND_DEPRECATE; | |
3701 | ||
3702 | if (*e == NULL) { | |
3703 | utf8n_flags = _UTF8_NO_CONFIDENCE_IN_CURLEN; | |
3704 | *e = p + UTF8SKIP(p); | |
3705 | ||
3706 | /* For mathoms.c calls, we use the function name we know is stored | |
c44e9413 | 3707 | * there. It could be part of a larger path */ |
607313a1 | 3708 | if (type == DEPRECATE_TO_UPPER) { |
c44e9413 | 3709 | name = instr(file, "mathoms.c") |
607313a1 KW |
3710 | ? "to_utf8_upper" |
3711 | : "toUPPER_utf8"; | |
3712 | alternative = "toUPPER_utf8_safe"; | |
3713 | } | |
3714 | else if (type == DEPRECATE_TO_TITLE) { | |
c44e9413 | 3715 | name = instr(file, "mathoms.c") |
607313a1 KW |
3716 | ? "to_utf8_title" |
3717 | : "toTITLE_utf8"; | |
3718 | alternative = "toTITLE_utf8_safe"; | |
3719 | } | |
3720 | else if (type == DEPRECATE_TO_LOWER) { | |
c44e9413 | 3721 | name = instr(file, "mathoms.c") |
607313a1 KW |
3722 | ? "to_utf8_lower" |
3723 | : "toLOWER_utf8"; | |
3724 | alternative = "toLOWER_utf8_safe"; | |
3725 | } | |
3726 | else if (type == DEPRECATE_TO_FOLD) { | |
c44e9413 | 3727 | name = instr(file, "mathoms.c") |
607313a1 KW |
3728 | ? "to_utf8_fold" |
3729 | : "toFOLD_utf8"; | |
3730 | alternative = "toFOLD_utf8_safe"; | |
3731 | } | |
3732 | else Perl_croak(aTHX_ "panic: Unexpected case change type"); | |
3733 | ||
3734 | warn_on_first_deprecated_use(name, alternative, use_locale, file, line); | |
3735 | } | |
3736 | else { | |
3737 | assert (p < *e); | |
3738 | } | |
3739 | ||
3740 | return utf8n_flags; | |
3741 | } | |
3742 | ||
eaf412bf KW |
3743 | /* The process for changing the case is essentially the same for the four case |
3744 | * change types, except there are complications for folding. Otherwise the | |
3745 | * difference is only which case to change to. To make sure that they all do | |
3746 | * the same thing, the bodies of the functions are extracted out into the | |
3747 | * following two macros. The functions are written with the same variable | |
3748 | * names, and these are known and used inside these macros. It would be | |
3749 | * better, of course, to have inline functions to do it, but since different | |
3750 | * macros are called, depending on which case is being changed to, this is not | |
3751 | * feasible in C (to khw's knowledge). Two macros are created so that the fold | |
3752 | * function can start with the common start macro, then finish with its special | |
3753 | * handling; while the other three cases can just use the common end macro. | |
3754 | * | |
3755 | * The algorithm is to use the proper (passed in) macro or function to change | |
3756 | * the case for code points that are below 256. The macro is used if using | |
3757 | * locale rules for the case change; the function if not. If the code point is | |
3758 | * above 255, it is computed from the input UTF-8, and another macro is called | |
3759 | * to do the conversion. If necessary, the output is converted to UTF-8. If | |
3760 | * using a locale, we have to check that the change did not cross the 255/256 | |
3761 | * boundary, see check_locale_boundary_crossing() for further details. | |
3762 | * | |
3763 | * The macros are split with the correct case change for the below-256 case | |
3764 | * stored into 'result', and in the middle of an else clause for the above-255 | |
3765 | * case. At that point in the 'else', 'result' is not the final result, but is | |
3766 | * the input code point calculated from the UTF-8. The fold code needs to | |
3767 | * realize all this and take it from there. | |
3768 | * | |
3769 | * If you read the two macros as sequential, it's easier to understand what's | |
3770 | * going on. */ | |
3771 | #define CASE_CHANGE_BODY_START(locale_flags, LC_L1_change_macro, L1_func, \ | |
3772 | L1_func_extra_param) \ | |
a239b1e2 | 3773 | \ |
eaf412bf | 3774 | if (flags & (locale_flags)) { \ |
8b7358b9 | 3775 | _CHECK_AND_WARN_PROBLEMATIC_LOCALE; \ |
eaf412bf KW |
3776 | /* Treat a UTF-8 locale as not being in locale at all */ \ |
3777 | if (IN_UTF8_CTYPE_LOCALE) { \ | |
3778 | flags &= ~(locale_flags); \ | |
3779 | } \ | |
eaf412bf KW |
3780 | } \ |
3781 | \ | |
3782 | if (UTF8_IS_INVARIANT(*p)) { \ | |
3783 | if (flags & (locale_flags)) { \ | |
3784 | result = LC_L1_change_macro(*p); \ | |
3785 | } \ | |
3786 | else { \ | |
3787 | return L1_func(*p, ustrp, lenp, L1_func_extra_param); \ | |
3788 | } \ | |
3789 | } \ | |
a239b1e2 | 3790 | else if UTF8_IS_NEXT_CHAR_DOWNGRADEABLE(p, e) { \ |
eaf412bf KW |
3791 | if (flags & (locale_flags)) { \ |
3792 | result = LC_L1_change_macro(EIGHT_BIT_UTF8_TO_NATIVE(*p, \ | |
3793 | *(p+1))); \ | |
3794 | } \ | |
3795 | else { \ | |
3796 | return L1_func(EIGHT_BIT_UTF8_TO_NATIVE(*p, *(p+1)), \ | |
3797 | ustrp, lenp, L1_func_extra_param); \ | |
3798 | } \ | |
3799 | } \ | |
fa8ab374 KW |
3800 | else { /* malformed UTF-8 or ord above 255 */ \ |
3801 | STRLEN len_result; \ | |
fa8ab374 KW |
3802 | result = utf8n_to_uvchr(p, e - p, &len_result, UTF8_CHECK_ONLY); \ |
3803 | if (len_result == (STRLEN) -1) { \ | |
607313a1 KW |
3804 | _force_out_malformed_utf8_message(p, e, utf8n_flags, \ |
3805 | 1 /* Die */ ); \ | |
fa8ab374 | 3806 | } |
eaf412bf KW |
3807 | |
3808 | #define CASE_CHANGE_BODY_END(locale_flags, change_macro) \ | |
3809 | result = change_macro(result, p, ustrp, lenp); \ | |
3810 | \ | |
3811 | if (flags & (locale_flags)) { \ | |
3812 | result = check_locale_boundary_crossing(p, result, ustrp, lenp); \ | |
3813 | } \ | |
3814 | return result; \ | |
3815 | } \ | |
3816 | \ | |
3817 | /* Here, used locale rules. Convert back to UTF-8 */ \ | |
3818 | if (UTF8_IS_INVARIANT(result)) { \ | |
3819 | *ustrp = (U8) result; \ | |
3820 | *lenp = 1; \ | |
3821 | } \ | |
3822 | else { \ | |
3823 | *ustrp = UTF8_EIGHT_BIT_HI((U8) result); \ | |
3824 | *(ustrp + 1) = UTF8_EIGHT_BIT_LO((U8) result); \ | |
3825 | *lenp = 2; \ | |
3826 | } \ | |
3827 | \ | |
3828 | return result; | |
3829 | ||
d3e79532 | 3830 | /* |
87cea99e | 3831 | =for apidoc to_utf8_upper |
d3e79532 | 3832 | |
a239b1e2 | 3833 | Instead use L</toUPPER_utf8_safe>. |
a1433954 | 3834 | |
d3e79532 JH |
3835 | =cut */ |
3836 | ||
051a06d4 | 3837 | /* Not currently externally documented, and subject to change: |
31f05a37 KW |
3838 | * <flags> is set iff iff the rules from the current underlying locale are to |
3839 | * be used. */ | |
051a06d4 | 3840 | |
2104c8d9 | 3841 | UV |
607313a1 KW |
3842 | Perl__to_utf8_upper_flags(pTHX_ const U8 *p, |
3843 | const U8 *e, | |
3844 | U8* ustrp, | |
3845 | STRLEN *lenp, | |
3846 | bool flags, | |
3847 | const char * const file, | |
3848 | const int line) | |
a0ed51b3 | 3849 | { |
051a06d4 | 3850 | UV result; |
607313a1 KW |
3851 | const U32 utf8n_flags = check_and_deprecate(p, &e, DEPRECATE_TO_UPPER, |
3852 | cBOOL(flags), file, line); | |
051a06d4 KW |
3853 | |
3854 | PERL_ARGS_ASSERT__TO_UTF8_UPPER_FLAGS; | |
7918f24d | 3855 | |
eaf412bf KW |
3856 | /* ~0 makes anything non-zero in 'flags' mean we are using locale rules */ |
3857 | /* 2nd char of uc(U+DF) is 'S' */ | |
3858 | CASE_CHANGE_BODY_START(~0, toUPPER_LC, _to_upper_title_latin1, 'S'); | |
3859 | CASE_CHANGE_BODY_END (~0, CALL_UPPER_CASE); | |
983ffd37 | 3860 | } |
a0ed51b3 | 3861 | |
d3e79532 | 3862 | /* |
87cea99e | 3863 | =for apidoc to_utf8_title |
d3e79532 | 3864 | |
a239b1e2 | 3865 | Instead use L</toTITLE_utf8_safe>. |
a1433954 | 3866 | |
d3e79532 JH |
3867 | =cut */ |
3868 | ||
051a06d4 | 3869 | /* Not currently externally documented, and subject to change: |
31f05a37 KW |
3870 | * <flags> is set iff the rules from the current underlying locale are to be |
3871 | * used. Since titlecase is not defined in POSIX, for other than a | |
3872 | * UTF-8 locale, uppercase is used instead for code points < 256. | |
445bf929 | 3873 | */ |
051a06d4 | 3874 | |
983ffd37 | 3875 | UV |
607313a1 KW |
3876 | Perl__to_utf8_title_flags(pTHX_ const U8 *p, |
3877 | const U8 *e, | |
3878 | U8* ustrp, | |
3879 | STRLEN *lenp, | |
3880 | bool flags, | |
3881 | const char * const file, | |
3882 | const int line) | |
983ffd37 | 3883 | { |
051a06d4 | 3884 | UV result; |
607313a1 KW |
3885 | const U32 utf8n_flags = check_and_deprecate(p, &e, DEPRECATE_TO_TITLE, |
3886 | cBOOL(flags), file, line); | |
051a06d4 KW |
3887 | |
3888 | PERL_ARGS_ASSERT__TO_UTF8_TITLE_FLAGS; | |
7918f24d | 3889 | |
eaf412bf KW |
3890 | /* 2nd char of ucfirst(U+DF) is 's' */ |
3891 | CASE_CHANGE_BODY_START(~0, toUPPER_LC, _to_upper_title_latin1, 's'); | |
3892 | CASE_CHANGE_BODY_END (~0, CALL_TITLE_CASE); | |
a0ed51b3 LW |
3893 | } |
3894 | ||
d3e79532 | 3895 | /* |
87cea99e | 3896 | =for apidoc to_utf8_lower |
d3e79532 | 3897 | |
a239b1e2 | 3898 | Instead use L</toLOWER_utf8_safe>. |
a1433954 | 3899 | |
d3e79532 JH |
3900 | =cut */ |
3901 | ||
051a06d4 | 3902 | /* Not currently externally documented, and subject to change: |
31f05a37 KW |
3903 | * <flags> is set iff iff the rules from the current underlying locale are to |
3904 | * be used. | |
3905 | */ | |
051a06d4 | 3906 | |
2104c8d9 | 3907 | UV |
607313a1 KW |
3908 | Perl__to_utf8_lower_flags(pTHX_ const U8 *p, |
3909 | const U8 *e, | |
3910 | U8* ustrp, | |
3911 | STRLEN *lenp, | |
3912 | bool flags, | |
3913 | const char * const file, | |
3914 | const int line) | |
a0ed51b3 | 3915 | { |
051a06d4 | 3916 | UV result; |
607313a1 KW |
3917 | const U32 utf8n_flags = check_and_deprecate(p, &e, DEPRECATE_TO_LOWER, |
3918 | cBOOL(flags), file, line); | |
051a06d4 | 3919 | |
051a06d4 | 3920 | PERL_ARGS_ASSERT__TO_UTF8_LOWER_FLAGS; |
7918f24d | 3921 | |
eaf412bf KW |
3922 | CASE_CHANGE_BODY_START(~0, toLOWER_LC, to_lower_latin1, 0 /* 0 is dummy */) |
3923 | CASE_CHANGE_BODY_END (~0, CALL_LOWER_CASE) | |
b4e400f9 JH |
3924 | } |
3925 | ||
d3e79532 | 3926 | /* |
87cea99e | 3927 | =for apidoc to_utf8_fold |
d3e79532 | 3928 | |
a239b1e2 | 3929 | Instead use L</toFOLD_utf8_safe>. |
a1433954 | 3930 | |
d3e79532 JH |
3931 | =cut */ |
3932 | ||
051a06d4 KW |
3933 | /* Not currently externally documented, and subject to change, |
3934 | * in <flags> | |
31f05a37 KW |
3935 | * bit FOLD_FLAGS_LOCALE is set iff the rules from the current underlying |
3936 | * locale are to be used. | |
051a06d4 KW |
3937 | * bit FOLD_FLAGS_FULL is set iff full case folds are to be used; |
3938 | * otherwise simple folds | |
a0270393 KW |
3939 | * bit FOLD_FLAGS_NOMIX_ASCII is set iff folds of non-ASCII to ASCII are |
3940 | * prohibited | |
445bf929 | 3941 | */ |
36bb2ab6 | 3942 | |
b4e400f9 | 3943 | UV |
607313a1 KW |
3944 | Perl__to_utf8_fold_flags(pTHX_ const U8 *p, |
3945 | const U8 *e, | |
3946 | U8* ustrp, | |
3947 | STRLEN *lenp, | |
3948 | U8 flags, | |
3949 | const char * const file, | |
3950 | const int line) | |
b4e400f9 | 3951 | { |
051a06d4 | 3952 | UV result; |
607313a1 KW |
3953 | const U32 utf8n_flags = check_and_deprecate(p, &e, DEPRECATE_TO_FOLD, |
3954 | cBOOL(flags), file, line); | |
051a06d4 | 3955 | |
36bb2ab6 | 3956 | PERL_ARGS_ASSERT__TO_UTF8_FOLD_FLAGS; |
7918f24d | 3957 | |
a0270393 KW |
3958 | /* These are mutually exclusive */ |
3959 | assert (! ((flags & FOLD_FLAGS_LOCALE) && (flags & FOLD_FLAGS_NOMIX_ASCII))); | |
3960 | ||
50ba90ff KW |
3961 | assert(p != ustrp); /* Otherwise overwrites */ |
3962 | ||
eaf412bf KW |
3963 | CASE_CHANGE_BODY_START(FOLD_FLAGS_LOCALE, toFOLD_LC, _to_fold_latin1, |
3964 | ((flags) & (FOLD_FLAGS_FULL | FOLD_FLAGS_NOMIX_ASCII))); | |
31f05a37 | 3965 | |
eaf412bf | 3966 | result = CALL_FOLD_CASE(result, p, ustrp, lenp, flags & FOLD_FLAGS_FULL); |
a1dde8de | 3967 | |
1ca267a5 KW |
3968 | if (flags & FOLD_FLAGS_LOCALE) { |
3969 | ||
76f2ffcd | 3970 | # define LONG_S_T LATIN_SMALL_LIGATURE_LONG_S_T_UTF8 |
0766489e KW |
3971 | # ifdef LATIN_CAPITAL_LETTER_SHARP_S_UTF8 |
3972 | # define CAP_SHARP_S LATIN_CAPITAL_LETTER_SHARP_S_UTF8 | |
76f2ffcd | 3973 | |
538e84ed KW |
3974 | /* Special case these two characters, as what normally gets |
3975 | * returned under locale doesn't work */ | |
b59bf0b2 | 3976 | if (memEQs((char *) p, UTF8SKIP(p), CAP_SHARP_S)) |
1ca267a5 | 3977 | { |
ab0b796c KW |
3978 | /* diag_listed_as: Can't do %s("%s") on non-UTF-8 locale; resolved to "%s". */ |
3979 | Perl_ck_warner(aTHX_ packWARN(WARN_LOCALE), | |
3980 | "Can't do fc(\"\\x{1E9E}\") on non-UTF-8 locale; " | |
3981 | "resolved to \"\\x{17F}\\x{17F}\"."); | |
1ca267a5 KW |
3982 | goto return_long_s; |
3983 | } | |
0766489e KW |
3984 | else |
3985 | #endif | |
b59bf0b2 | 3986 | if (memEQs((char *) p, UTF8SKIP(p), LONG_S_T)) |
9fc2026f | 3987 | { |
ab0b796c KW |
3988 | /* diag_listed_as: Can't do %s("%s") on non-UTF-8 locale; resolved to "%s". */ |
3989 | Perl_ck_warner(aTHX_ packWARN(WARN_LOCALE), | |
3990 | "Can't do fc(\"\\x{FB05}\") on non-UTF-8 locale; " | |
3991 | "resolved to \"\\x{FB06}\"."); | |
9fc2026f KW |
3992 | goto return_ligature_st; |
3993 | } | |
74894415 KW |
3994 | |
3995 | #if UNICODE_MAJOR_VERSION == 3 \ | |
3996 | && UNICODE_DOT_VERSION == 0 \ | |
3997 | && UNICODE_DOT_DOT_VERSION == 1 | |
3998 | # define DOTTED_I LATIN_CAPITAL_LETTER_I_WITH_DOT_ABOVE_UTF8 | |
3999 | ||
4000 | /* And special case this on this Unicode version only, for the same | |
4001 | * reaons the other two are special cased. They would cross the | |
4002 | * 255/256 boundary which is forbidden under /l, and so the code | |
4003 | * wouldn't catch that they are equivalent (which they are only in | |
4004 | * this release) */ | |
b59bf0b2 | 4005 | else if (memEQs((char *) p, UTF8SKIP(p), DOTTED_I)) { |
74894415 KW |
4006 | /* diag_listed_as: Can't do %s("%s") on non-UTF-8 locale; resolved to "%s". */ |
4007 | Perl_ck_warner(aTHX_ packWARN(WARN_LOCALE), | |
4008 | "Can't do fc(\"\\x{0130}\") on non-UTF-8 locale; " | |
4009 | "resolved to \"\\x{0131}\"."); | |
4010 | goto return_dotless_i; | |
4011 | } | |
4012 | #endif | |
4013 | ||
357aadde | 4014 | return check_locale_boundary_crossing(p, result, ustrp, lenp); |
051a06d4 | 4015 | } |
a0270393 KW |
4016 | else if (! (flags & FOLD_FLAGS_NOMIX_ASCII)) { |
4017 | return result; | |
4018 | } | |
4019 | else { | |
4a4088c4 | 4020 | /* This is called when changing the case of a UTF-8-encoded |
9fc2026f KW |
4021 | * character above the ASCII range, and the result should not |
4022 | * contain an ASCII character. */ | |
a0270393 KW |
4023 | |
4024 | UV original; /* To store the first code point of <p> */ | |
4025 | ||
4026 | /* Look at every character in the result; if any cross the | |
4027 | * boundary, the whole thing is disallowed */ | |
4028 | U8* s = ustrp; | |
4029 | U8* e = ustrp + *lenp; | |
4030 | while (s < e) { | |
4031 | if (isASCII(*s)) { | |
4032 | /* Crossed, have to return the original */ | |
4033 | original = valid_utf8_to_uvchr(p, lenp); | |
1ca267a5 | 4034 | |
9fc2026f | 4035 | /* But in these instances, there is an alternative we can |
1ca267a5 | 4036 | * return that is valid */ |
0766489e KW |
4037 | if (original == LATIN_SMALL_LETTER_SHARP_S |
4038 | #ifdef LATIN_CAPITAL_LETTER_SHARP_S /* not defined in early Unicode releases */ | |
4039 | || original == LATIN_CAPITAL_LETTER_SHARP_S | |
4040 | #endif | |
4041 | ) { | |
1ca267a5 KW |
4042 | goto return_long_s; |
4043 | } | |
9fc2026f KW |
4044 | else if (original == LATIN_SMALL_LIGATURE_LONG_S_T) { |
4045 | goto return_ligature_st; | |
4046 | } | |
74894415 KW |
4047 | #if UNICODE_MAJOR_VERSION == 3 \ |
4048 | && UNICODE_DOT_VERSION == 0 \ | |
4049 | && UNICODE_DOT_DOT_VERSION == 1 | |
4050 | ||
4051 | else if (original == LATIN_CAPITAL_LETTER_I_WITH_DOT_ABOVE) { | |
4052 | goto return_dotless_i; | |
4053 | } | |
4054 | #endif | |
a0270393 KW |
4055 | Copy(p, ustrp, *lenp, char); |
4056 | return original; | |
4057 | } | |
4058 | s += UTF8SKIP(s); | |
4059 | } | |
051a06d4 | 4060 | |
a0270393 KW |
4061 | /* Here, no characters crossed, result is ok as-is */ |
4062 | return result; | |
4063 | } | |
051a06d4 KW |
4064 | } |
4065 | ||
4a4088c4 | 4066 | /* Here, used locale rules. Convert back to UTF-8 */ |
051a06d4 KW |
4067 | if (UTF8_IS_INVARIANT(result)) { |
4068 | *ustrp = (U8) result; | |
4069 | *lenp = 1; | |
4070 | } | |
4071 | else { | |
62cb07ea KW |
4072 | *ustrp = UTF8_EIGHT_BIT_HI((U8) result); |
4073 | *(ustrp + 1) = UTF8_EIGHT_BIT_LO((U8) result); | |
051a06d4 KW |
4074 | *lenp = 2; |
4075 | } | |
4076 | ||
051a06d4 | 4077 | return result; |
1ca267a5 KW |
4078 | |
4079 | return_long_s: | |
4080 | /* Certain folds to 'ss' are prohibited by the options, but they do allow | |
4081 | * folds to a string of two of these characters. By returning this | |
4082 | * instead, then, e.g., | |
4083 | * fc("\x{1E9E}") eq fc("\x{17F}\x{17F}") | |
4084 | * works. */ | |
4085 | ||
4086 | *lenp = 2 * sizeof(LATIN_SMALL_LETTER_LONG_S_UTF8) - 2; | |
4087 | Copy(LATIN_SMALL_LETTER_LONG_S_UTF8 LATIN_SMALL_LETTER_LONG_S_UTF8, | |
4088 | ustrp, *lenp, U8); | |
4089 | return LATIN_SMALL_LETTER_LONG_S; | |
9fc2026f KW |
4090 | |
4091 | return_ligature_st: | |
4092 | /* Two folds to 'st' are prohibited by the options; instead we pick one and | |
4093 | * have the other one fold to it */ | |
4094 | ||
4095 | *lenp = sizeof(LATIN_SMALL_LIGATURE_ST_UTF8) - 1; | |
4096 | Copy(LATIN_SMALL_LIGATURE_ST_UTF8, ustrp, *lenp, U8); | |
4097 | return LATIN_SMALL_LIGATURE_ST; | |
74894415 KW |
4098 | |
4099 | #if UNICODE_MAJOR_VERSION == 3 \ | |
4100 | && UNICODE_DOT_VERSION == 0 \ | |
4101 | && UNICODE_DOT_DOT_VERSION == 1 | |
4102 | ||
4103 | return_dotless_i: | |
4104 | *lenp = sizeof(LATIN_SMALL_LETTER_DOTLESS_I_UTF8) - 1; | |
4105 | Copy(LATIN_SMALL_LETTER_DOTLESS_I_UTF8, ustrp, *lenp, U8); | |
4106 | return LATIN_SMALL_LETTER_DOTLESS_I; | |
4107 | ||
4108 | #endif | |
4109 | ||
a0ed51b3 LW |
4110 | } |
4111 | ||
711a919c | 4112 | /* Note: |
f90a9a02 | 4113 | * Returns a "swash" which is a hash described in utf8.c:Perl_swash_fetch(). |
711a919c TS |
4114 | * C<pkg> is a pointer to a package name for SWASHNEW, should be "utf8". |
4115 | * For other parameters, see utf8::SWASHNEW in lib/utf8_heavy.pl. | |
4116 | */ | |
c4a5db0c | 4117 | |
a0ed51b3 | 4118 | SV* |
56576a04 KW |
4119 | Perl_swash_init(pTHX_ const char* pkg, const char* name, SV *listsv, |
4120 | I32 minbits, I32 none) | |
a0ed51b3 | 4121 | { |
c4a5db0c KW |
4122 | PERL_ARGS_ASSERT_SWASH_INIT; |
4123 | ||
4124 | /* Returns a copy of a swash initiated by the called function. This is the | |
4125 | * public interface, and returning a copy prevents others from doing | |
4126 | * mischief on the original */ | |
4127 | ||
56576a04 KW |
4128 | return newSVsv(_core_swash_init(pkg, name, listsv, minbits, none, |
4129 | NULL, NULL)); | |
c4a5db0c KW |
4130 | } |
4131 | ||
4132 | SV* | |
56576a04 KW |
4133 | Perl__core_swash_init(pTHX_ const char* pkg, const char* name, SV *listsv, |
4134 | I32 minbits, I32 none, SV* invlist, | |
4135 | U8* const flags_p) | |
c4a5db0c | 4136 | { |
2c1f00b9 YO |
4137 | |
4138 | /*NOTE NOTE NOTE - If you want to use "return" in this routine you MUST | |
4139 | * use the following define */ | |
4140 | ||
4141 | #define CORE_SWASH_INIT_RETURN(x) \ | |
4142 | PL_curpm= old_PL_curpm; \ | |
4143 | return x | |
4144 | ||
c4a5db0c | 4145 | /* Initialize and return a swash, creating it if necessary. It does this |
87367d5f KW |
4146 | * by calling utf8_heavy.pl in the general case. The returned value may be |
4147 | * the swash's inversion list instead if the input parameters allow it. | |
4148 | * Which is returned should be immaterial to callers, as the only | |
923b6d4e KW |
4149 | * operations permitted on a swash, swash_fetch(), _get_swash_invlist(), |
4150 | * and swash_to_invlist() handle both these transparently. | |
c4a5db0c KW |
4151 | * |
4152 | * This interface should only be used by functions that won't destroy or | |
4153 | * adversely change the swash, as doing so affects all other uses of the | |
4154 | * swash in the program; the general public should use 'Perl_swash_init' | |
4155 | * instead. | |
4156 | * | |
4157 | * pkg is the name of the package that <name> should be in. | |
4158 | * name is the name of the swash to find. Typically it is a Unicode | |
4159 | * property name, including user-defined ones | |
4160 | * listsv is a string to initialize the swash with. It must be of the form | |
4161 | * documented as the subroutine return value in | |
4162 | * L<perlunicode/User-Defined Character Properties> | |
4163 | * minbits is the number of bits required to represent each data element. | |
4164 | * It is '1' for binary properties. | |
4165 | * none I (khw) do not understand this one, but it is used only in tr///. | |
9a53f6cf | 4166 | * invlist is an inversion list to initialize the swash with (or NULL) |
83199d38 KW |
4167 | * flags_p if non-NULL is the address of various input and output flag bits |
4168 | * to the routine, as follows: ('I' means is input to the routine; | |
4169 | * 'O' means output from the routine. Only flags marked O are | |
4170 | * meaningful on return.) | |
4171 | * _CORE_SWASH_INIT_USER_DEFINED_PROPERTY indicates if the swash | |
4172 | * came from a user-defined property. (I O) | |
5d3d13d1 KW |
4173 | * _CORE_SWASH_INIT_RETURN_IF_UNDEF indicates that instead of croaking |
4174 | * when the swash cannot be located, to simply return NULL. (I) | |
87367d5f KW |
4175 | * _CORE_SWASH_INIT_ACCEPT_INVLIST indicates that the caller will accept a |
4176 | * return of an inversion list instead of a swash hash if this routine | |
4177 | * thinks that would result in faster execution of swash_fetch() later | |
4178 | * on. (I) | |
9a53f6cf KW |
4179 | * |
4180 | * Thus there are three possible inputs to find the swash: <name>, | |
4181 | * <listsv>, and <invlist>. At least one must be specified. The result | |
4182 | * will be the union of the specified ones, although <listsv>'s various | |
aabbdbda KW |
4183 | * actions can intersect, etc. what <name> gives. To avoid going out to |
4184 | * disk at all, <invlist> should specify completely what the swash should | |
4185 | * have, and <listsv> should be &PL_sv_undef and <name> should be "". | |
9a53f6cf KW |
4186 | * |
4187 | * <invlist> is only valid for binary properties */ | |
c4a5db0c | 4188 | |
2c1f00b9 YO |
4189 | PMOP *old_PL_curpm= PL_curpm; /* save away the old PL_curpm */ |
4190 | ||
c4a5db0c | 4191 | SV* retval = &PL_sv_undef; |
83199d38 | 4192 | HV* swash_hv = NULL; |
87367d5f KW |
4193 | const int invlist_swash_boundary = |
4194 | (flags_p && *flags_p & _CORE_SWASH_INIT_ACCEPT_INVLIST) | |
4195 | ? 512 /* Based on some benchmarking, but not extensive, see commit | |
4196 | message */ | |
4197 | : -1; /* Never return just an inversion list */ | |
9a53f6cf KW |
4198 | |
4199 | assert(listsv != &PL_sv_undef || strNE(name, "") || invlist); | |
4200 | assert(! invlist || minbits == 1); | |
4201 | ||
56576a04 KW |
4202 | PL_curpm= NULL; /* reset PL_curpm so that we dont get confused between the |
4203 | regex that triggered the swash init and the swash init | |
4204 | perl logic itself. See perl #122747 */ | |
2c1f00b9 | 4205 | |
9a53f6cf KW |
4206 | /* If data was passed in to go out to utf8_heavy to find the swash of, do |
4207 | * so */ | |
4208 | if (listsv != &PL_sv_undef || strNE(name, "")) { | |
69794297 KW |
4209 | dSP; |
4210 | const size_t pkg_len = strlen(pkg); | |
4211 | const size_t name_len = strlen(name); | |
4212 | HV * const stash = gv_stashpvn(pkg, pkg_len, 0); | |
4213 | SV* errsv_save; | |
4214 | GV *method; | |
4215 | ||
4216 | PERL_ARGS_ASSERT__CORE_SWASH_INIT; | |
4217 | ||
4218 | PUSHSTACKi(PERLSI_MAGIC); | |
ce3b816e | 4219 | ENTER; |
69794297 | 4220 | SAVEHINTS(); |
2782061f | 4221 | save_re_context(); |
650f067c JL |
4222 | /* We might get here via a subroutine signature which uses a utf8 |
4223 | * parameter name, at which point PL_subname will have been set | |
4224 | * but not yet used. */ | |
4225 | save_item(PL_subname); | |
69794297 KW |
4226 | if (PL_parser && PL_parser->error_count) |
4227 | SAVEI8(PL_parser->error_count), PL_parser->error_count = 0; | |
4228 | method = gv_fetchmeth(stash, "SWASHNEW", 8, -1); | |
4a4088c4 | 4229 | if (!method) { /* demand load UTF-8 */ |
69794297 | 4230 | ENTER; |
db2c6cb3 FC |
4231 | if ((errsv_save = GvSV(PL_errgv))) SAVEFREESV(errsv_save); |
4232 | GvSV(PL_errgv) = NULL; | |
1a419e6b | 4233 | #ifndef NO_TAINT_SUPPORT |
69794297 KW |
4234 | /* It is assumed that callers of this routine are not passing in |
4235 | * any user derived data. */ | |
2782061f DM |
4236 | /* Need to do this after save_re_context() as it will set |
4237 | * PL_tainted to 1 while saving $1 etc (see the code after getrx: | |
4238 | * in Perl_magic_get). Even line to create errsv_save can turn on | |
4239 | * PL_tainted. */ | |
284167a5 S |
4240 | SAVEBOOL(TAINT_get); |
4241 | TAINT_NOT; | |
4242 | #endif | |
69794297 KW |
4243 | Perl_load_module(aTHX_ PERL_LOADMOD_NOIMPORT, newSVpvn(pkg,pkg_len), |
4244 | NULL); | |
eed484f9 | 4245 | { |
db2c6cb3 FC |
4246 | /* Not ERRSV, as there is no need to vivify a scalar we are |
4247 | about to discard. */ | |
4248 | SV * const errsv = GvSV(PL_errgv); | |
4249 | if (!SvTRUE(errsv)) { | |
4250 | GvSV(PL_errgv) = SvREFCNT_inc_simple(errsv_save); | |
4251 | SvREFCNT_dec(errsv); | |
4252 | } | |
eed484f9 | 4253 | } |
69794297 KW |
4254 | LEAVE; |
4255 | } | |
4256 | SPAGAIN; | |
4257 | PUSHMARK(SP); | |
4258 | EXTEND(SP,5); | |
4259 | mPUSHp(pkg, pkg_len); | |
4260 | mPUSHp(name, name_len); | |
4261 | PUSHs(listsv); | |
4262 | mPUSHi(minbits); | |
4263 | mPUSHi(none); | |
4264 | PUTBACK; | |
db2c6cb3 FC |
4265 | if ((errsv_save = GvSV(PL_errgv))) SAVEFREESV(errsv_save); |
4266 | GvSV(PL_errgv) = NULL; | |
69794297 KW |
4267 | /* If we already have a pointer to the method, no need to use |
4268 | * call_method() to repeat the lookup. */ | |
c41800a8 KW |
4269 | if (method |
4270 | ? call_sv(MUTABLE_SV(method), G_SCALAR) | |
69794297 KW |
4271 | : call_sv(newSVpvs_flags("SWASHNEW", SVs_TEMP), G_SCALAR | G_METHOD)) |
4272 | { | |
4273 | retval = *PL_stack_sp--; | |
4274 | SvREFCNT_inc(retval); | |
4275 | } | |
eed484f9 | 4276 | { |
db2c6cb3 FC |
4277 | /* Not ERRSV. See above. */ |
4278 | SV * const errsv = GvSV(PL_errgv); | |
4279 | if (!SvTRUE(errsv)) { | |
4280 | GvSV(PL_errgv) = SvREFCNT_inc_simple(errsv_save); | |
4281 | SvREFCNT_dec(errsv); | |
4282 | } | |
eed484f9 | 4283 | } |
ce3b816e | 4284 | LEAVE; |
69794297 KW |
4285 | POPSTACK; |
4286 | if (IN_PERL_COMPILETIME) { | |
4287 | CopHINTS_set(PL_curcop, PL_hints); | |
4288 | } | |
4289 | if (!SvROK(retval) || SvTYPE(SvRV(retval)) != SVt_PVHV) { | |
d95e4a00 | 4290 | if (SvPOK(retval)) { |
69794297 KW |
4291 | |
4292 | /* If caller wants to handle missing properties, let them */ | |
5d3d13d1 | 4293 | if (flags_p && *flags_p & _CORE_SWASH_INIT_RETURN_IF_UNDEF) { |
2c1f00b9 | 4294 | CORE_SWASH_INIT_RETURN(NULL); |
69794297 KW |
4295 | } |
4296 | Perl_croak(aTHX_ | |
147e3846 | 4297 | "Can't find Unicode property definition \"%" SVf "\"", |
69794297 | 4298 | SVfARG(retval)); |
a25b5927 | 4299 | NOT_REACHED; /* NOTREACHED */ |
d95e4a00 | 4300 | } |
69794297 | 4301 | } |
9a53f6cf | 4302 | } /* End of calling the module to find the swash */ |
36eb48b4 | 4303 | |
83199d38 KW |
4304 | /* If this operation fetched a swash, and we will need it later, get it */ |
4305 | if (retval != &PL_sv_undef | |
4306 | && (minbits == 1 || (flags_p | |
4307 | && ! (*flags_p | |
4308 | & _CORE_SWASH_INIT_USER_DEFINED_PROPERTY)))) | |
4309 | { | |
4310 | swash_hv = MUTABLE_HV(SvRV(retval)); | |
4311 | ||
4312 | /* If we don't already know that there is a user-defined component to | |
4313 | * this swash, and the user has indicated they wish to know if there is | |
4314 | * one (by passing <flags_p>), find out */ | |
4315 | if (flags_p && ! (*flags_p & _CORE_SWASH_INIT_USER_DEFINED_PROPERTY)) { | |
4316 | SV** user_defined = hv_fetchs(swash_hv, "USER_DEFINED", FALSE); | |
4317 | if (user_defined && SvUV(*user_defined)) { | |
4318 | *flags_p |= _CORE_SWASH_INIT_USER_DEFINED_PROPERTY; | |
4319 | } | |
4320 | } | |
4321 | } | |
4322 | ||
36eb48b4 KW |
4323 | /* Make sure there is an inversion list for binary properties */ |
4324 | if (minbits == 1) { | |
4325 | SV** swash_invlistsvp = NULL; | |
4326 | SV* swash_invlist = NULL; | |
9a53f6cf | 4327 | bool invlist_in_swash_is_valid = FALSE; |
02c85471 FC |
4328 | bool swash_invlist_unclaimed = FALSE; /* whether swash_invlist has |
4329 | an unclaimed reference count */ | |
36eb48b4 | 4330 | |
9a53f6cf | 4331 | /* If this operation fetched a swash, get its already existing |
83199d38 | 4332 | * inversion list, or create one for it */ |
36eb48b4 | 4333 | |
83199d38 | 4334 | if (swash_hv) { |
5c9f4bd2 | 4335 | swash_invlistsvp = hv_fetchs(swash_hv, "V", FALSE); |
9a53f6cf KW |
4336 | if (swash_invlistsvp) { |
4337 | swash_invlist = *swash_invlistsvp; | |
4338 | invlist_in_swash_is_valid = TRUE; | |
4339 | } | |
4340 | else { | |
36eb48b4 | 4341 | swash_invlist = _swash_to_invlist(retval); |
02c85471 | 4342 | swash_invlist_unclaimed = TRUE; |
9a53f6cf KW |
4343 | } |
4344 | } | |
4345 | ||
4346 | /* If an inversion list was passed in, have to include it */ | |
4347 | if (invlist) { | |
4348 | ||
4349 | /* Any fetched swash will by now have an inversion list in it; | |
4350 | * otherwise <swash_invlist> will be NULL, indicating that we | |
4351 | * didn't fetch a swash */ | |
4352 | if (swash_invlist) { | |
4353 | ||
4354 | /* Add the passed-in inversion list, which invalidates the one | |
4355 | * already stored in the swash */ | |
4356 | invlist_in_swash_is_valid = FALSE; | |
eee4c920 | 4357 | SvREADONLY_off(swash_invlist); /* Turned on again below */ |
9a53f6cf KW |
4358 | _invlist_union(invlist, swash_invlist, &swash_invlist); |
4359 | } | |
4360 | else { | |
4361 | ||
87367d5f KW |
4362 | /* Here, there is no swash already. Set up a minimal one, if |
4363 | * we are going to return a swash */ | |
4364 | if ((int) _invlist_len(invlist) > invlist_swash_boundary) { | |
971d486f | 4365 | swash_hv = newHV(); |
4aca0fe6 | 4366 | retval = newRV_noinc(MUTABLE_SV(swash_hv)); |
87367d5f | 4367 | } |
9a53f6cf KW |
4368 | swash_invlist = invlist; |
4369 | } | |
9a53f6cf KW |
4370 | } |
4371 | ||
4372 | /* Here, we have computed the union of all the passed-in data. It may | |
4373 | * be that there was an inversion list in the swash which didn't get | |
538e84ed | 4374 | * touched; otherwise save the computed one */ |
87367d5f KW |
4375 | if (! invlist_in_swash_is_valid |
4376 | && (int) _invlist_len(swash_invlist) > invlist_swash_boundary) | |
4377 | { | |
5c9f4bd2 | 4378 | if (! hv_stores(MUTABLE_HV(SvRV(retval)), "V", swash_invlist)) |
69794297 KW |
4379 | { |
4380 | Perl_croak(aTHX_ "panic: hv_store() unexpectedly failed"); | |
4381 | } | |
cc34d8c5 FC |
4382 | /* We just stole a reference count. */ |
4383 | if (swash_invlist_unclaimed) swash_invlist_unclaimed = FALSE; | |
4384 | else SvREFCNT_inc_simple_void_NN(swash_invlist); | |
9a53f6cf | 4385 | } |
87367d5f | 4386 | |
ee3222e3 | 4387 | /* The result is immutable. Forbid attempts to change it. */ |
dbfdbd26 KW |
4388 | SvREADONLY_on(swash_invlist); |
4389 | ||
c41800a8 | 4390 | /* Use the inversion list stand-alone if small enough */ |
87367d5f KW |
4391 | if ((int) _invlist_len(swash_invlist) <= invlist_swash_boundary) { |
4392 | SvREFCNT_dec(retval); | |
02c85471 FC |
4393 | if (!swash_invlist_unclaimed) |
4394 | SvREFCNT_inc_simple_void_NN(swash_invlist); | |
4395 | retval = newRV_noinc(swash_invlist); | |
87367d5f | 4396 | } |
36eb48b4 KW |
4397 | } |
4398 | ||
2c1f00b9 YO |
4399 | CORE_SWASH_INIT_RETURN(retval); |
4400 | #undef CORE_SWASH_INIT_RETURN | |
a0ed51b3 LW |
4401 | } |
4402 | ||
035d37be JH |
4403 | |
4404 | /* This API is wrong for special case conversions since we may need to | |
4405 | * return several Unicode characters for a single Unicode character | |
4406 | * (see lib/unicore/SpecCase.txt) The SWASHGET in lib/utf8_heavy.pl is | |
4407 | * the lower-level routine, and it is similarly broken for returning | |
38684baa | 4408 | * multiple values. --jhi |
b9992569 | 4409 | * For those, you should use S__to_utf8_case() instead */ |
b0e3252e | 4410 | /* Now SWASHGET is recasted into S_swatch_get in this file. */ |
680c470c TS |
4411 | |
4412 | /* Note: | |
4413 | * Returns the value of property/mapping C<swash> for the first character | |
4414 | * of the string C<ptr>. If C<do_utf8> is true, the string C<ptr> is | |
4a4088c4 | 4415 | * assumed to be in well-formed UTF-8. If C<do_utf8> is false, the string C<ptr> |
3d0f8846 | 4416 | * is assumed to be in native 8-bit encoding. Caches the swatch in C<swash>. |
af2af982 KW |
4417 | * |
4418 | * A "swash" is a hash which contains initially the keys/values set up by | |
4419 | * SWASHNEW. The purpose is to be able to completely represent a Unicode | |
4420 | * property for all possible code points. Things are stored in a compact form | |
4421 | * (see utf8_heavy.pl) so that calculation is required to find the actual | |
4422 | * property value for a given code point. As code points are looked up, new | |
4423 | * key/value pairs are added to the hash, so that the calculation doesn't have | |
4424 | * to ever be re-done. Further, each calculation is done, not just for the | |
4425 | * desired one, but for a whole block of code points adjacent to that one. | |
4426 | * For binary properties on ASCII machines, the block is usually for 64 code | |
4427 | * points, starting with a code point evenly divisible by 64. Thus if the | |
4428 | * property value for code point 257 is requested, the code goes out and | |
4429 | * calculates the property values for all 64 code points between 256 and 319, | |
4430 | * and stores these as a single 64-bit long bit vector, called a "swatch", | |
4431 | * under the key for code point 256. The key is the UTF-8 encoding for code | |
4432 | * point 256, minus the final byte. Thus, if the length of the UTF-8 encoding | |
4433 | * for a code point is 13 bytes, the key will be 12 bytes long. If the value | |
4434 | * for code point 258 is then requested, this code realizes that it would be | |
4435 | * stored under the key for 256, and would find that value and extract the | |
4436 | * relevant bit, offset from 256. | |
4437 | * | |
4438 | * Non-binary properties are stored in as many bits as necessary to represent | |
4439 | * their values (32 currently, though the code is more general than that), not | |
fc273927 | 4440 | * as single bits, but the principle is the same: the value for each key is a |
af2af982 KW |
4441 | * vector that encompasses the property values for all code points whose UTF-8 |
4442 | * representations are represented by the key. That is, for all code points | |
4443 | * whose UTF-8 representations are length N bytes, and the key is the first N-1 | |
4444 | * bytes of that. | |
680c470c | 4445 | */ |
a0ed51b3 | 4446 | UV |
680c470c | 4447 | Perl_swash_fetch(pTHX_ SV *swash, const U8 *ptr, bool do_utf8) |
a0ed51b3 | 4448 | { |
ef8f7699 | 4449 | HV *const hv = MUTABLE_HV(SvRV(swash)); |
3568d838 JH |
4450 | U32 klen; |
4451 | U32 off; | |
9b56a019 | 4452 | STRLEN slen = 0; |
7d85a32c | 4453 | STRLEN needents; |
cfd0369c | 4454 | const U8 *tmps = NULL; |
979f2922 | 4455 | SV *swatch; |
08fb1ac5 | 4456 | const U8 c = *ptr; |
3568d838 | 4457 | |
7918f24d NC |
4458 | PERL_ARGS_ASSERT_SWASH_FETCH; |
4459 | ||
87367d5f KW |
4460 | /* If it really isn't a hash, it isn't really swash; must be an inversion |
4461 | * list */ | |
4462 | if (SvTYPE(hv) != SVt_PVHV) { | |
4463 | return _invlist_contains_cp((SV*)hv, | |
4464 | (do_utf8) | |
4465 | ? valid_utf8_to_uvchr(ptr, NULL) | |
4466 | : c); | |
4467 | } | |
4468 | ||
08fb1ac5 KW |
4469 | /* We store the values in a "swatch" which is a vec() value in a swash |
4470 | * hash. Code points 0-255 are a single vec() stored with key length | |
4471 | * (klen) 0. All other code points have a UTF-8 representation | |
4472 | * 0xAA..0xYY,0xZZ. A vec() is constructed containing all of them which | |
4473 | * share 0xAA..0xYY, which is the key in the hash to that vec. So the key | |
4474 | * length for them is the length of the encoded char - 1. ptr[klen] is the | |
4475 | * final byte in the sequence representing the character */ | |
4476 | if (!do_utf8 || UTF8_IS_INVARIANT(c)) { | |
4477 | klen = 0; | |
4478 | needents = 256; | |
4479 | off = c; | |
3568d838 | 4480 | } |
08fb1ac5 KW |
4481 | else if (UTF8_IS_DOWNGRADEABLE_START(c)) { |
4482 | klen = 0; | |
4483 | needents = 256; | |
a62b247b | 4484 | off = EIGHT_BIT_UTF8_TO_NATIVE(c, *(ptr + 1)); |
979f2922 TS |
4485 | } |
4486 | else { | |
08fb1ac5 KW |
4487 | klen = UTF8SKIP(ptr) - 1; |
4488 | ||
4489 | /* Each vec() stores 2**UTF_ACCUMULATION_SHIFT values. The offset into | |
4490 | * the vec is the final byte in the sequence. (In EBCDIC this is | |
4491 | * converted to I8 to get consecutive values.) To help you visualize | |
4492 | * all this: | |
4493 | * Straight 1047 After final byte | |
4494 | * UTF-8 UTF-EBCDIC I8 transform | |
4495 | * U+0400: \xD0\x80 \xB8\x41\x41 \xB8\x41\xA0 | |
4496 | * U+0401: \xD0\x81 \xB8\x41\x42 \xB8\x41\xA1 | |
4497 | * ... | |
4498 | * U+0409: \xD0\x89 \xB8\x41\x4A \xB8\x41\xA9 | |
4499 | * U+040A: \xD0\x8A \xB8\x41\x51 \xB8\x41\xAA | |
4500 | * ... | |
4501 | * U+0412: \xD0\x92 \xB8\x41\x59 \xB8\x41\xB2 | |
4502 | * U+0413: \xD0\x93 \xB8\x41\x62 \xB8\x41\xB3 | |
4503 | * ... | |
4504 | * U+041B: \xD0\x9B \xB8\x41\x6A \xB8\x41\xBB | |
4505 | * U+041C: \xD0\x9C \xB8\x41\x70 \xB8\x41\xBC | |
4506 | * ... | |
4507 | * U+041F: \xD0\x9F \xB8\x41\x73 \xB8\x41\xBF | |
4508 | * U+0420: \xD0\xA0 \xB8\x42\x41 \xB8\x42\x41 | |
4509 | * | |
4510 | * (There are no discontinuities in the elided (...) entries.) | |
4511 | * The UTF-8 key for these 33 code points is '\xD0' (which also is the | |
4512 | * key for the next 31, up through U+043F, whose UTF-8 final byte is | |
4513 | * \xBF). Thus in UTF-8, each key is for a vec() for 64 code points. | |
4514 | * The final UTF-8 byte, which ranges between \x80 and \xBF, is an | |
4515 | * index into the vec() swatch (after subtracting 0x80, which we | |
4516 | * actually do with an '&'). | |
4517 | * In UTF-EBCDIC, each key is for a 32 code point vec(). The first 32 | |
4518 | * code points above have key '\xB8\x41'. The final UTF-EBCDIC byte has | |
4519 | * dicontinuities which go away by transforming it into I8, and we | |
4520 | * effectively subtract 0xA0 to get the index. */ | |
979f2922 | 4521 | needents = (1 << UTF_ACCUMULATION_SHIFT); |
bc3632a8 | 4522 | off = NATIVE_UTF8_TO_I8(ptr[klen]) & UTF_CONTINUATION_MASK; |
979f2922 | 4523 | } |
7d85a32c | 4524 | |
a0ed51b3 | 4525 | /* |
4a4088c4 | 4526 | * This single-entry cache saves about 1/3 of the UTF-8 overhead in test |
a0ed51b3 LW |
4527 | * suite. (That is, only 7-8% overall over just a hash cache. Still, |
4528 | * it's nothing to sniff at.) Pity we usually come through at least | |
4529 | * two function calls to get here... | |
4530 | * | |
4531 | * NB: this code assumes that swatches are never modified, once generated! | |
4532 | */ | |
4533 | ||
3568d838 | 4534 | if (hv == PL_last_swash_hv && |
a0ed51b3 | 4535 | klen == PL_last_swash_klen && |
27da23d5 | 4536 | (!klen || memEQ((char *)ptr, (char *)PL_last_swash_key, klen)) ) |
a0ed51b3 LW |
4537 | { |
4538 | tmps = PL_last_swash_tmps; | |
4539 | slen = PL_last_swash_slen; | |
4540 | } | |
4541 | else { | |
4542 | /* Try our second-level swatch cache, kept in a hash. */ | |
e1ec3a88 | 4543 | SV** svp = hv_fetch(hv, (const char*)ptr, klen, FALSE); |
a0ed51b3 | 4544 | |
b0e3252e | 4545 | /* If not cached, generate it via swatch_get */ |
979f2922 | 4546 | if (!svp || !SvPOK(*svp) |
08fb1ac5 KW |
4547 | || !(tmps = (const U8*)SvPV_const(*svp, slen))) |
4548 | { | |
4549 | if (klen) { | |
4550 | const UV code_point = valid_utf8_to_uvchr(ptr, NULL); | |
4551 | swatch = swatch_get(swash, | |
4552 | code_point & ~((UV)needents - 1), | |
4553 | needents); | |
4554 | } | |
4555 | else { /* For the first 256 code points, the swatch has a key of | |
4556 | length 0 */ | |
4557 | swatch = swatch_get(swash, 0, needents); | |
4558 | } | |
979f2922 | 4559 | |
923e4eb5 | 4560 | if (IN_PERL_COMPILETIME) |
623e6609 | 4561 | CopHINTS_set(PL_curcop, PL_hints); |
a0ed51b3 | 4562 | |
979f2922 | 4563 | svp = hv_store(hv, (const char *)ptr, klen, swatch, 0); |
a0ed51b3 | 4564 | |
979f2922 TS |
4565 | if (!svp || !(tmps = (U8*)SvPV(*svp, slen)) |
4566 | || (slen << 3) < needents) | |
5637ef5b | 4567 | Perl_croak(aTHX_ "panic: swash_fetch got improper swatch, " |
147e3846 | 4568 | "svp=%p, tmps=%p, slen=%" UVuf ", needents=%" UVuf, |
5637ef5b | 4569 | svp, tmps, (UV)slen, (UV)needents); |
a0ed51b3 LW |
4570 | } |
4571 | ||
4572 | PL_last_swash_hv = hv; | |
16d8f38a | 4573 | assert(klen <= sizeof(PL_last_swash_key)); |
eac04b2e | 4574 | PL_last_swash_klen = (U8)klen; |
cfd0369c NC |
4575 | /* FIXME change interpvar.h? */ |
4576 | PL_last_swash_tmps = (U8 *) tmps; | |
a0ed51b3 LW |
4577 | PL_last_swash_slen = slen; |
4578 | if (klen) | |
4579 | Copy(ptr, PL_last_swash_key, klen, U8); | |
4580 | } | |
4581 | ||
9faf8d75 | 4582 | switch ((int)((slen << 3) / needents)) { |
a0ed51b3 | 4583 | case 1: |
e7aca353 | 4584 | return ((UV) tmps[off >> 3] & (1 << (off & 7))) != 0; |
a0ed51b3 | 4585 | case 8: |
e7aca353 | 4586 | return ((UV) tmps[off]); |
a0ed51b3 LW |
4587 | case 16: |
4588 | off <<= 1; | |
e7aca353 JH |
4589 | return |
4590 | ((UV) tmps[off ] << 8) + | |
4591 | ((UV) tmps[off + 1]); | |
a0ed51b3 LW |
4592 | case 32: |
4593 | off <<= 2; | |
e7aca353 JH |
4594 | return |
4595 | ((UV) tmps[off ] << 24) + | |
4596 | ((UV) tmps[off + 1] << 16) + | |
4597 | ((UV) tmps[off + 2] << 8) + | |
4598 | ((UV) tmps[off + 3]); | |
a0ed51b3 | 4599 | } |
5637ef5b | 4600 | Perl_croak(aTHX_ "panic: swash_fetch got swatch of unexpected bit width, " |
147e3846 | 4601 | "slen=%" UVuf ", needents=%" UVuf, (UV)slen, (UV)needents); |
670f1322 | 4602 | NORETURN_FUNCTION_END; |
a0ed51b3 | 4603 | } |
2b9d42f0 | 4604 | |
319009ee KW |
4605 | /* Read a single line of the main body of the swash input text. These are of |
4606 | * the form: | |
4607 | * 0053 0056 0073 | |
4608 | * where each number is hex. The first two numbers form the minimum and | |
4609 | * maximum of a range, and the third is the value associated with the range. | |
4610 | * Not all swashes should have a third number | |
4611 | * | |
4612 | * On input: l points to the beginning of the line to be examined; it points | |
4613 | * to somewhere in the string of the whole input text, and is | |
4614 | * terminated by a \n or the null string terminator. | |
4615 | * lend points to the null terminator of that string | |
4616 | * wants_value is non-zero if the swash expects a third number | |
4617 | * typestr is the name of the swash's mapping, like 'ToLower' | |
4618 | * On output: *min, *max, and *val are set to the values read from the line. | |
4619 | * returns a pointer just beyond the line examined. If there was no | |
4620 | * valid min number on the line, returns lend+1 | |
4621 | */ | |
4622 | ||
4623 | STATIC U8* | |
4624 | S_swash_scan_list_line(pTHX_ U8* l, U8* const lend, UV* min, UV* max, UV* val, | |
4625 | const bool wants_value, const U8* const typestr) | |
4626 | { | |
4627 | const int typeto = typestr[0] == 'T' && typestr[1] == 'o'; | |
4628 | STRLEN numlen; /* Length of the number */ | |
02470786 KW |
4629 | I32 flags = PERL_SCAN_SILENT_ILLDIGIT |
4630 | | PERL_SCAN_DISALLOW_PREFIX | |
4631 | | PERL_SCAN_SILENT_NON_PORTABLE; | |
319009ee KW |
4632 | |
4633 | /* nl points to the next \n in the scan */ | |
4634 | U8* const nl = (U8*)memchr(l, '\n', lend - l); | |
4635 | ||
95543e92 KW |
4636 | PERL_ARGS_ASSERT_SWASH_SCAN_LIST_LINE; |
4637 | ||
319009ee KW |
4638 | /* Get the first number on the line: the range minimum */ |
4639 | numlen = lend - l; | |
4640 | *min = grok_hex((char *)l, &numlen, &flags, NULL); | |
c88850db | 4641 | *max = *min; /* So can never return without setting max */ |
319009ee KW |
4642 | if (numlen) /* If found a hex number, position past it */ |
4643 | l += numlen; | |
4644 | else if (nl) { /* Else, go handle next line, if any */ | |
4645 | return nl + 1; /* 1 is length of "\n" */ | |
4646 | } | |
4647 | else { /* Else, no next line */ | |
4648 | return lend + 1; /* to LIST's end at which \n is not found */ | |
4649 | } | |
4650 | ||
4651 | /* The max range value follows, separated by a BLANK */ | |
4652 | if (isBLANK(*l)) { | |
4653 | ++l; | |
02470786 KW |
4654 | flags = PERL_SCAN_SILENT_ILLDIGIT |
4655 | | PERL_SCAN_DISALLOW_PREFIX | |
4656 | | PERL_SCAN_SILENT_NON_PORTABLE; | |
319009ee KW |
4657 | numlen = lend - l; |
4658 | *max = grok_hex((char *)l, &numlen, &flags, NULL); | |
4659 | if (numlen) | |
4660 | l += numlen; | |
4661 | else /* If no value here, it is a single element range */ | |
4662 | *max = *min; | |
4663 | ||
4664 | /* Non-binary tables have a third entry: what the first element of the | |
24303724 | 4665 | * range maps to. The map for those currently read here is in hex */ |
319009ee KW |
4666 | if (wants_value) { |
4667 | if (isBLANK(*l)) { | |
4668 | ++l; | |
f2a7d0fc KW |
4669 | flags = PERL_SCAN_SILENT_ILLDIGIT |
4670 | | PERL_SCAN_DISALLOW_PREFIX | |
4671 | | PERL_SCAN_SILENT_NON_PORTABLE; | |
4672 | numlen = lend - l; | |
4673 | *val = grok_hex((char *)l, &numlen, &flags, NULL); | |
4674 | if (numlen) | |
4675 | l += numlen; | |
4676 | else | |
4677 | *val = 0; | |
319009ee KW |
4678 | } |
4679 | else { | |
4680 | *val = 0; | |
4681 | if (typeto) { | |
dcbac5bb | 4682 | /* diag_listed_as: To%s: illegal mapping '%s' */ |
319009ee KW |
4683 | Perl_croak(aTHX_ "%s: illegal mapping '%s'", |
4684 | typestr, l); | |
4685 | } | |
4686 | } | |
4687 | } | |
4688 | else | |
4689 | *val = 0; /* bits == 1, then any val should be ignored */ | |
4690 | } | |
4691 | else { /* Nothing following range min, should be single element with no | |
4692 | mapping expected */ | |
319009ee KW |
4693 | if (wants_value) { |
4694 | *val = 0; | |
4695 | if (typeto) { | |
dcbac5bb | 4696 | /* diag_listed_as: To%s: illegal mapping '%s' */ |
319009ee KW |
4697 | Perl_croak(aTHX_ "%s: illegal mapping '%s'", typestr, l); |
4698 | } | |
4699 | } | |
4700 | else | |
4701 | *val = 0; /* bits == 1, then val should be ignored */ | |
4702 | } | |
4703 | ||
4704 | /* Position to next line if any, or EOF */ | |
4705 | if (nl) | |
4706 | l = nl + 1; | |
4707 | else | |
4708 | l = lend; | |
4709 | ||
4710 | return l; | |
4711 | } | |
4712 | ||
979f2922 TS |
4713 | /* Note: |
4714 | * Returns a swatch (a bit vector string) for a code point sequence | |
4715 | * that starts from the value C<start> and comprises the number C<span>. | |
4716 | * A C<swash> must be an object created by SWASHNEW (see lib/utf8_heavy.pl). | |
4717 | * Should be used via swash_fetch, which will cache the swatch in C<swash>. | |
4718 | */ | |
4719 | STATIC SV* | |
b0e3252e | 4720 | S_swatch_get(pTHX_ SV* swash, UV start, UV span) |
979f2922 TS |
4721 | { |
4722 | SV *swatch; | |
77f9f126 | 4723 | U8 *l, *lend, *x, *xend, *s, *send; |
979f2922 | 4724 | STRLEN lcur, xcur, scur; |
ef8f7699 | 4725 | HV *const hv = MUTABLE_HV(SvRV(swash)); |
5c9f4bd2 | 4726 | SV** const invlistsvp = hv_fetchs(hv, "V", FALSE); |
36eb48b4 | 4727 | |
88d45d28 KW |
4728 | SV** listsvp = NULL; /* The string containing the main body of the table */ |
4729 | SV** extssvp = NULL; | |
4730 | SV** invert_it_svp = NULL; | |
4731 | U8* typestr = NULL; | |
786861f5 KW |
4732 | STRLEN bits; |
4733 | STRLEN octets; /* if bits == 1, then octets == 0 */ | |
4734 | UV none; | |
4735 | UV end = start + span; | |
972dd592 | 4736 | |
36eb48b4 | 4737 | if (invlistsvp == NULL) { |
786861f5 KW |
4738 | SV** const bitssvp = hv_fetchs(hv, "BITS", FALSE); |
4739 | SV** const nonesvp = hv_fetchs(hv, "NONE", FALSE); | |
4740 | SV** const typesvp = hv_fetchs(hv, "TYPE", FALSE); | |
4741 | extssvp = hv_fetchs(hv, "EXTRAS", FALSE); | |
4742 | listsvp = hv_fetchs(hv, "LIST", FALSE); | |
4743 | invert_it_svp = hv_fetchs(hv, "INVERT_IT", FALSE); | |
4744 | ||
4745 | bits = SvUV(*bitssvp); | |
4746 | none = SvUV(*nonesvp); | |
4747 | typestr = (U8*)SvPV_nolen(*typesvp); | |
4748 | } | |
36eb48b4 KW |
4749 | else { |
4750 | bits = 1; | |
4751 | none = 0; | |
4752 | } | |
786861f5 | 4753 | octets = bits >> 3; /* if bits == 1, then octets == 0 */ |
979f2922 | 4754 | |
b0e3252e | 4755 | PERL_ARGS_ASSERT_SWATCH_GET; |
7918f24d | 4756 | |
979f2922 | 4757 | if (bits != 1 && bits != 8 && bits != 16 && bits != 32) { |
147e3846 | 4758 | Perl_croak(aTHX_ "panic: swatch_get doesn't expect bits %" UVuf, |
660a4616 | 4759 | (UV)bits); |
979f2922 TS |
4760 | } |
4761 | ||
84ea5ef6 KW |
4762 | /* If overflowed, use the max possible */ |
4763 | if (end < start) { | |
4764 | end = UV_MAX; | |
4765 | span = end - start; | |
4766 | } | |
4767 | ||
979f2922 | 4768 | /* create and initialize $swatch */ |
979f2922 | 4769 | scur = octets ? (span * octets) : (span + 7) / 8; |
e524fe40 NC |
4770 | swatch = newSV(scur); |
4771 | SvPOK_on(swatch); | |
979f2922 TS |
4772 | s = (U8*)SvPVX(swatch); |
4773 | if (octets && none) { | |
0bd48802 | 4774 | const U8* const e = s + scur; |
979f2922 TS |
4775 | while (s < e) { |
4776 | if (bits == 8) | |
4777 | *s++ = (U8)(none & 0xff); | |
4778 | else if (bits == 16) { | |
4779 | *s++ = (U8)((none >> 8) & 0xff); | |
4780 | *s++ = (U8)( none & 0xff); | |
4781 | } | |
4782 | else if (bits == 32) { | |
4783 | *s++ = (U8)((none >> 24) & 0xff); | |
4784 | *s++ = (U8)((none >> 16) & 0xff); | |
4785 | *s++ = (U8)((none >> 8) & 0xff); | |
4786 | *s++ = (U8)( none & 0xff); | |
4787 | } | |
4788 | } | |
4789 | *s = '\0'; | |
4790 | } | |
4791 | else { | |
4792 | (void)memzero((U8*)s, scur + 1); | |
4793 | } | |
4794 | SvCUR_set(swatch, scur); | |
4795 | s = (U8*)SvPVX(swatch); | |
4796 | ||
36eb48b4 KW |
4797 | if (invlistsvp) { /* If has an inversion list set up use that */ |
4798 | _invlist_populate_swatch(*invlistsvp, start, end, s); | |
4799 | return swatch; | |
4800 | } | |
4801 | ||
4802 | /* read $swash->{LIST} */ | |
979f2922 TS |
4803 | l = (U8*)SvPV(*listsvp, lcur); |
4804 | lend = l + lcur; | |
4805 | while (l < lend) { | |
8ed25d53 | 4806 | UV min, max, val, upper; |
95543e92 KW |
4807 | l = swash_scan_list_line(l, lend, &min, &max, &val, |
4808 | cBOOL(octets), typestr); | |
319009ee | 4809 | if (l > lend) { |
979f2922 TS |
4810 | break; |
4811 | } | |
4812 | ||
972dd592 | 4813 | /* If looking for something beyond this range, go try the next one */ |
979f2922 TS |
4814 | if (max < start) |
4815 | continue; | |
4816 | ||
8ed25d53 KW |
4817 | /* <end> is generally 1 beyond where we want to set things, but at the |
4818 | * platform's infinity, where we can't go any higher, we want to | |
4819 | * include the code point at <end> */ | |
4820 | upper = (max < end) | |
4821 | ? max | |
4822 | : (max != UV_MAX || end != UV_MAX) | |
4823 | ? end - 1 | |
4824 | : end; | |
4825 | ||
979f2922 | 4826 | if (octets) { |
35da51f7 | 4827 | UV key; |
979f2922 TS |
4828 | if (min < start) { |
4829 | if (!none || val < none) { | |
4830 | val += start - min; | |
4831 | } | |
4832 | min = start; | |
4833 | } | |
8ed25d53 | 4834 | for (key = min; key <= upper; key++) { |
979f2922 | 4835 | STRLEN offset; |
979f2922 TS |
4836 | /* offset must be non-negative (start <= min <= key < end) */ |
4837 | offset = octets * (key - start); | |
4838 | if (bits == 8) | |
4839 | s[offset] = (U8)(val & 0xff); | |
4840 | else if (bits == 16) { | |
4841 | s[offset ] = (U8)((val >> 8) & 0xff); | |
4842 | s[offset + 1] = (U8)( val & 0xff); | |
4843 | } | |
4844 | else if (bits == 32) { | |
4845 | s[offset ] = (U8)((val >> 24) & 0xff); | |
4846 | s[offset + 1] = (U8)((val >> 16) & 0xff); | |
4847 | s[offset + 2] = (U8)((val >> 8) & 0xff); | |
4848 | s[offset + 3] = (U8)( val & 0xff); | |
4849 | } | |
4850 | ||
4851 | if (!none || val < none) | |
4852 | ++val; | |
4853 | } | |
4854 | } | |
711a919c | 4855 | else { /* bits == 1, then val should be ignored */ |
35da51f7 | 4856 | UV key; |
979f2922 TS |
4857 | if (min < start) |
4858 | min = start; | |
6cb05c12 | 4859 | |
8ed25d53 | 4860 | for (key = min; key <= upper; key++) { |
0bd48802 | 4861 | const STRLEN offset = (STRLEN)(key - start); |
979f2922 TS |
4862 | s[offset >> 3] |= 1 << (offset & 7); |
4863 | } | |
4864 | } | |
4865 | } /* while */ | |
979f2922 | 4866 | |
9479a769 | 4867 | /* Invert if the data says it should be. Assumes that bits == 1 */ |
77f9f126 | 4868 | if (invert_it_svp && SvUV(*invert_it_svp)) { |
0bda3001 KW |
4869 | |
4870 | /* Unicode properties should come with all bits above PERL_UNICODE_MAX | |
4871 | * be 0, and their inversion should also be 0, as we don't succeed any | |
4872 | * Unicode property matches for non-Unicode code points */ | |
4873 | if (start <= PERL_UNICODE_MAX) { | |
4874 | ||
4875 | /* The code below assumes that we never cross the | |
4876 | * Unicode/above-Unicode boundary in a range, as otherwise we would | |
4877 | * have to figure out where to stop flipping the bits. Since this | |
4878 | * boundary is divisible by a large power of 2, and swatches comes | |
4879 | * in small powers of 2, this should be a valid assumption */ | |
4880 | assert(start + span - 1 <= PERL_UNICODE_MAX); | |
4881 | ||
507a8485 KW |
4882 | send = s + scur; |
4883 | while (s < send) { | |
4884 | *s = ~(*s); | |
4885 | s++; | |
4886 | } | |
0bda3001 | 4887 | } |
77f9f126 KW |
4888 | } |
4889 | ||
d73c39c5 KW |
4890 | /* read $swash->{EXTRAS} |
4891 | * This code also copied to swash_to_invlist() below */ | |
979f2922 TS |
4892 | x = (U8*)SvPV(*extssvp, xcur); |
4893 | xend = x + xcur; | |
4894 | while (x < xend) { | |
4895 | STRLEN namelen; | |
4896 | U8 *namestr; | |
4897 | SV** othersvp; | |
4898 | HV* otherhv; | |
4899 | STRLEN otherbits; | |
4900 | SV **otherbitssvp, *other; | |
711a919c | 4901 | U8 *s, *o, *nl; |
979f2922 TS |
4902 | STRLEN slen, olen; |
4903 | ||
35da51f7 | 4904 | const U8 opc = *x++; |
979f2922 TS |
4905 | if (opc == '\n') |
4906 | continue; | |
4907 | ||
4908 | nl = (U8*)memchr(x, '\n', xend - x); | |
4909 | ||
4910 | if (opc != '-' && opc != '+' && opc != '!' && opc != '&') { | |
4911 | if (nl) { | |
4912 | x = nl + 1; /* 1 is length of "\n" */ | |
4913 | continue; | |
4914 | } | |
4915 | else { | |
4916 | x = xend; /* to EXTRAS' end at which \n is not found */ | |
4917 | break; | |
4918 | } | |
4919 | } | |
4920 | ||
4921 | namestr = x; | |
4922 | if (nl) { | |
4923 | namelen = nl - namestr; | |
4924 | x = nl + 1; | |
4925 | } | |
4926 | else { | |
4927 | namelen = xend - namestr; | |
4928 | x = xend; | |
4929 | } | |
4930 | ||
4931 | othersvp = hv_fetch(hv, (char *)namestr, namelen, FALSE); | |
ef8f7699 | 4932 | otherhv = MUTABLE_HV(SvRV(*othersvp)); |
017a3ce5 | 4933 | otherbitssvp = hv_fetchs(otherhv, "BITS", FALSE); |
979f2922 TS |
4934 | otherbits = (STRLEN)SvUV(*otherbitssvp); |
4935 | if (bits < otherbits) | |
5637ef5b | 4936 | Perl_croak(aTHX_ "panic: swatch_get found swatch size mismatch, " |
147e3846 | 4937 | "bits=%" UVuf ", otherbits=%" UVuf, (UV)bits, (UV)otherbits); |
979f2922 TS |
4938 | |
4939 | /* The "other" swatch must be destroyed after. */ | |
b0e3252e | 4940 | other = swatch_get(*othersvp, start, span); |
979f2922 TS |
4941 | o = (U8*)SvPV(other, olen); |
4942 | ||
4943 | if (!olen) | |
b0e3252e | 4944 | Perl_croak(aTHX_ "panic: swatch_get got improper swatch"); |
979f2922 TS |
4945 | |
4946 | s = (U8*)SvPV(swatch, slen); | |
4947 | if (bits == 1 && otherbits == 1) { | |
4948 | if (slen != olen) | |
5637ef5b | 4949 | Perl_croak(aTHX_ "panic: swatch_get found swatch length " |
147e3846 | 4950 | "mismatch, slen=%" UVuf ", olen=%" UVuf, |
5637ef5b | 4951 | (UV)slen, (UV)olen); |
979f2922 TS |
4952 | |
4953 | switch (opc) { | |
4954 | case '+': | |
4955 | while (slen--) | |
4956 | *s++ |= *o++; | |
4957 | break; | |
4958 | case '!': | |
4959 | while (slen--) | |
4960 | *s++ |= ~*o++; | |
4961 | break; | |
4962 | case '-': | |
4963 | while (slen--) | |
4964 | *s++ &= ~*o++; | |
4965 | break; | |
4966 | case '&': | |
4967 | while (slen--) | |
4968 | *s++ &= *o++; | |
4969 | break; | |
4970 | default: | |
4971 | break; | |
4972 | } | |
4973 | } | |
711a919c | 4974 | else { |
979f2922 TS |
4975 | STRLEN otheroctets = otherbits >> 3; |
4976 | STRLEN offset = 0; | |
35da51f7 | 4977 | U8* const send = s + slen; |
979f2922 TS |
4978 | |
4979 | while (s < send) { | |
4980 | UV otherval = 0; | |
4981 | ||
4982 | if (otherbits == 1) { | |
4983 | otherval = (o[offset >> 3] >> (offset & 7)) & 1; | |
4984 | ++offset; | |
4985 | } | |
4986 | else { | |
4987 | STRLEN vlen = otheroctets; | |
4988 | otherval = *o++; | |
4989 | while (--vlen) { | |
4990 | otherval <<= 8; | |
4991 | otherval |= *o++; | |
4992 | } | |
4993 | } | |
4994 | ||
711a919c | 4995 | if (opc == '+' && otherval) |
6f207bd3 | 4996 | NOOP; /* replace with otherval */ |
979f2922 TS |
4997 | else if (opc == '!' && !otherval) |
4998 | otherval = 1; | |
4999 | else if (opc == '-' && otherval) | |
5000 | otherval = 0; | |
5001 | else if (opc == '&' && !otherval) | |
5002 | otherval = 0; | |
5003 | else { | |
711a919c | 5004 | s += octets; /* no replacement */ |
979f2922 TS |
5005 | continue; |
5006 | } | |
5007 | ||
5008 | if (bits == 8) | |
5009 | *s++ = (U8)( otherval & 0xff); | |
5010 | else if (bits == 16) { | |
5011 | *s++ = (U8)((otherval >> 8) & 0xff); | |
5012 | *s++ = (U8)( otherval & 0xff); | |
5013 | } | |
5014 | else if (bits == 32) { | |
5015 | *s++ = (U8)((otherval >> 24) & 0xff); | |
5016 | *s++ = (U8)((otherval >> 16) & 0xff); | |
5017 | *s++ = (U8)((otherval >> 8) & 0xff); | |
5018 | *s++ = (U8)( otherval & 0xff); | |
5019 | } | |
5020 | } | |
5021 | } | |
5022 | sv_free(other); /* through with it! */ | |
5023 | } /* while */ | |
5024 | return swatch; | |
5025 | } | |
5026 | ||
064c021d | 5027 | HV* |
4c2e1131 | 5028 | Perl__swash_inversion_hash(pTHX_ SV* const swash) |
064c021d KW |
5029 | { |
5030 | ||
79a2a0e8 | 5031 | /* Subject to change or removal. For use only in regcomp.c and regexec.c |
5662e334 KW |
5032 | * Can't be used on a property that is subject to user override, as it |
5033 | * relies on the value of SPECIALS in the swash which would be set by | |
5034 | * utf8_heavy.pl to the hash in the non-overriden file, and hence is not set | |
5035 | * for overridden properties | |
064c021d KW |
5036 | * |
5037 | * Returns a hash which is the inversion and closure of a swash mapping. | |
5038 | * For example, consider the input lines: | |
5039 | * 004B 006B | |
5040 | * 004C 006C | |
5041 | * 212A 006B | |
5042 | * | |
4a4088c4 | 5043 | * The returned hash would have two keys, the UTF-8 for 006B and the UTF-8 for |
064c021d | 5044 | * 006C. The value for each key is an array. For 006C, the array would |
4a4088c4 KW |
5045 | * have two elements, the UTF-8 for itself, and for 004C. For 006B, there |
5046 | * would be three elements in its array, the UTF-8 for 006B, 004B and 212A. | |
064c021d | 5047 | * |
538e84ed KW |
5048 | * Note that there are no elements in the hash for 004B, 004C, 212A. The |
5049 | * keys are only code points that are folded-to, so it isn't a full closure. | |
5050 | * | |
064c021d KW |
5051 | * Essentially, for any code point, it gives all the code points that map to |
5052 | * it, or the list of 'froms' for that point. | |
5053 | * | |
5662e334 KW |
5054 | * Currently it ignores any additions or deletions from other swashes, |
5055 | * looking at just the main body of the swash, and if there are SPECIALS | |
5056 | * in the swash, at that hash | |
5057 | * | |
5058 | * The specials hash can be extra code points, and most likely consists of | |
5059 | * maps from single code points to multiple ones (each expressed as a string | |
4a4088c4 | 5060 | * of UTF-8 characters). This function currently returns only 1-1 mappings. |
5662e334 KW |
5061 | * However consider this possible input in the specials hash: |
5062 | * "\xEF\xAC\x85" => "\x{0073}\x{0074}", # U+FB05 => 0073 0074 | |
5063 | * "\xEF\xAC\x86" => "\x{0073}\x{0074}", # U+FB06 => 0073 0074 | |
5064 | * | |
5065 | * Both FB05 and FB06 map to the same multi-char sequence, which we don't | |
5066 | * currently handle. But it also means that FB05 and FB06 are equivalent in | |
5067 | * a 1-1 mapping which we should handle, and this relationship may not be in | |
5068 | * the main table. Therefore this function examines all the multi-char | |
74894415 KW |
5069 | * sequences and adds the 1-1 mappings that come out of that. |
5070 | * | |
5071 | * XXX This function was originally intended to be multipurpose, but its | |
5072 | * only use is quite likely to remain for constructing the inversion of | |
5073 | * the CaseFolding (//i) property. If it were more general purpose for | |
5074 | * regex patterns, it would have to do the FB05/FB06 game for simple folds, | |
5075 | * because certain folds are prohibited under /iaa and /il. As an example, | |
5076 | * in Unicode 3.0.1 both U+0130 and U+0131 fold to 'i', and hence are both | |
5077 | * equivalent under /i. But under /iaa and /il, the folds to 'i' are | |
5078 | * prohibited, so we would not figure out that they fold to each other. | |
5079 | * Code could be written to automatically figure this out, similar to the | |
5080 | * code that does this for multi-character folds, but this is the only case | |
5081 | * where something like this is ever likely to happen, as all the single | |
7ee537e6 | 5082 | * char folds to the 0-255 range are now quite settled. Instead there is a |
74894415 KW |
5083 | * little special code that is compiled only for this Unicode version. This |
5084 | * is smaller and didn't require much coding time to do. But this makes | |
5085 | * this routine strongly tied to being used just for CaseFolding. If ever | |
5086 | * it should be generalized, this would have to be fixed */ | |
064c021d KW |
5087 | |
5088 | U8 *l, *lend; | |
5089 | STRLEN lcur; | |
5090 | HV *const hv = MUTABLE_HV(SvRV(swash)); | |
5091 | ||
923b6d4e KW |
5092 | /* The string containing the main body of the table. This will have its |
5093 | * assertion fail if the swash has been converted to its inversion list */ | |
064c021d KW |
5094 | SV** const listsvp = hv_fetchs(hv, "LIST", FALSE); |
5095 | ||
5096 | SV** const typesvp = hv_fetchs(hv, "TYPE", FALSE); | |
5097 | SV** const bitssvp = hv_fetchs(hv, "BITS", FALSE); | |
5098 | SV** const nonesvp = hv_fetchs(hv, "NONE", FALSE); | |
5099 | /*SV** const extssvp = hv_fetchs(hv, "EXTRAS", FALSE);*/ | |
5100 | const U8* const typestr = (U8*)SvPV_nolen(*typesvp); | |
5101 | const STRLEN bits = SvUV(*bitssvp); | |
5102 | const STRLEN octets = bits >> 3; /* if bits == 1, then octets == 0 */ | |
5103 | const UV none = SvUV(*nonesvp); | |
5662e334 | 5104 | SV **specials_p = hv_fetchs(hv, "SPECIALS", 0); |
064c021d KW |
5105 | |
5106 | HV* ret = newHV(); | |
5107 | ||
5108 | PERL_ARGS_ASSERT__SWASH_INVERSION_HASH; | |
5109 | ||
5110 | /* Must have at least 8 bits to get the mappings */ | |
5111 | if (bits != 8 && bits != 16 && bits != 32) { | |
56576a04 KW |
5112 | Perl_croak(aTHX_ "panic: swash_inversion_hash doesn't expect bits %" |
5113 | UVuf, (UV)bits); | |
064c021d KW |
5114 | } |
5115 | ||
5662e334 KW |
5116 | if (specials_p) { /* It might be "special" (sometimes, but not always, a |
5117 | mapping to more than one character */ | |
5118 | ||
5119 | /* Construct an inverse mapping hash for the specials */ | |
5120 | HV * const specials_hv = MUTABLE_HV(SvRV(*specials_p)); | |
5121 | HV * specials_inverse = newHV(); | |
5122 | char *char_from; /* the lhs of the map */ | |
5123 | I32 from_len; /* its byte length */ | |
5124 | char *char_to; /* the rhs of the map */ | |
5125 | I32 to_len; /* its byte length */ | |
5126 | SV *sv_to; /* and in a sv */ | |
5127 | AV* from_list; /* list of things that map to each 'to' */ | |
5128 | ||
5129 | hv_iterinit(specials_hv); | |
5130 | ||
4a4088c4 KW |
5131 | /* The keys are the characters (in UTF-8) that map to the corresponding |
5132 | * UTF-8 string value. Iterate through the list creating the inverse | |
5662e334 KW |
5133 | * list. */ |
5134 | while ((sv_to = hv_iternextsv(specials_hv, &char_from, &from_len))) { | |
5135 | SV** listp; | |
5136 | if (! SvPOK(sv_to)) { | |
5637ef5b NC |
5137 | Perl_croak(aTHX_ "panic: value returned from hv_iternextsv() " |
5138 | "unexpectedly is not a string, flags=%lu", | |
5139 | (unsigned long)SvFLAGS(sv_to)); | |
5662e334 | 5140 | } |
147e3846 | 5141 | /*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 |
5142 | |
5143 | /* Each key in the inverse list is a mapped-to value, and the key's | |
4a4088c4 | 5144 | * hash value is a list of the strings (each in UTF-8) that map to |
5662e334 KW |
5145 | * it. Those strings are all one character long */ |
5146 | if ((listp = hv_fetch(specials_inverse, | |
5147 | SvPVX(sv_to), | |
5148 | SvCUR(sv_to), 0))) | |
5149 | { | |
5150 | from_list = (AV*) *listp; | |
5151 | } | |
5152 | else { /* No entry yet for it: create one */ | |
5153 | from_list = newAV(); | |
5154 | if (! hv_store(specials_inverse, | |
5155 | SvPVX(sv_to), | |
5156 | SvCUR(sv_to), | |
5157 | (SV*) from_list, 0)) | |
5158 | { | |
5159 | Perl_croak(aTHX_ "panic: hv_store() unexpectedly failed"); | |
5160 | } | |
5161 | } | |
5162 | ||
5163 | /* Here have the list associated with this 'to' (perhaps newly | |
5164 | * created and empty). Just add to it. Note that we ASSUME that | |
5165 | * the input is guaranteed to not have duplications, so we don't | |
5166 | * check for that. Duplications just slow down execution time. */ | |
5167 | av_push(from_list, newSVpvn_utf8(char_from, from_len, TRUE)); | |
5168 | } | |
5169 | ||
5170 | /* Here, 'specials_inverse' contains the inverse mapping. Go through | |
5171 | * it looking for cases like the FB05/FB06 examples above. There would | |
5172 | * be an entry in the hash like | |
5173 | * 'st' => [ FB05, FB06 ] | |
5174 | * In this example we will create two lists that get stored in the | |
5175 | * returned hash, 'ret': | |
5176 | * FB05 => [ FB05, FB06 ] | |
5177 | * FB06 => [ FB05, FB06 ] | |
5178 | * | |
5179 | * Note that there is nothing to do if the array only has one element. | |
5180 | * (In the normal 1-1 case handled below, we don't have to worry about | |
5181 | * two lists, as everything gets tied to the single list that is | |
5182 | * generated for the single character 'to'. But here, we are omitting | |
5183 | * that list, ('st' in the example), so must have multiple lists.) */ | |
5184 | while ((from_list = (AV *) hv_iternextsv(specials_inverse, | |
5185 | &char_to, &to_len))) | |
5186 | { | |
9506e945 | 5187 | if (av_tindex_skip_len_mg(from_list) > 0) { |
c70927a6 | 5188 | SSize_t i; |
5662e334 KW |
5189 | |
5190 | /* We iterate over all combinations of i,j to place each code | |
5191 | * point on each list */ | |
9506e945 | 5192 | for (i = 0; i <= av_tindex_skip_len_mg(from_list); i++) { |
c70927a6 | 5193 | SSize_t j; |
5662e334 KW |
5194 | AV* i_list = newAV(); |
5195 | SV** entryp = av_fetch(from_list, i, FALSE); | |
5196 | if (entryp == NULL) { | |
56576a04 KW |
5197 | Perl_croak(aTHX_ "panic: av_fetch() unexpectedly" |
5198 | " failed"); | |
5662e334 KW |
5199 | } |
5200 | if (hv_fetch(ret, SvPVX(*entryp), SvCUR(*entryp), FALSE)) { | |
56576a04 KW |
5201 | Perl_croak(aTHX_ "panic: unexpected entry for %s", |
5202 | SvPVX(*entryp)); | |
5662e334 KW |
5203 | } |
5204 | if (! hv_store(ret, SvPVX(*entryp), SvCUR(*entryp), | |
5205 | (SV*) i_list, FALSE)) | |
5206 | { | |
5207 | Perl_croak(aTHX_ "panic: hv_store() unexpectedly failed"); | |
5208 | } | |
5209 | ||
538e84ed | 5210 | /* For DEBUG_U: UV u = valid_utf8_to_uvchr((U8*) SvPVX(*entryp), 0);*/ |
9506e945 | 5211 | for (j = 0; j <= av_tindex_skip_len_mg(from_list); j++) { |
5662e334 KW |
5212 | entryp = av_fetch(from_list, j, FALSE); |
5213 | if (entryp == NULL) { | |
5214 | Perl_croak(aTHX_ "panic: av_fetch() unexpectedly failed"); | |
5215 | } | |
5216 | ||
5217 | /* When i==j this adds itself to the list */ | |
4b88fb76 KW |
5218 | av_push(i_list, newSVuv(utf8_to_uvchr_buf( |
5219 | (U8*) SvPVX(*entryp), | |
5220 | (U8*) SvPVX(*entryp) + SvCUR(*entryp), | |
5221 | 0))); | |
147e3846 | 5222 | /*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 |
5223 | } |
5224 | } | |
5225 | } | |
5226 | } | |
5227 | SvREFCNT_dec(specials_inverse); /* done with it */ | |
5228 | } /* End of specials */ | |
5229 | ||
064c021d | 5230 | /* read $swash->{LIST} */ |
74894415 KW |
5231 | |
5232 | #if UNICODE_MAJOR_VERSION == 3 \ | |
5233 | && UNICODE_DOT_VERSION == 0 \ | |
5234 | && UNICODE_DOT_DOT_VERSION == 1 | |
5235 | ||
5236 | /* For this version only U+130 and U+131 are equivalent under qr//i. Add a | |
5237 | * rule so that things work under /iaa and /il */ | |
5238 | ||
5239 | SV * mod_listsv = sv_mortalcopy(*listsvp); | |
5240 | sv_catpv(mod_listsv, "130\t130\t131\n"); | |
5241 | l = (U8*)SvPV(mod_listsv, lcur); | |
5242 | ||
5243 | #else | |
5244 | ||
064c021d | 5245 | l = (U8*)SvPV(*listsvp, lcur); |
74894415 KW |
5246 | |
5247 | #endif | |
5248 | ||
064c021d KW |
5249 | lend = l + lcur; |
5250 | ||
5251 | /* Go through each input line */ | |
5252 | while (l < lend) { | |
5253 | UV min, max, val; | |
5254 | UV inverse; | |
95543e92 KW |
5255 | l = swash_scan_list_line(l, lend, &min, &max, &val, |
5256 | cBOOL(octets), typestr); | |
064c021d KW |
5257 | if (l > lend) { |
5258 | break; | |
5259 | } | |
5260 | ||
5261 | /* Each element in the range is to be inverted */ | |
5262 | for (inverse = min; inverse <= max; inverse++) { | |
5263 | AV* list; | |
064c021d KW |
5264 | SV** listp; |
5265 | IV i; | |
5266 | bool found_key = FALSE; | |
5662e334 | 5267 | bool found_inverse = FALSE; |
064c021d KW |
5268 | |
5269 | /* The key is the inverse mapping */ | |
5270 | char key[UTF8_MAXBYTES+1]; | |
c80e42f3 | 5271 | char* key_end = (char *) uvchr_to_utf8((U8*) key, val); |
064c021d KW |
5272 | STRLEN key_len = key_end - key; |
5273 | ||
064c021d KW |
5274 | /* Get the list for the map */ |
5275 | if ((listp = hv_fetch(ret, key, key_len, FALSE))) { | |
5276 | list = (AV*) *listp; | |
5277 | } | |
5278 | else { /* No entry yet for it: create one */ | |
5279 | list = newAV(); | |
5280 | if (! hv_store(ret, key, key_len, (SV*) list, FALSE)) { | |
5281 | Perl_croak(aTHX_ "panic: hv_store() unexpectedly failed"); | |
5282 | } | |
5283 | } | |
5284 | ||
5662e334 KW |
5285 | /* Look through list to see if this inverse mapping already is |
5286 | * listed, or if there is a mapping to itself already */ | |
9506e945 | 5287 | for (i = 0; i <= av_tindex_skip_len_mg(list); i++) { |
064c021d KW |
5288 | SV** entryp = av_fetch(list, i, FALSE); |
5289 | SV* entry; | |
414db8a1 | 5290 | UV uv; |
064c021d KW |
5291 | if (entryp == NULL) { |
5292 | Perl_croak(aTHX_ "panic: av_fetch() unexpectedly failed"); | |
5293 | } | |
5294 | entry = *entryp; | |
414db8a1 | 5295 | uv = SvUV(entry); |
147e3846 | 5296 | /*DEBUG_U(PerlIO_printf(Perl_debug_log, "list for %" UVXf " contains %" UVXf "\n", val, uv));*/ |
414db8a1 | 5297 | if (uv == val) { |
064c021d | 5298 | found_key = TRUE; |
5662e334 | 5299 | } |
414db8a1 | 5300 | if (uv == inverse) { |
5662e334 KW |
5301 | found_inverse = TRUE; |
5302 | } | |
5303 | ||
5304 | /* No need to continue searching if found everything we are | |
5305 | * looking for */ | |
5306 | if (found_key && found_inverse) { | |
064c021d KW |
5307 | break; |
5308 | } | |
5309 | } | |
56ca34ca KW |
5310 | |
5311 | /* Make sure there is a mapping to itself on the list */ | |
064c021d | 5312 | if (! found_key) { |
d397ff6a | 5313 | av_push(list, newSVuv(val)); |
147e3846 | 5314 | /*DEBUG_U(PerlIO_printf(Perl_debug_log, "%s: %d: Adding %" UVXf " to list for %" UVXf "\n", __FILE__, __LINE__, val, val));*/ |
064c021d KW |
5315 | } |
5316 | ||
5317 | ||
5318 | /* Simply add the value to the list */ | |
5662e334 KW |
5319 | if (! found_inverse) { |
5320 | av_push(list, newSVuv(inverse)); | |
147e3846 | 5321 | /*DEBUG_U(PerlIO_printf(Perl_debug_log, "%s: %d: Adding %" UVXf " to list for %" UVXf "\n", __FILE__, __LINE__, inverse, val));*/ |
5662e334 | 5322 | } |
064c021d | 5323 | |
b0e3252e | 5324 | /* swatch_get() increments the value of val for each element in the |
064c021d KW |
5325 | * range. That makes more compact tables possible. You can |
5326 | * express the capitalization, for example, of all consecutive | |
5327 | * letters with a single line: 0061\t007A\t0041 This maps 0061 to | |
5328 | * 0041, 0062 to 0042, etc. I (khw) have never understood 'none', | |
bd3f2f94 | 5329 | * and it's not documented; it appears to be used only in |
b0e3252e | 5330 | * implementing tr//; I copied the semantics from swatch_get(), just |
bd3f2f94 | 5331 | * in case */ |
064c021d KW |
5332 | if (!none || val < none) { |
5333 | ++val; | |
5334 | } | |
5335 | } | |
5336 | } | |
5337 | ||
5338 | return ret; | |
5339 | } | |
5340 | ||
a25abddc | 5341 | SV* |
d764b54e KW |
5342 | Perl__swash_to_invlist(pTHX_ SV* const swash) |
5343 | { | |
5344 | ||
ed92f1b3 KW |
5345 | /* Subject to change or removal. For use only in one place in regcomp.c. |
5346 | * Ownership is given to one reference count in the returned SV* */ | |
d764b54e KW |
5347 | |
5348 | U8 *l, *lend; | |
5349 | char *loc; | |
5350 | STRLEN lcur; | |
5351 | HV *const hv = MUTABLE_HV(SvRV(swash)); | |
5352 | UV elements = 0; /* Number of elements in the inversion list */ | |
b443038a | 5353 | U8 empty[] = ""; |
923b6d4e KW |
5354 | SV** listsvp; |
5355 | SV** typesvp; | |
5356 | SV** bitssvp; | |
5357 | SV** extssvp; | |
5358 | SV** invert_it_svp; | |
d764b54e | 5359 | |
923b6d4e KW |
5360 | U8* typestr; |
5361 | STRLEN bits; | |
5362 | STRLEN octets; /* if bits == 1, then octets == 0 */ | |
d73c39c5 KW |
5363 | U8 *x, *xend; |
5364 | STRLEN xcur; | |
d764b54e | 5365 | |
a25abddc | 5366 | SV* invlist; |
d764b54e | 5367 | |
b81740c0 KW |
5368 | PERL_ARGS_ASSERT__SWASH_TO_INVLIST; |
5369 | ||
923b6d4e KW |
5370 | /* If not a hash, it must be the swash's inversion list instead */ |
5371 | if (SvTYPE(hv) != SVt_PVHV) { | |
ed92f1b3 | 5372 | return SvREFCNT_inc_simple_NN((SV*) hv); |
923b6d4e KW |
5373 | } |
5374 | ||
5375 | /* The string containing the main body of the table */ | |
5376 | listsvp = hv_fetchs(hv, "LIST", FALSE); | |
5377 | typesvp = hv_fetchs(hv, "TYPE", FALSE); | |
5378 | bitssvp = hv_fetchs(hv, "BITS", FALSE); | |
5379 | extssvp = hv_fetchs(hv, "EXTRAS", FALSE); | |
5380 | invert_it_svp = hv_fetchs(hv, "INVERT_IT", FALSE); | |
5381 | ||
5382 | typestr = (U8*)SvPV_nolen(*typesvp); | |
5383 | bits = SvUV(*bitssvp); | |
5384 | octets = bits >> 3; /* if bits == 1, then octets == 0 */ | |
5385 | ||
d764b54e | 5386 | /* read $swash->{LIST} */ |
b443038a KW |
5387 | if (SvPOK(*listsvp)) { |
5388 | l = (U8*)SvPV(*listsvp, lcur); | |
5389 | } | |
5390 | else { | |
5391 | /* LIST legitimately doesn't contain a string during compilation phases | |
5392 | * of Perl itself, before the Unicode tables are generated. In this | |
5393 | * case, just fake things up by creating an empty list */ | |
5394 | l = empty; | |
5395 | lcur = 0; | |
5396 | } | |
d764b54e KW |
5397 | loc = (char *) l; |
5398 | lend = l + lcur; | |
5399 | ||
31aa6e0b | 5400 | if (*l == 'V') { /* Inversion list format */ |
99944878 | 5401 | const char *after_atou = (char *) lend; |
31aa6e0b KW |
5402 | UV element0; |
5403 | UV* other_elements_ptr; | |
5404 | ||
5405 | /* The first number is a count of the rest */ | |
5406 | l++; | |
22ff3130 | 5407 | if (!grok_atoUV((const char *)l, &elements, &after_atou)) { |
56576a04 KW |
5408 | Perl_croak(aTHX_ "panic: Expecting a valid count of elements" |
5409 | " at start of inversion list"); | |
22ff3130 | 5410 | } |
eb092534 KW |
5411 | if (elements == 0) { |
5412 | invlist = _new_invlist(0); | |
5413 | } | |
5414 | else { | |
99944878 | 5415 | l = (U8 *) after_atou; |
1f9f7d4c | 5416 | |
56576a04 KW |
5417 | /* Get the 0th element, which is needed to setup the inversion list |
5418 | * */ | |
99944878 | 5419 | while (isSPACE(*l)) l++; |
22ff3130 | 5420 | if (!grok_atoUV((const char *)l, &element0, &after_atou)) { |
56576a04 KW |
5421 | Perl_croak(aTHX_ "panic: Expecting a valid 0th element for" |
5422 | " inversion list"); | |
22ff3130 | 5423 | } |
99944878 | 5424 | l = (U8 *) after_atou; |
56576a04 KW |
5425 | invlist = _setup_canned_invlist(elements, element0, |
5426 | &other_elements_ptr); | |
1f9f7d4c KW |
5427 | elements--; |
5428 | ||
5429 | /* Then just populate the rest of the input */ | |
5430 | while (elements-- > 0) { | |
5431 | if (l > lend) { | |
56576a04 KW |
5432 | Perl_croak(aTHX_ "panic: Expecting %" UVuf " more" |
5433 | " elements than available", elements); | |
1f9f7d4c | 5434 | } |
99944878 | 5435 | while (isSPACE(*l)) l++; |
56576a04 KW |
5436 | if (!grok_atoUV((const char *)l, other_elements_ptr++, |
5437 | &after_atou)) | |
5438 | { | |
5439 | Perl_croak(aTHX_ "panic: Expecting a valid element" | |
5440 | " in inversion list"); | |
22ff3130 | 5441 | } |
99944878 | 5442 | l = (U8 *) after_atou; |
1f9f7d4c | 5443 | } |
eb092534 | 5444 | } |
31aa6e0b KW |
5445 | } |
5446 | else { | |
5447 | ||
1784d2f9 KW |
5448 | /* Scan the input to count the number of lines to preallocate array |
5449 | * size based on worst possible case, which is each line in the input | |
5450 | * creates 2 elements in the inversion list: 1) the beginning of a | |
5451 | * range in the list; 2) the beginning of a range not in the list. */ | |
320d4f1b | 5452 | while ((loc = (char *) memchr(loc, '\n', lend - (U8 *) loc)) != NULL) { |
1784d2f9 KW |
5453 | elements += 2; |
5454 | loc++; | |
5455 | } | |
d764b54e | 5456 | |
1784d2f9 KW |
5457 | /* If the ending is somehow corrupt and isn't a new line, add another |
5458 | * element for the final range that isn't in the inversion list */ | |
5459 | if (! (*lend == '\n' | |
5460 | || (*lend == '\0' && (lcur == 0 || *(lend - 1) == '\n')))) | |
5461 | { | |
5462 | elements++; | |
5463 | } | |
d764b54e | 5464 | |
1784d2f9 | 5465 | invlist = _new_invlist(elements); |
d764b54e | 5466 | |
1784d2f9 KW |
5467 | /* Now go through the input again, adding each range to the list */ |
5468 | while (l < lend) { | |
5469 | UV start, end; | |
5470 | UV val; /* Not used by this function */ | |
d764b54e | 5471 | |
95543e92 KW |
5472 | l = swash_scan_list_line(l, lend, &start, &end, &val, |
5473 | cBOOL(octets), typestr); | |
d764b54e | 5474 | |
1784d2f9 KW |
5475 | if (l > lend) { |
5476 | break; | |
5477 | } | |
5478 | ||
5479 | invlist = _add_range_to_invlist(invlist, start, end); | |
5480 | } | |
31aa6e0b | 5481 | } |
d764b54e | 5482 | |
77f9f126 KW |
5483 | /* Invert if the data says it should be */ |
5484 | if (invert_it_svp && SvUV(*invert_it_svp)) { | |
25151030 | 5485 | _invlist_invert(invlist); |
77f9f126 KW |
5486 | } |
5487 | ||
b0e3252e | 5488 | /* This code is copied from swatch_get() |
d73c39c5 KW |
5489 | * read $swash->{EXTRAS} */ |
5490 | x = (U8*)SvPV(*extssvp, xcur); | |
5491 | xend = x + xcur; | |
5492 | while (x < xend) { | |
5493 | STRLEN namelen; | |
5494 | U8 *namestr; | |
5495 | SV** othersvp; | |
5496 | HV* otherhv; | |
5497 | STRLEN otherbits; | |
5498 | SV **otherbitssvp, *other; | |
5499 | U8 *nl; | |
5500 | ||
5501 | const U8 opc = *x++; | |
5502 | if (opc == '\n') | |
5503 | continue; | |
5504 | ||
5505 | nl = (U8*)memchr(x, '\n', xend - x); | |
5506 | ||
5507 | if (opc != '-' && opc != '+' && opc != '!' && opc != '&') { | |
5508 | if (nl) { | |
5509 | x = nl + 1; /* 1 is length of "\n" */ | |
5510 | continue; | |
5511 | } | |
5512 | else { | |
5513 | x = xend; /* to EXTRAS' end at which \n is not found */ | |
5514 | break; | |
5515 | } | |
5516 | } | |
5517 | ||
5518 | namestr = x; | |
5519 | if (nl) { | |
5520 | namelen = nl - namestr; | |
5521 | x = nl + 1; | |
5522 | } | |
5523 | else { | |
5524 | namelen = xend - namestr; | |
5525 | x = xend; | |
5526 | } | |
5527 | ||
5528 | othersvp = hv_fetch(hv, (char *)namestr, namelen, FALSE); | |
5529 | otherhv = MUTABLE_HV(SvRV(*othersvp)); | |
5530 | otherbitssvp = hv_fetchs(otherhv, "BITS", FALSE); | |
5531 | otherbits = (STRLEN)SvUV(*otherbitssvp); | |
5532 | ||
5533 | if (bits != otherbits || bits != 1) { | |
5637ef5b | 5534 | Perl_croak(aTHX_ "panic: _swash_to_invlist only operates on boolean " |
147e3846 | 5535 | "properties, bits=%" UVuf ", otherbits=%" UVuf, |
5637ef5b | 5536 | (UV)bits, (UV)otherbits); |
d73c39c5 KW |
5537 | } |
5538 | ||
5539 | /* The "other" swatch must be destroyed after. */ | |
5540 | other = _swash_to_invlist((SV *)*othersvp); | |
5541 | ||
b0e3252e | 5542 | /* End of code copied from swatch_get() */ |
d73c39c5 KW |
5543 | switch (opc) { |
5544 | case '+': | |
5545 | _invlist_union(invlist, other, &invlist); | |
5546 | break; | |
5547 | case '!': | |
6c46377d | 5548 | _invlist_union_maybe_complement_2nd(invlist, other, TRUE, &invlist); |
d73c39c5 KW |
5549 | break; |
5550 | case '-': | |
5551 | _invlist_subtract(invlist, other, &invlist); | |
5552 | break; | |
5553 | case '&': | |
5554 | _invlist_intersection(invlist, other, &invlist); | |
5555 | break; | |
5556 | default: | |
5557 | break; | |
5558 | } | |
5559 | sv_free(other); /* through with it! */ | |
5560 | } | |
5561 | ||
dbfdbd26 | 5562 | SvREADONLY_on(invlist); |
d764b54e KW |
5563 | return invlist; |
5564 | } | |
5565 | ||
3fdfee00 KW |
5566 | SV* |
5567 | Perl__get_swash_invlist(pTHX_ SV* const swash) | |
5568 | { | |
872dd7e0 | 5569 | SV** ptr; |
3fdfee00 KW |
5570 | |
5571 | PERL_ARGS_ASSERT__GET_SWASH_INVLIST; | |
5572 | ||
87367d5f | 5573 | if (! SvROK(swash)) { |
872dd7e0 KW |
5574 | return NULL; |
5575 | } | |
5576 | ||
87367d5f KW |
5577 | /* If it really isn't a hash, it isn't really swash; must be an inversion |
5578 | * list */ | |
5579 | if (SvTYPE(SvRV(swash)) != SVt_PVHV) { | |
5580 | return SvRV(swash); | |
5581 | } | |
872dd7e0 | 5582 | |
87367d5f | 5583 | ptr = hv_fetchs(MUTABLE_HV(SvRV(swash)), "V", FALSE); |
3fdfee00 KW |
5584 | if (! ptr) { |
5585 | return NULL; | |
5586 | } | |
5587 | ||
5588 | return *ptr; | |
5589 | } | |
5590 | ||
0876b9a0 | 5591 | bool |
5aaab254 | 5592 | Perl_check_utf8_print(pTHX_ const U8* s, const STRLEN len) |
0876b9a0 KW |
5593 | { |
5594 | /* May change: warns if surrogates, non-character code points, or | |
56576a04 KW |
5595 | * non-Unicode code points are in 's' which has length 'len' bytes. |
5596 | * Returns TRUE if none found; FALSE otherwise. The only other validity | |
d22ec717 KW |
5597 | * check is to make sure that this won't exceed the string's length nor |
5598 | * overflow */ | |
0876b9a0 KW |
5599 | |
5600 | const U8* const e = s + len; | |
5601 | bool ok = TRUE; | |
5602 | ||
5603 | PERL_ARGS_ASSERT_CHECK_UTF8_PRINT; | |
5604 | ||
5605 | while (s < e) { | |
5606 | if (UTF8SKIP(s) > len) { | |
5607 | Perl_ck_warner_d(aTHX_ packWARN(WARN_UTF8), | |
5608 | "%s in %s", unees, PL_op ? OP_DESC(PL_op) : "print"); | |
5609 | return FALSE; | |
5610 | } | |
ac6f1fbe | 5611 | if (UNLIKELY(isUTF8_POSSIBLY_PROBLEMATIC(*s))) { |
f2bf18cc | 5612 | if (UNLIKELY(UTF8_IS_SUPER(s, e))) { |
760c7c2f | 5613 | if ( ckWARN_d(WARN_NON_UNICODE) |
e050c007 KW |
5614 | || UNLIKELY(0 < does_utf8_overflow(s, s + len, |
5615 | 0 /* Don't consider overlongs */ | |
5616 | ))) | |
5617 | { | |
15ca5930 | 5618 | /* A side effect of this function will be to warn */ |
2db24202 | 5619 | (void) utf8n_to_uvchr(s, e - s, NULL, UTF8_WARN_SUPER); |
7ee537e6 KW |
5620 | ok = FALSE; |
5621 | } | |
0876b9a0 | 5622 | } |
f2bf18cc | 5623 | else if (UNLIKELY(UTF8_IS_SURROGATE(s, e))) { |
8457b38f | 5624 | if (ckWARN_d(WARN_SURROGATE)) { |
15ca5930 KW |
5625 | /* This has a different warning than the one the called |
5626 | * function would output, so can't just call it, unlike we | |
5627 | * do for the non-chars and above-unicodes */ | |
2db24202 | 5628 | UV uv = utf8_to_uvchr_buf(s, e, NULL); |
8457b38f | 5629 | Perl_warner(aTHX_ packWARN(WARN_SURROGATE), |
56576a04 KW |
5630 | "Unicode surrogate U+%04" UVXf " is illegal in UTF-8", |
5631 | uv); | |
8457b38f KW |
5632 | ok = FALSE; |
5633 | } | |
0876b9a0 | 5634 | } |
56576a04 KW |
5635 | else if ( UNLIKELY(UTF8_IS_NONCHAR(s, e)) |
5636 | && (ckWARN_d(WARN_NONCHAR))) | |
5637 | { | |
15ca5930 | 5638 | /* A side effect of this function will be to warn */ |
2db24202 | 5639 | (void) utf8n_to_uvchr(s, e - s, NULL, UTF8_WARN_NONCHAR); |
0876b9a0 KW |
5640 | ok = FALSE; |
5641 | } | |
5642 | } | |
5643 | s += UTF8SKIP(s); | |
5644 | } | |
5645 | ||
5646 | return ok; | |
5647 | } | |
5648 | ||
0f830e0b | 5649 | /* |
87cea99e | 5650 | =for apidoc pv_uni_display |
d2cc3551 | 5651 | |
a1433954 KW |
5652 | Build to the scalar C<dsv> a displayable version of the string C<spv>, |
5653 | length C<len>, the displayable version being at most C<pvlim> bytes long | |
796b6530 | 5654 | (if longer, the rest is truncated and C<"..."> will be appended). |
0a2ef054 | 5655 | |
796b6530 KW |
5656 | The C<flags> argument can have C<UNI_DISPLAY_ISPRINT> set to display |
5657 | C<isPRINT()>able characters as themselves, C<UNI_DISPLAY_BACKSLASH> | |
5658 | to display the C<\\[nrfta\\]> as the backslashed versions (like C<"\n">) | |
5659 | (C<UNI_DISPLAY_BACKSLASH> is preferred over C<UNI_DISPLAY_ISPRINT> for C<"\\">). | |
5660 | C<UNI_DISPLAY_QQ> (and its alias C<UNI_DISPLAY_REGEX>) have both | |
5661 | C<UNI_DISPLAY_BACKSLASH> and C<UNI_DISPLAY_ISPRINT> turned on. | |
0a2ef054 | 5662 | |
a1433954 | 5663 | The pointer to the PV of the C<dsv> is returned. |
d2cc3551 | 5664 | |
119bc988 KW |
5665 | See also L</sv_uni_display>. |
5666 | ||
d2cc3551 | 5667 | =cut */ |
e6b2e755 | 5668 | char * |
56576a04 KW |
5669 | Perl_pv_uni_display(pTHX_ SV *dsv, const U8 *spv, STRLEN len, STRLEN pvlim, |
5670 | UV flags) | |
e6b2e755 JH |
5671 | { |
5672 | int truncated = 0; | |
e1ec3a88 | 5673 | const char *s, *e; |
e6b2e755 | 5674 | |
7918f24d NC |
5675 | PERL_ARGS_ASSERT_PV_UNI_DISPLAY; |
5676 | ||
9e2aa2e7 | 5677 | SvPVCLEAR(dsv); |
7fddd944 | 5678 | SvUTF8_off(dsv); |
e1ec3a88 | 5679 | for (s = (const char *)spv, e = s + len; s < e; s += UTF8SKIP(s)) { |
e6b2e755 | 5680 | UV u; |
a49f32c6 NC |
5681 | /* This serves double duty as a flag and a character to print after |
5682 | a \ when flags & UNI_DISPLAY_BACKSLASH is true. | |
5683 | */ | |
5684 | char ok = 0; | |
c728cb41 | 5685 | |
e6b2e755 JH |
5686 | if (pvlim && SvCUR(dsv) >= pvlim) { |
5687 | truncated++; | |
5688 | break; | |
5689 | } | |
4b88fb76 | 5690 | u = utf8_to_uvchr_buf((U8*)s, (U8*)e, 0); |
c728cb41 | 5691 | if (u < 256) { |
a3b680e6 | 5692 | const unsigned char c = (unsigned char)u & 0xFF; |
0bd48802 | 5693 | if (flags & UNI_DISPLAY_BACKSLASH) { |
a49f32c6 | 5694 | switch (c) { |
c728cb41 | 5695 | case '\n': |
a49f32c6 | 5696 | ok = 'n'; break; |
c728cb41 | 5697 | case '\r': |
a49f32c6 | 5698 | ok = 'r'; break; |
c728cb41 | 5699 | case '\t': |
a49f32c6 | 5700 | ok = 't'; break; |
c728cb41 | 5701 | case '\f': |
a49f32c6 | 5702 | ok = 'f'; break; |
c728cb41 | 5703 | case '\a': |
a49f32c6 | 5704 | ok = 'a'; break; |
c728cb41 | 5705 | case '\\': |
a49f32c6 | 5706 | ok = '\\'; break; |
c728cb41 JH |
5707 | default: break; |
5708 | } | |
a49f32c6 | 5709 | if (ok) { |
88c9ea1e | 5710 | const char string = ok; |
76f68e9b | 5711 | sv_catpvs(dsv, "\\"); |
5e7aa789 | 5712 | sv_catpvn(dsv, &string, 1); |
a49f32c6 | 5713 | } |
c728cb41 | 5714 | } |
00e86452 | 5715 | /* isPRINT() is the locale-blind version. */ |
a49f32c6 | 5716 | if (!ok && (flags & UNI_DISPLAY_ISPRINT) && isPRINT(c)) { |
88c9ea1e | 5717 | const char string = c; |
5e7aa789 | 5718 | sv_catpvn(dsv, &string, 1); |
a49f32c6 | 5719 | ok = 1; |
0a2ef054 | 5720 | } |
c728cb41 JH |
5721 | } |
5722 | if (!ok) | |
147e3846 | 5723 | Perl_sv_catpvf(aTHX_ dsv, "\\x{%" UVxf "}", u); |
e6b2e755 JH |
5724 | } |
5725 | if (truncated) | |
396482e1 | 5726 | sv_catpvs(dsv, "..."); |
48ef279e | 5727 | |
e6b2e755 JH |
5728 | return SvPVX(dsv); |
5729 | } | |
2b9d42f0 | 5730 | |
d2cc3551 | 5731 | /* |
87cea99e | 5732 | =for apidoc sv_uni_display |
d2cc3551 | 5733 | |
a1433954 KW |
5734 | Build to the scalar C<dsv> a displayable version of the scalar C<sv>, |
5735 | the displayable version being at most C<pvlim> bytes long | |
d2cc3551 | 5736 | (if longer, the rest is truncated and "..." will be appended). |
0a2ef054 | 5737 | |
a1433954 | 5738 | The C<flags> argument is as in L</pv_uni_display>(). |
0a2ef054 | 5739 | |
a1433954 | 5740 | The pointer to the PV of the C<dsv> is returned. |
d2cc3551 | 5741 | |
d4c19fe8 AL |
5742 | =cut |
5743 | */ | |
e6b2e755 JH |
5744 | char * |
5745 | Perl_sv_uni_display(pTHX_ SV *dsv, SV *ssv, STRLEN pvlim, UV flags) | |
5746 | { | |
8cdde9f8 NC |
5747 | const char * const ptr = |
5748 | isREGEXP(ssv) ? RX_WRAPPED((REGEXP*)ssv) : SvPVX_const(ssv); | |
5749 | ||
7918f24d NC |
5750 | PERL_ARGS_ASSERT_SV_UNI_DISPLAY; |
5751 | ||
8cdde9f8 | 5752 | return Perl_pv_uni_display(aTHX_ dsv, (const U8*)ptr, |
cfd0369c | 5753 | SvCUR(ssv), pvlim, flags); |
701a277b JH |
5754 | } |
5755 | ||
d2cc3551 | 5756 | /* |
e6226b18 | 5757 | =for apidoc foldEQ_utf8 |
d2cc3551 | 5758 | |
56576a04 KW |
5759 | Returns true if the leading portions of the strings C<s1> and C<s2> (either or |
5760 | both of which may be in UTF-8) are the same case-insensitively; false | |
5761 | otherwise. How far into the strings to compare is determined by other input | |
5762 | parameters. | |
8b35872c | 5763 | |
a1433954 | 5764 | If C<u1> is true, the string C<s1> is assumed to be in UTF-8-encoded Unicode; |
56576a04 KW |
5765 | otherwise it is assumed to be in native 8-bit encoding. Correspondingly for |
5766 | C<u2> with respect to C<s2>. | |
5767 | ||
5768 | If the byte length C<l1> is non-zero, it says how far into C<s1> to check for | |
5769 | fold equality. In other words, C<s1>+C<l1> will be used as a goal to reach. | |
5770 | The scan will not be considered to be a match unless the goal is reached, and | |
5771 | scanning won't continue past that goal. Correspondingly for C<l2> with respect | |
5772 | to C<s2>. | |
5773 | ||
5774 | If C<pe1> is non-C<NULL> and the pointer it points to is not C<NULL>, that | |
5775 | pointer is considered an end pointer to the position 1 byte past the maximum | |
5776 | point in C<s1> beyond which scanning will not continue under any circumstances. | |
03bb5c85 | 5777 | (This routine assumes that UTF-8 encoded input strings are not malformed; |
56576a04 KW |
5778 | malformed input can cause it to read past C<pe1>). This means that if both |
5779 | C<l1> and C<pe1> are specified, and C<pe1> is less than C<s1>+C<l1>, the match | |
5780 | will never be successful because it can never | |
d51c1b21 | 5781 | get as far as its goal (and in fact is asserted against). Correspondingly for |
a1433954 | 5782 | C<pe2> with respect to C<s2>. |
8b35872c | 5783 | |
a1433954 KW |
5784 | At least one of C<s1> and C<s2> must have a goal (at least one of C<l1> and |
5785 | C<l2> must be non-zero), and if both do, both have to be | |
8b35872c KW |
5786 | reached for a successful match. Also, if the fold of a character is multiple |
5787 | characters, all of them must be matched (see tr21 reference below for | |
5788 | 'folding'). | |
5789 | ||
796b6530 | 5790 | Upon a successful match, if C<pe1> is non-C<NULL>, |
a1433954 KW |
5791 | it will be set to point to the beginning of the I<next> character of C<s1> |
5792 | beyond what was matched. Correspondingly for C<pe2> and C<s2>. | |
d2cc3551 JH |
5793 | |
5794 | For case-insensitiveness, the "casefolding" of Unicode is used | |
5795 | instead of upper/lowercasing both the characters, see | |
a1433954 | 5796 | L<http://www.unicode.org/unicode/reports/tr21/> (Case Mappings). |
d2cc3551 JH |
5797 | |
5798 | =cut */ | |
a33c29bc KW |
5799 | |
5800 | /* A flags parameter has been added which may change, and hence isn't | |
5801 | * externally documented. Currently it is: | |
5802 | * 0 for as-documented above | |
5803 | * FOLDEQ_UTF8_NOMIX_ASCII meaning that if a non-ASCII character folds to an | |
5804 | ASCII one, to not match | |
31f05a37 KW |
5805 | * FOLDEQ_LOCALE is set iff the rules from the current underlying |
5806 | * locale are to be used. | |
5807 | * FOLDEQ_S1_ALREADY_FOLDED s1 has already been folded before calling this | |
aa8ebe62 KW |
5808 | * routine. This allows that step to be skipped. |
5809 | * Currently, this requires s1 to be encoded as UTF-8 | |
5810 | * (u1 must be true), which is asserted for. | |
d635b710 KW |
5811 | * FOLDEQ_S1_FOLDS_SANE With either NOMIX_ASCII or LOCALE, no folds may |
5812 | * cross certain boundaries. Hence, the caller should | |
5813 | * let this function do the folding instead of | |
5814 | * pre-folding. This code contains an assertion to | |
5815 | * that effect. However, if the caller knows what | |
5816 | * it's doing, it can pass this flag to indicate that, | |
5817 | * and the assertion is skipped. | |
31f05a37 | 5818 | * FOLDEQ_S2_ALREADY_FOLDED Similarly. |
d635b710 | 5819 | * FOLDEQ_S2_FOLDS_SANE |
a33c29bc | 5820 | */ |
701a277b | 5821 | I32 |
56576a04 KW |
5822 | Perl_foldEQ_utf8_flags(pTHX_ const char *s1, char **pe1, UV l1, bool u1, |
5823 | const char *s2, char **pe2, UV l2, bool u2, | |
5824 | U32 flags) | |
332ddc25 | 5825 | { |
eb578fdb KW |
5826 | const U8 *p1 = (const U8*)s1; /* Point to current char */ |
5827 | const U8 *p2 = (const U8*)s2; | |
5828 | const U8 *g1 = NULL; /* goal for s1 */ | |
5829 | const U8 *g2 = NULL; | |
5830 | const U8 *e1 = NULL; /* Don't scan s1 past this */ | |
5831 | U8 *f1 = NULL; /* Point to current folded */ | |
5832 | const U8 *e2 = NULL; | |
5833 | U8 *f2 = NULL; | |
48ef279e | 5834 | STRLEN n1 = 0, n2 = 0; /* Number of bytes in current char */ |
8b35872c KW |
5835 | U8 foldbuf1[UTF8_MAXBYTES_CASE+1]; |
5836 | U8 foldbuf2[UTF8_MAXBYTES_CASE+1]; | |
1d39b2cd | 5837 | U8 flags_for_folder = FOLD_FLAGS_FULL; |
8b35872c | 5838 | |
eda9cac1 | 5839 | PERL_ARGS_ASSERT_FOLDEQ_UTF8_FLAGS; |
8b35872c | 5840 | |
cea315b6 | 5841 | assert( ! ((flags & (FOLDEQ_UTF8_NOMIX_ASCII | FOLDEQ_LOCALE)) |
d635b710 KW |
5842 | && (((flags & FOLDEQ_S1_ALREADY_FOLDED) |
5843 | && !(flags & FOLDEQ_S1_FOLDS_SANE)) | |
5844 | || ((flags & FOLDEQ_S2_ALREADY_FOLDED) | |
5845 | && !(flags & FOLDEQ_S2_FOLDS_SANE))))); | |
b08f1bd5 KW |
5846 | /* The algorithm is to trial the folds without regard to the flags on |
5847 | * the first line of the above assert(), and then see if the result | |
5848 | * violates them. This means that the inputs can't be pre-folded to a | |
5849 | * violating result, hence the assert. This could be changed, with the | |
5850 | * addition of extra tests here for the already-folded case, which would | |
5851 | * slow it down. That cost is more than any possible gain for when these | |
5852 | * flags are specified, as the flags indicate /il or /iaa matching which | |
5853 | * is less common than /iu, and I (khw) also believe that real-world /il | |
5854 | * and /iaa matches are most likely to involve code points 0-255, and this | |
5855 | * function only under rare conditions gets called for 0-255. */ | |
18f762c3 | 5856 | |
1d39b2cd KW |
5857 | if (flags & FOLDEQ_LOCALE) { |
5858 | if (IN_UTF8_CTYPE_LOCALE) { | |
5859 | flags &= ~FOLDEQ_LOCALE; | |
5860 | } | |
5861 | else { | |
5862 | flags_for_folder |= FOLD_FLAGS_LOCALE; | |
5863 | } | |
31f05a37 KW |
5864 | } |
5865 | ||
8b35872c | 5866 | if (pe1) { |
48ef279e | 5867 | e1 = *(U8**)pe1; |
8b35872c KW |
5868 | } |
5869 | ||
5870 | if (l1) { | |
48ef279e | 5871 | g1 = (const U8*)s1 + l1; |
8b35872c KW |
5872 | } |
5873 | ||
5874 | if (pe2) { | |
48ef279e | 5875 | e2 = *(U8**)pe2; |
8b35872c KW |
5876 | } |
5877 | ||
5878 | if (l2) { | |
48ef279e | 5879 | g2 = (const U8*)s2 + l2; |
8b35872c KW |
5880 | } |
5881 | ||
5882 | /* Must have at least one goal */ | |
5883 | assert(g1 || g2); | |
5884 | ||
5885 | if (g1) { | |
5886 | ||
48ef279e KW |
5887 | /* Will never match if goal is out-of-bounds */ |
5888 | assert(! e1 || e1 >= g1); | |
8b35872c | 5889 | |
48ef279e KW |
5890 | /* Here, there isn't an end pointer, or it is beyond the goal. We |
5891 | * only go as far as the goal */ | |
5892 | e1 = g1; | |
8b35872c | 5893 | } |
313b38e5 NC |
5894 | else { |
5895 | assert(e1); /* Must have an end for looking at s1 */ | |
5896 | } | |
8b35872c KW |
5897 | |
5898 | /* Same for goal for s2 */ | |
5899 | if (g2) { | |
48ef279e KW |
5900 | assert(! e2 || e2 >= g2); |
5901 | e2 = g2; | |
8b35872c | 5902 | } |
313b38e5 NC |
5903 | else { |
5904 | assert(e2); | |
5905 | } | |
8b35872c | 5906 | |
18f762c3 KW |
5907 | /* If both operands are already folded, we could just do a memEQ on the |
5908 | * whole strings at once, but it would be better if the caller realized | |
5909 | * this and didn't even call us */ | |
5910 | ||
8b35872c KW |
5911 | /* Look through both strings, a character at a time */ |
5912 | while (p1 < e1 && p2 < e2) { | |
5913 | ||
d51c1b21 | 5914 | /* If at the beginning of a new character in s1, get its fold to use |
1d39b2cd | 5915 | * and the length of the fold. */ |
48ef279e | 5916 | if (n1 == 0) { |
18f762c3 KW |
5917 | if (flags & FOLDEQ_S1_ALREADY_FOLDED) { |
5918 | f1 = (U8 *) p1; | |
aa8ebe62 | 5919 | assert(u1); |
18f762c3 | 5920 | n1 = UTF8SKIP(f1); |
18f762c3 KW |
5921 | } |
5922 | else { | |
1d39b2cd KW |
5923 | if (isASCII(*p1) && ! (flags & FOLDEQ_LOCALE)) { |
5924 | ||
5925 | /* We have to forbid mixing ASCII with non-ASCII if the | |
5926 | * flags so indicate. And, we can short circuit having to | |
5927 | * call the general functions for this common ASCII case, | |
5928 | * all of whose non-locale folds are also ASCII, and hence | |
5929 | * UTF-8 invariants, so the UTF8ness of the strings is not | |
5930 | * relevant. */ | |
5931 | if ((flags & FOLDEQ_UTF8_NOMIX_ASCII) && ! isASCII(*p2)) { | |
5932 | return 0; | |
5933 | } | |
5934 | n1 = 1; | |
5935 | *foldbuf1 = toFOLD(*p1); | |
5936 | } | |
5937 | else if (u1) { | |
a1a5ec35 | 5938 | _toFOLD_utf8_flags(p1, e1, foldbuf1, &n1, flags_for_folder); |
1d39b2cd | 5939 | } |
4a4088c4 | 5940 | else { /* Not UTF-8, get UTF-8 fold */ |
1d39b2cd KW |
5941 | _to_uni_fold_flags(*p1, foldbuf1, &n1, flags_for_folder); |
5942 | } | |
5943 | f1 = foldbuf1; | |
5944 | } | |
48ef279e | 5945 | } |
8b35872c | 5946 | |
48ef279e | 5947 | if (n2 == 0) { /* Same for s2 */ |
18f762c3 KW |
5948 | if (flags & FOLDEQ_S2_ALREADY_FOLDED) { |
5949 | f2 = (U8 *) p2; | |
aa8ebe62 | 5950 | assert(u2); |
18f762c3 KW |
5951 | n2 = UTF8SKIP(f2); |
5952 | } | |
5953 | else { | |
1d39b2cd KW |
5954 | if (isASCII(*p2) && ! (flags & FOLDEQ_LOCALE)) { |
5955 | if ((flags & FOLDEQ_UTF8_NOMIX_ASCII) && ! isASCII(*p1)) { | |
5956 | return 0; | |
5957 | } | |
5958 | n2 = 1; | |
5959 | *foldbuf2 = toFOLD(*p2); | |
5960 | } | |
5961 | else if (u2) { | |
a1a5ec35 | 5962 | _toFOLD_utf8_flags(p2, e2, foldbuf2, &n2, flags_for_folder); |
1d39b2cd KW |
5963 | } |
5964 | else { | |
5965 | _to_uni_fold_flags(*p2, foldbuf2, &n2, flags_for_folder); | |
5966 | } | |
5967 | f2 = foldbuf2; | |
18f762c3 | 5968 | } |
48ef279e | 5969 | } |
8b35872c | 5970 | |
5001101e | 5971 | /* Here f1 and f2 point to the beginning of the strings to compare. |
227968da | 5972 | * These strings are the folds of the next character from each input |
4a4088c4 | 5973 | * string, stored in UTF-8. */ |
5e64d0fa | 5974 | |
48ef279e KW |
5975 | /* While there is more to look for in both folds, see if they |
5976 | * continue to match */ | |
5977 | while (n1 && n2) { | |
5978 | U8 fold_length = UTF8SKIP(f1); | |
5979 | if (fold_length != UTF8SKIP(f2) | |
5980 | || (fold_length == 1 && *f1 != *f2) /* Short circuit memNE | |
5981 | function call for single | |
a6d5f321 | 5982 | byte */ |
48ef279e KW |
5983 | || memNE((char*)f1, (char*)f2, fold_length)) |
5984 | { | |
e6226b18 | 5985 | return 0; /* mismatch */ |
48ef279e KW |
5986 | } |
5987 | ||
5988 | /* Here, they matched, advance past them */ | |
5989 | n1 -= fold_length; | |
5990 | f1 += fold_length; | |
5991 | n2 -= fold_length; | |
5992 | f2 += fold_length; | |
5993 | } | |
8b35872c | 5994 | |
48ef279e KW |
5995 | /* When reach the end of any fold, advance the input past it */ |
5996 | if (n1 == 0) { | |
5997 | p1 += u1 ? UTF8SKIP(p1) : 1; | |
5998 | } | |
5999 | if (n2 == 0) { | |
6000 | p2 += u2 ? UTF8SKIP(p2) : 1; | |
6001 | } | |
8b35872c KW |
6002 | } /* End of loop through both strings */ |
6003 | ||
6004 | /* A match is defined by each scan that specified an explicit length | |
6005 | * reaching its final goal, and the other not having matched a partial | |
6006 | * character (which can happen when the fold of a character is more than one | |
6007 | * character). */ | |
6008 | if (! ((g1 == 0 || p1 == g1) && (g2 == 0 || p2 == g2)) || n1 || n2) { | |
e6226b18 | 6009 | return 0; |
8b35872c KW |
6010 | } |
6011 | ||
6012 | /* Successful match. Set output pointers */ | |
6013 | if (pe1) { | |
48ef279e | 6014 | *pe1 = (char*)p1; |
8b35872c KW |
6015 | } |
6016 | if (pe2) { | |
48ef279e | 6017 | *pe2 = (char*)p2; |
8b35872c | 6018 | } |
e6226b18 | 6019 | return 1; |
e6b2e755 | 6020 | } |
701a277b | 6021 | |
f2645549 | 6022 | /* XXX The next two functions should likely be moved to mathoms.c once all |
37e7596b KW |
6023 | * occurrences of them are removed from the core; some cpan-upstream modules |
6024 | * still use them */ | |
6025 | ||
6026 | U8 * | |
6027 | Perl_uvuni_to_utf8(pTHX_ U8 *d, UV uv) | |
6028 | { | |
6029 | PERL_ARGS_ASSERT_UVUNI_TO_UTF8; | |
6030 | ||
33f38593 | 6031 | return uvoffuni_to_utf8_flags(d, uv, 0); |
37e7596b KW |
6032 | } |
6033 | ||
e505af10 KW |
6034 | /* |
6035 | =for apidoc utf8n_to_uvuni | |
6036 | ||
6037 | Instead use L</utf8_to_uvchr_buf>, or rarely, L</utf8n_to_uvchr>. | |
6038 | ||
6039 | This function was useful for code that wanted to handle both EBCDIC and | |
6040 | ASCII platforms with Unicode properties, but starting in Perl v5.20, the | |
6041 | distinctions between the platforms have mostly been made invisible to most | |
6042 | code, so this function is quite unlikely to be what you want. If you do need | |
6043 | this precise functionality, use instead | |
6044 | C<L<NATIVE_TO_UNI(utf8_to_uvchr_buf(...))|/utf8_to_uvchr_buf>> | |
6045 | or C<L<NATIVE_TO_UNI(utf8n_to_uvchr(...))|/utf8n_to_uvchr>>. | |
6046 | ||
6047 | =cut | |
6048 | */ | |
6049 | ||
37e7596b KW |
6050 | UV |
6051 | Perl_utf8n_to_uvuni(pTHX_ const U8 *s, STRLEN curlen, STRLEN *retlen, U32 flags) | |
6052 | { | |
6053 | PERL_ARGS_ASSERT_UTF8N_TO_UVUNI; | |
6054 | ||
6055 | return NATIVE_TO_UNI(utf8n_to_uvchr(s, curlen, retlen, flags)); | |
6056 | } | |
6057 | ||
6058 | /* | |
6059 | =for apidoc uvuni_to_utf8_flags | |
6060 | ||
6061 | Instead you almost certainly want to use L</uvchr_to_utf8> or | |
efa9cd84 | 6062 | L</uvchr_to_utf8_flags>. |
37e7596b KW |
6063 | |
6064 | This function is a deprecated synonym for L</uvoffuni_to_utf8_flags>, | |
6065 | which itself, while not deprecated, should be used only in isolated | |
6066 | circumstances. These functions were useful for code that wanted to handle | |
6067 | both EBCDIC and ASCII platforms with Unicode properties, but starting in Perl | |
6068 | v5.20, the distinctions between the platforms have mostly been made invisible | |
6069 | to most code, so this function is quite unlikely to be what you want. | |
6070 | ||
6071 | =cut | |
6072 | */ | |
6073 | ||
6074 | U8 * | |
6075 | Perl_uvuni_to_utf8_flags(pTHX_ U8 *d, UV uv, UV flags) | |
6076 | { | |
6077 | PERL_ARGS_ASSERT_UVUNI_TO_UTF8_FLAGS; | |
6078 | ||
6079 | return uvoffuni_to_utf8_flags(d, uv, flags); | |
6080 | } | |
6081 | ||
6082 | /* | |
14d04a33 | 6083 | * ex: set ts=8 sts=4 sw=4 et: |
37442d52 | 6084 | */ |