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 | |
27da23d5 JH |
36 | static const char unees[] = |
37 | "Malformed UTF-8 character (unexpected end of string)"; | |
760c7c2f | 38 | static const char cp_above_legal_max[] = |
2d212e86 | 39 | "Use of code point 0x%"UVXf" is deprecated; the permissible max is 0x%"UVXf""; |
760c7c2f | 40 | |
114d9c4d | 41 | #define MAX_NON_DEPRECATED_CP ((UV) (IV_MAX)) |
901b21bf | 42 | |
48ef279e | 43 | /* |
ccfc67b7 | 44 | =head1 Unicode Support |
7fefc6c1 | 45 | These are various utility functions for manipulating UTF8-encoded |
72d33970 | 46 | strings. For the uninitiated, this is a method of representing arbitrary |
61296642 | 47 | Unicode characters as a variable number of bytes, in such a way that |
56da48f7 DM |
48 | characters in the ASCII range are unmodified, and a zero byte never appears |
49 | within non-zero characters. | |
166f8a29 | 50 | |
eaf7a4d2 CS |
51 | =cut |
52 | */ | |
53 | ||
54 | /* | |
7bbfa158 | 55 | =for apidoc is_invariant_string |
eaf7a4d2 | 56 | |
b8ed1ca9 KW |
57 | Returns true iff the first C<len> bytes of the string C<s> are the same |
58 | regardless of the UTF-8 encoding of the string (or UTF-EBCDIC encoding on | |
59 | EBCDIC machines). That is, if they are UTF-8 invariant. On ASCII-ish | |
60 | machines, all the ASCII characters and only the ASCII characters fit this | |
7bbfa158 | 61 | definition. On EBCDIC machines, the ASCII-range characters are invariant, but |
b8ed1ca9 KW |
62 | so also are the C1 controls and C<\c?> (which isn't in the ASCII range on |
63 | EBCDIC). | |
eaf7a4d2 | 64 | |
75200dff KW |
65 | If C<len> is 0, it will be calculated using C<strlen(s)>, (which means if you |
66 | use this option, that C<s> can't have embedded C<NUL> characters and has to | |
67 | have a terminating C<NUL> byte). | |
9f7e3d64 | 68 | |
a1433954 | 69 | See also L</is_utf8_string>(), L</is_utf8_string_loclen>(), and L</is_utf8_string_loc>(). |
eaf7a4d2 CS |
70 | |
71 | =cut | |
72 | */ | |
73 | ||
74 | bool | |
7bbfa158 | 75 | Perl_is_invariant_string(const U8 *s, STRLEN len) |
eaf7a4d2 CS |
76 | { |
77 | const U8* const send = s + (len ? len : strlen((const char *)s)); | |
78 | const U8* x = s; | |
79 | ||
7bbfa158 | 80 | PERL_ARGS_ASSERT_IS_INVARIANT_STRING; |
eaf7a4d2 CS |
81 | |
82 | for (; x < send; ++x) { | |
83 | if (!UTF8_IS_INVARIANT(*x)) | |
84 | break; | |
85 | } | |
86 | ||
87 | return x == send; | |
88 | } | |
89 | ||
90 | /* | |
378516de | 91 | =for apidoc uvoffuni_to_utf8_flags |
eebe1485 | 92 | |
a27992cc | 93 | THIS FUNCTION SHOULD BE USED IN ONLY VERY SPECIALIZED CIRCUMSTANCES. |
de69f3af KW |
94 | Instead, B<Almost all code should use L</uvchr_to_utf8> or |
95 | L</uvchr_to_utf8_flags>>. | |
a27992cc | 96 | |
de69f3af KW |
97 | This function is like them, but the input is a strict Unicode |
98 | (as opposed to native) code point. Only in very rare circumstances should code | |
99 | not be using the native code point. | |
949cf498 | 100 | |
efa9cd84 | 101 | For details, see the description for L</uvchr_to_utf8_flags>. |
949cf498 | 102 | |
eebe1485 SC |
103 | =cut |
104 | */ | |
105 | ||
8ee1cdcb KW |
106 | #define HANDLE_UNICODE_SURROGATE(uv, flags) \ |
107 | STMT_START { \ | |
108 | if (flags & UNICODE_WARN_SURROGATE) { \ | |
109 | Perl_ck_warner_d(aTHX_ packWARN(WARN_SURROGATE), \ | |
110 | "UTF-16 surrogate U+%04"UVXf, uv); \ | |
111 | } \ | |
112 | if (flags & UNICODE_DISALLOW_SURROGATE) { \ | |
113 | return NULL; \ | |
114 | } \ | |
115 | } STMT_END; | |
116 | ||
117 | #define HANDLE_UNICODE_NONCHAR(uv, flags) \ | |
118 | STMT_START { \ | |
119 | if (flags & UNICODE_WARN_NONCHAR) { \ | |
120 | Perl_ck_warner_d(aTHX_ packWARN(WARN_NONCHAR), \ | |
121 | "Unicode non-character U+%04"UVXf" is not " \ | |
122 | "recommended for open interchange", uv); \ | |
123 | } \ | |
124 | if (flags & UNICODE_DISALLOW_NONCHAR) { \ | |
125 | return NULL; \ | |
126 | } \ | |
127 | } STMT_END; | |
128 | ||
ba6ed43c KW |
129 | /* Use shorter names internally in this file */ |
130 | #define SHIFT UTF_ACCUMULATION_SHIFT | |
131 | #undef MARK | |
132 | #define MARK UTF_CONTINUATION_MARK | |
133 | #define MASK UTF_CONTINUATION_MASK | |
134 | ||
dfe13c55 | 135 | U8 * |
378516de | 136 | Perl_uvoffuni_to_utf8_flags(pTHX_ U8 *d, UV uv, UV flags) |
a0ed51b3 | 137 | { |
378516de | 138 | PERL_ARGS_ASSERT_UVOFFUNI_TO_UTF8_FLAGS; |
7918f24d | 139 | |
2d1545e5 | 140 | if (OFFUNI_IS_INVARIANT(uv)) { |
4c8cd605 | 141 | *d++ = LATIN1_TO_NATIVE(uv); |
d9432125 KW |
142 | return d; |
143 | } | |
facc1dc2 | 144 | |
3ea68d71 | 145 | if (uv <= MAX_UTF8_TWO_BYTE) { |
facc1dc2 KW |
146 | *d++ = I8_TO_NATIVE_UTF8(( uv >> SHIFT) | UTF_START_MARK(2)); |
147 | *d++ = I8_TO_NATIVE_UTF8(( uv & MASK) | MARK); | |
3ea68d71 KW |
148 | return d; |
149 | } | |
d9432125 | 150 | |
ba6ed43c KW |
151 | /* Not 2-byte; test for and handle 3-byte result. In the test immediately |
152 | * below, the 16 is for start bytes E0-EF (which are all the possible ones | |
153 | * for 3 byte characters). The 2 is for 2 continuation bytes; these each | |
154 | * contribute SHIFT bits. This yields 0x4000 on EBCDIC platforms, 0x1_0000 | |
155 | * on ASCII; so 3 bytes covers the range 0x400-0x3FFF on EBCDIC; | |
156 | * 0x800-0xFFFF on ASCII */ | |
157 | if (uv < (16 * (1U << (2 * SHIFT)))) { | |
158 | *d++ = I8_TO_NATIVE_UTF8(( uv >> ((3 - 1) * SHIFT)) | UTF_START_MARK(3)); | |
159 | *d++ = I8_TO_NATIVE_UTF8(((uv >> ((2 - 1) * SHIFT)) & MASK) | MARK); | |
160 | *d++ = I8_TO_NATIVE_UTF8(( uv /* (1 - 1) */ & MASK) | MARK); | |
161 | ||
162 | #ifndef EBCDIC /* These problematic code points are 4 bytes on EBCDIC, so | |
163 | aren't tested here */ | |
164 | /* The most likely code points in this range are below the surrogates. | |
165 | * Do an extra test to quickly exclude those. */ | |
166 | if (UNLIKELY(uv >= UNICODE_SURROGATE_FIRST)) { | |
167 | if (UNLIKELY( UNICODE_IS_32_CONTIGUOUS_NONCHARS(uv) | |
168 | || UNICODE_IS_END_PLANE_NONCHAR_GIVEN_NOT_SUPER(uv))) | |
169 | { | |
8ee1cdcb KW |
170 | HANDLE_UNICODE_NONCHAR(uv, flags); |
171 | } | |
172 | else if (UNLIKELY(UNICODE_IS_SURROGATE(uv))) { | |
173 | HANDLE_UNICODE_SURROGATE(uv, flags); | |
760c7c2f | 174 | } |
ba6ed43c KW |
175 | } |
176 | #endif | |
177 | return d; | |
178 | } | |
179 | ||
180 | /* Not 3-byte; that means the code point is at least 0x1_0000 on ASCII | |
181 | * platforms, and 0x4000 on EBCDIC. There are problematic cases that can | |
182 | * happen starting with 4-byte characters on ASCII platforms. We unify the | |
183 | * code for these with EBCDIC, even though some of them require 5-bytes on | |
184 | * those, because khw believes the code saving is worth the very slight | |
185 | * performance hit on these high EBCDIC code points. */ | |
186 | ||
187 | if (UNLIKELY(UNICODE_IS_SUPER(uv))) { | |
a5bf80e0 KW |
188 | if ( UNLIKELY(uv > MAX_NON_DEPRECATED_CP) |
189 | && ckWARN_d(WARN_DEPRECATED)) | |
190 | { | |
191 | Perl_warner(aTHX_ packWARN(WARN_DEPRECATED), | |
192 | cp_above_legal_max, uv, MAX_NON_DEPRECATED_CP); | |
193 | } | |
194 | if ( (flags & UNICODE_WARN_SUPER) | |
195 | || ( UNICODE_IS_ABOVE_31_BIT(uv) | |
196 | && (flags & UNICODE_WARN_ABOVE_31_BIT))) | |
197 | { | |
198 | Perl_ck_warner_d(aTHX_ packWARN(WARN_NON_UNICODE), | |
199 | ||
200 | /* Choose the more dire applicable warning */ | |
201 | (UNICODE_IS_ABOVE_31_BIT(uv)) | |
202 | ? "Code point 0x%"UVXf" is not Unicode, and not portable" | |
203 | : "Code point 0x%"UVXf" is not Unicode, may not be portable", | |
204 | uv); | |
205 | } | |
206 | if (flags & UNICODE_DISALLOW_SUPER | |
207 | || ( UNICODE_IS_ABOVE_31_BIT(uv) | |
208 | && (flags & UNICODE_DISALLOW_ABOVE_31_BIT))) | |
209 | { | |
210 | return NULL; | |
211 | } | |
212 | } | |
ba6ed43c KW |
213 | else if (UNLIKELY(UNICODE_IS_END_PLANE_NONCHAR_GIVEN_NOT_SUPER(uv))) { |
214 | HANDLE_UNICODE_NONCHAR(uv, flags); | |
507b9800 | 215 | } |
d9432125 | 216 | |
ba6ed43c KW |
217 | /* Test for and handle 4-byte result. In the test immediately below, the |
218 | * 8 is for start bytes F0-F7 (which are all the possible ones for 4 byte | |
219 | * characters). The 3 is for 3 continuation bytes; these each contribute | |
220 | * SHIFT bits. This yields 0x4_0000 on EBCDIC platforms, 0x20_0000 on | |
221 | * ASCII, so 4 bytes covers the range 0x4000-0x3_FFFF on EBCDIC; | |
222 | * 0x1_0000-0x1F_FFFF on ASCII */ | |
223 | if (uv < (8 * (1U << (3 * SHIFT)))) { | |
224 | *d++ = I8_TO_NATIVE_UTF8(( uv >> ((4 - 1) * SHIFT)) | UTF_START_MARK(4)); | |
225 | *d++ = I8_TO_NATIVE_UTF8(((uv >> ((3 - 1) * SHIFT)) & MASK) | MARK); | |
226 | *d++ = I8_TO_NATIVE_UTF8(((uv >> ((2 - 1) * SHIFT)) & MASK) | MARK); | |
227 | *d++ = I8_TO_NATIVE_UTF8(( uv /* (1 - 1) */ & MASK) | MARK); | |
228 | ||
229 | #ifdef EBCDIC /* These were handled on ASCII platforms in the code for 3-byte | |
230 | characters. The end-plane non-characters for EBCDIC were | |
231 | handled just above */ | |
232 | if (UNLIKELY(UNICODE_IS_32_CONTIGUOUS_NONCHARS(uv))) { | |
233 | HANDLE_UNICODE_NONCHAR(uv, flags); | |
d528804a | 234 | } |
ba6ed43c KW |
235 | else if (UNLIKELY(UNICODE_IS_SURROGATE(uv))) { |
236 | HANDLE_UNICODE_SURROGATE(uv, flags); | |
237 | } | |
238 | #endif | |
239 | ||
240 | return d; | |
241 | } | |
242 | ||
243 | /* Not 4-byte; that means the code point is at least 0x20_0000 on ASCII | |
244 | * platforms, and 0x4000 on EBCDIC. At this point we switch to a loop | |
245 | * format. The unrolled version above turns out to not save all that much | |
246 | * time, and at these high code points (well above the legal Unicode range | |
247 | * on ASCII platforms, and well above anything in common use in EBCDIC), | |
248 | * khw believes that less code outweighs slight performance gains. */ | |
249 | ||
d9432125 | 250 | { |
5aaebcb3 | 251 | STRLEN len = OFFUNISKIP(uv); |
1d72bdf6 NIS |
252 | U8 *p = d+len-1; |
253 | while (p > d) { | |
4c8cd605 | 254 | *p-- = I8_TO_NATIVE_UTF8((uv & UTF_CONTINUATION_MASK) | UTF_CONTINUATION_MARK); |
1d72bdf6 NIS |
255 | uv >>= UTF_ACCUMULATION_SHIFT; |
256 | } | |
4c8cd605 | 257 | *p = I8_TO_NATIVE_UTF8((uv & UTF_START_MASK(len)) | UTF_START_MARK(len)); |
1d72bdf6 NIS |
258 | return d+len; |
259 | } | |
a0ed51b3 | 260 | } |
a5bf80e0 | 261 | |
646ca15d | 262 | /* |
07693fe6 KW |
263 | =for apidoc uvchr_to_utf8 |
264 | ||
bcb1a2d4 | 265 | Adds the UTF-8 representation of the native code point C<uv> to the end |
f2fc1b45 | 266 | of the string C<d>; C<d> should have at least C<UVCHR_SKIP(uv)+1> (up to |
c749c9fd KW |
267 | C<UTF8_MAXBYTES+1>) free bytes available. The return value is the pointer to |
268 | the byte after the end of the new character. In other words, | |
07693fe6 KW |
269 | |
270 | d = uvchr_to_utf8(d, uv); | |
271 | ||
272 | is the recommended wide native character-aware way of saying | |
273 | ||
274 | *(d++) = uv; | |
275 | ||
760c7c2f KW |
276 | This function accepts any UV as input, but very high code points (above |
277 | C<IV_MAX> on the platform) will raise a deprecation warning. This is | |
278 | typically 0x7FFF_FFFF in a 32-bit word. | |
279 | ||
280 | It is possible to forbid or warn on non-Unicode code points, or those that may | |
281 | be problematic by using L</uvchr_to_utf8_flags>. | |
de69f3af | 282 | |
07693fe6 KW |
283 | =cut |
284 | */ | |
285 | ||
de69f3af KW |
286 | /* This is also a macro */ |
287 | PERL_CALLCONV U8* Perl_uvchr_to_utf8(pTHX_ U8 *d, UV uv); | |
288 | ||
07693fe6 KW |
289 | U8 * |
290 | Perl_uvchr_to_utf8(pTHX_ U8 *d, UV uv) | |
291 | { | |
de69f3af | 292 | return uvchr_to_utf8(d, uv); |
07693fe6 KW |
293 | } |
294 | ||
de69f3af KW |
295 | /* |
296 | =for apidoc uvchr_to_utf8_flags | |
297 | ||
298 | Adds the UTF-8 representation of the native code point C<uv> to the end | |
f2fc1b45 | 299 | of the string C<d>; C<d> should have at least C<UVCHR_SKIP(uv)+1> (up to |
c749c9fd KW |
300 | C<UTF8_MAXBYTES+1>) free bytes available. The return value is the pointer to |
301 | the byte after the end of the new character. In other words, | |
de69f3af KW |
302 | |
303 | d = uvchr_to_utf8_flags(d, uv, flags); | |
304 | ||
305 | or, in most cases, | |
306 | ||
307 | d = uvchr_to_utf8_flags(d, uv, 0); | |
308 | ||
309 | This is the Unicode-aware way of saying | |
310 | ||
311 | *(d++) = uv; | |
312 | ||
760c7c2f KW |
313 | If C<flags> is 0, this function accepts any UV as input, but very high code |
314 | points (above C<IV_MAX> for the platform) will raise a deprecation warning. | |
315 | This is typically 0x7FFF_FFFF in a 32-bit word. | |
316 | ||
317 | Specifying C<flags> can further restrict what is allowed and not warned on, as | |
318 | follows: | |
de69f3af | 319 | |
796b6530 | 320 | If C<uv> is a Unicode surrogate code point and C<UNICODE_WARN_SURROGATE> is set, |
7ee537e6 KW |
321 | the function will raise a warning, provided UTF8 warnings are enabled. If |
322 | instead C<UNICODE_DISALLOW_SURROGATE> is set, the function will fail and return | |
323 | NULL. If both flags are set, the function will both warn and return NULL. | |
de69f3af | 324 | |
760c7c2f KW |
325 | Similarly, the C<UNICODE_WARN_NONCHAR> and C<UNICODE_DISALLOW_NONCHAR> flags |
326 | affect how the function handles a Unicode non-character. | |
93e6dbd6 | 327 | |
760c7c2f KW |
328 | And likewise, the C<UNICODE_WARN_SUPER> and C<UNICODE_DISALLOW_SUPER> flags |
329 | affect the handling of code points that are above the Unicode maximum of | |
330 | 0x10FFFF. Languages other than Perl may not be able to accept files that | |
331 | contain these. | |
93e6dbd6 KW |
332 | |
333 | The flag C<UNICODE_WARN_ILLEGAL_INTERCHANGE> selects all three of | |
334 | the above WARN flags; and C<UNICODE_DISALLOW_ILLEGAL_INTERCHANGE> selects all | |
335 | three DISALLOW flags. | |
336 | ||
ab8e6d41 KW |
337 | Code points above 0x7FFF_FFFF (2**31 - 1) were never specified in any standard, |
338 | so using them is more problematic than other above-Unicode code points. Perl | |
339 | invented an extension to UTF-8 to represent the ones above 2**36-1, so it is | |
340 | likely that non-Perl languages will not be able to read files that contain | |
341 | these that written by the perl interpreter; nor would Perl understand files | |
342 | written by something that uses a different extension. For these reasons, there | |
343 | is a separate set of flags that can warn and/or disallow these extremely high | |
344 | code points, even if other above-Unicode ones are accepted. These are the | |
760c7c2f KW |
345 | C<UNICODE_WARN_ABOVE_31_BIT> and C<UNICODE_DISALLOW_ABOVE_31_BIT> flags. These |
346 | are entirely independent from the deprecation warning for code points above | |
347 | C<IV_MAX>. On 32-bit machines, it will eventually be forbidden to have any | |
348 | code point that needs more than 31 bits to represent. When that happens, | |
349 | effectively the C<UNICODE_DISALLOW_ABOVE_31_BIT> flag will always be set on | |
350 | 32-bit machines. (Of course C<UNICODE_DISALLOW_SUPER> will treat all | |
ab8e6d41 KW |
351 | above-Unicode code points, including these, as malformations; and |
352 | C<UNICODE_WARN_SUPER> warns on these.) | |
353 | ||
354 | On EBCDIC platforms starting in Perl v5.24, the Perl extension for representing | |
355 | extremely high code points kicks in at 0x3FFF_FFFF (2**30 -1), which is lower | |
356 | than on ASCII. Prior to that, code points 2**31 and higher were simply | |
357 | unrepresentable, and a different, incompatible method was used to represent | |
358 | code points between 2**30 and 2**31 - 1. The flags C<UNICODE_WARN_ABOVE_31_BIT> | |
359 | and C<UNICODE_DISALLOW_ABOVE_31_BIT> have the same function as on ASCII | |
360 | platforms, warning and disallowing 2**31 and higher. | |
de69f3af | 361 | |
de69f3af KW |
362 | =cut |
363 | */ | |
364 | ||
365 | /* This is also a macro */ | |
366 | PERL_CALLCONV U8* Perl_uvchr_to_utf8_flags(pTHX_ U8 *d, UV uv, UV flags); | |
367 | ||
07693fe6 KW |
368 | U8 * |
369 | Perl_uvchr_to_utf8_flags(pTHX_ U8 *d, UV uv, UV flags) | |
370 | { | |
de69f3af | 371 | return uvchr_to_utf8_flags(d, uv, flags); |
07693fe6 KW |
372 | } |
373 | ||
374 | /* | |
87cea99e | 375 | =for apidoc is_utf8_string |
6662521e | 376 | |
a1433954 | 377 | Returns true if the first C<len> bytes of string C<s> form a valid |
9f7e3d64 | 378 | UTF-8 string, false otherwise. If C<len> is 0, it will be calculated |
75200dff KW |
379 | using C<strlen(s)> (which means if you use this option, that C<s> can't have |
380 | embedded C<NUL> characters and has to have a terminating C<NUL> byte). Note | |
381 | that all characters being ASCII constitute 'a valid UTF-8 string'. | |
6662521e | 382 | |
7bbfa158 | 383 | See also L</is_invariant_string>(), L</is_utf8_string_loclen>(), and L</is_utf8_string_loc>(). |
768c67ee | 384 | |
6662521e GS |
385 | =cut |
386 | */ | |
387 | ||
8e84507e | 388 | bool |
668b6d8d | 389 | Perl_is_utf8_string(const U8 *s, STRLEN len) |
6662521e | 390 | { |
35da51f7 | 391 | const U8* const send = s + (len ? len : strlen((const char *)s)); |
7fc63493 | 392 | const U8* x = s; |
067a85ef | 393 | |
7918f24d | 394 | PERL_ARGS_ASSERT_IS_UTF8_STRING; |
1aa99e6b | 395 | |
6662521e | 396 | while (x < send) { |
6302f837 KW |
397 | STRLEN len = isUTF8_CHAR(x, send); |
398 | if (UNLIKELY(! len)) { | |
399 | return FALSE; | |
400 | } | |
401 | x += len; | |
6662521e | 402 | } |
768c67ee | 403 | |
067a85ef | 404 | return TRUE; |
6662521e GS |
405 | } |
406 | ||
67e989fb | 407 | /* |
814fafa7 NC |
408 | Implemented as a macro in utf8.h |
409 | ||
87cea99e | 410 | =for apidoc is_utf8_string_loc |
814fafa7 | 411 | |
a1433954 KW |
412 | Like L</is_utf8_string> but stores the location of the failure (in the |
413 | case of "utf8ness failure") or the location C<s>+C<len> (in the case of | |
814fafa7 NC |
414 | "utf8ness success") in the C<ep>. |
415 | ||
a1433954 | 416 | See also L</is_utf8_string_loclen>() and L</is_utf8_string>(). |
814fafa7 | 417 | |
87cea99e | 418 | =for apidoc is_utf8_string_loclen |
81cd54e3 | 419 | |
a1433954 KW |
420 | Like L</is_utf8_string>() but stores the location of the failure (in the |
421 | case of "utf8ness failure") or the location C<s>+C<len> (in the case of | |
768c67ee JH |
422 | "utf8ness success") in the C<ep>, and the number of UTF-8 |
423 | encoded characters in the C<el>. | |
424 | ||
a1433954 | 425 | See also L</is_utf8_string_loc>() and L</is_utf8_string>(). |
81cd54e3 JH |
426 | |
427 | =cut | |
428 | */ | |
429 | ||
430 | bool | |
668b6d8d | 431 | Perl_is_utf8_string_loclen(const U8 *s, STRLEN len, const U8 **ep, STRLEN *el) |
81cd54e3 | 432 | { |
35da51f7 | 433 | const U8* const send = s + (len ? len : strlen((const char *)s)); |
7fc63493 | 434 | const U8* x = s; |
3ebfea28 | 435 | STRLEN outlen = 0; |
7918f24d NC |
436 | |
437 | PERL_ARGS_ASSERT_IS_UTF8_STRING_LOCLEN; | |
81cd54e3 | 438 | |
81cd54e3 | 439 | while (x < send) { |
6302f837 KW |
440 | STRLEN len = isUTF8_CHAR(x, send); |
441 | if (UNLIKELY(! len)) { | |
442 | goto out; | |
443 | } | |
444 | x += len; | |
445 | outlen++; | |
81cd54e3 | 446 | } |
768c67ee JH |
447 | |
448 | out: | |
3ebfea28 AL |
449 | if (el) |
450 | *el = outlen; | |
451 | ||
768c67ee JH |
452 | if (ep) |
453 | *ep = x; | |
3ebfea28 | 454 | return (x == send); |
81cd54e3 JH |
455 | } |
456 | ||
457 | /* | |
768c67ee | 458 | |
de69f3af | 459 | =for apidoc utf8n_to_uvchr |
378516de KW |
460 | |
461 | THIS FUNCTION SHOULD BE USED IN ONLY VERY SPECIALIZED CIRCUMSTANCES. | |
de69f3af | 462 | Most code should use L</utf8_to_uvchr_buf>() rather than call this directly. |
67e989fb | 463 | |
9041c2e3 | 464 | Bottom level UTF-8 decode routine. |
de69f3af | 465 | Returns the native code point value of the first character in the string C<s>, |
746afd53 KW |
466 | which is assumed to be in UTF-8 (or UTF-EBCDIC) encoding, and no longer than |
467 | C<curlen> bytes; C<*retlen> (if C<retlen> isn't NULL) will be set to | |
468 | the length, in bytes, of that character. | |
949cf498 KW |
469 | |
470 | The value of C<flags> determines the behavior when C<s> does not point to a | |
471 | well-formed UTF-8 character. If C<flags> is 0, when a malformation is found, | |
524080c4 KW |
472 | zero is returned and C<*retlen> is set so that (S<C<s> + C<*retlen>>) is the |
473 | next possible position in C<s> that could begin a non-malformed character. | |
474 | Also, if UTF-8 warnings haven't been lexically disabled, a warning is raised. | |
949cf498 KW |
475 | |
476 | Various ALLOW flags can be set in C<flags> to allow (and not warn on) | |
477 | individual types of malformations, such as the sequence being overlong (that | |
478 | is, when there is a shorter sequence that can express the same code point; | |
479 | overlong sequences are expressly forbidden in the UTF-8 standard due to | |
480 | potential security issues). Another malformation example is the first byte of | |
481 | a character not being a legal first byte. See F<utf8.h> for the list of such | |
524080c4 KW |
482 | flags. For allowed 0 length strings, this function returns 0; for allowed |
483 | overlong sequences, the computed code point is returned; for all other allowed | |
484 | malformations, the Unicode REPLACEMENT CHARACTER is returned, as these have no | |
485 | determinable reasonable value. | |
949cf498 | 486 | |
796b6530 | 487 | The C<UTF8_CHECK_ONLY> flag overrides the behavior when a non-allowed (by other |
949cf498 KW |
488 | flags) malformation is found. If this flag is set, the routine assumes that |
489 | the caller will raise a warning, and this function will silently just set | |
d088425d KW |
490 | C<retlen> to C<-1> (cast to C<STRLEN>) and return zero. |
491 | ||
75200dff | 492 | Note that this API requires disambiguation between successful decoding a C<NUL> |
796b6530 | 493 | character, and an error return (unless the C<UTF8_CHECK_ONLY> flag is set), as |
d088425d | 494 | in both cases, 0 is returned. To disambiguate, upon a zero return, see if the |
75200dff KW |
495 | first byte of C<s> is 0 as well. If so, the input was a C<NUL>; if not, the |
496 | input had an error. | |
949cf498 KW |
497 | |
498 | Certain code points are considered problematic. These are Unicode surrogates, | |
746afd53 | 499 | Unicode non-characters, and code points above the Unicode maximum of 0x10FFFF. |
949cf498 | 500 | By default these are considered regular code points, but certain situations |
5eafe189 | 501 | warrant special handling for them. If C<flags> contains |
796b6530 KW |
502 | C<UTF8_DISALLOW_ILLEGAL_INTERCHANGE>, all three classes are treated as |
503 | malformations and handled as such. The flags C<UTF8_DISALLOW_SURROGATE>, | |
504 | C<UTF8_DISALLOW_NONCHAR>, and C<UTF8_DISALLOW_SUPER> (meaning above the legal | |
505 | Unicode maximum) can be set to disallow these categories individually. | |
506 | ||
507 | The flags C<UTF8_WARN_ILLEGAL_INTERCHANGE>, C<UTF8_WARN_SURROGATE>, | |
508 | C<UTF8_WARN_NONCHAR>, and C<UTF8_WARN_SUPER> will cause warning messages to be | |
509 | raised for their respective categories, but otherwise the code points are | |
510 | considered valid (not malformations). To get a category to both be treated as | |
511 | a malformation and raise a warning, specify both the WARN and DISALLOW flags. | |
949cf498 | 512 | (But note that warnings are not raised if lexically disabled nor if |
796b6530 | 513 | C<UTF8_CHECK_ONLY> is also specified.) |
949cf498 | 514 | |
760c7c2f KW |
515 | It is now deprecated to have very high code points (above C<IV_MAX> on the |
516 | platforms) and this function will raise a deprecation warning for these (unless | |
517 | such warnings are turned off). This value, is typically 0x7FFF_FFFF (2**31 -1) | |
518 | in a 32-bit word. | |
ab8e6d41 KW |
519 | |
520 | Code points above 0x7FFF_FFFF (2**31 - 1) were never specified in any standard, | |
521 | so using them is more problematic than other above-Unicode code points. Perl | |
522 | invented an extension to UTF-8 to represent the ones above 2**36-1, so it is | |
523 | likely that non-Perl languages will not be able to read files that contain | |
524 | these that written by the perl interpreter; nor would Perl understand files | |
525 | written by something that uses a different extension. For these reasons, there | |
526 | is a separate set of flags that can warn and/or disallow these extremely high | |
527 | code points, even if other above-Unicode ones are accepted. These are the | |
760c7c2f KW |
528 | C<UTF8_WARN_ABOVE_31_BIT> and C<UTF8_DISALLOW_ABOVE_31_BIT> flags. These |
529 | are entirely independent from the deprecation warning for code points above | |
530 | C<IV_MAX>. On 32-bit machines, it will eventually be forbidden to have any | |
531 | code point that needs more than 31 bits to represent. When that happens, | |
532 | effectively the C<UTF8_DISALLOW_ABOVE_31_BIT> flag will always be set on | |
533 | 32-bit machines. (Of course C<UTF8_DISALLOW_SUPER> will treat all | |
ab8e6d41 KW |
534 | above-Unicode code points, including these, as malformations; and |
535 | C<UTF8_WARN_SUPER> warns on these.) | |
536 | ||
537 | On EBCDIC platforms starting in Perl v5.24, the Perl extension for representing | |
538 | extremely high code points kicks in at 0x3FFF_FFFF (2**30 -1), which is lower | |
539 | than on ASCII. Prior to that, code points 2**31 and higher were simply | |
540 | unrepresentable, and a different, incompatible method was used to represent | |
541 | code points between 2**30 and 2**31 - 1. The flags C<UTF8_WARN_ABOVE_31_BIT> | |
542 | and C<UTF8_DISALLOW_ABOVE_31_BIT> have the same function as on ASCII | |
543 | platforms, warning and disallowing 2**31 and higher. | |
949cf498 KW |
544 | |
545 | All other code points corresponding to Unicode characters, including private | |
546 | use and those yet to be assigned, are never considered malformed and never | |
547 | warn. | |
67e989fb | 548 | |
37607a96 PK |
549 | =cut |
550 | */ | |
67e989fb | 551 | |
a0ed51b3 | 552 | UV |
de69f3af | 553 | Perl_utf8n_to_uvchr(pTHX_ const U8 *s, STRLEN curlen, STRLEN *retlen, U32 flags) |
a0ed51b3 | 554 | { |
d4c19fe8 | 555 | const U8 * const s0 = s; |
eb83ed87 | 556 | U8 overflow_byte = '\0'; /* Save byte in case of overflow */ |
0b8d30e8 | 557 | U8 * send; |
eb83ed87 KW |
558 | UV uv = *s; |
559 | STRLEN expectlen; | |
949cf498 | 560 | SV* sv = NULL; |
eb83ed87 KW |
561 | UV outlier_ret = 0; /* return value when input is in error or problematic |
562 | */ | |
563 | UV pack_warn = 0; /* Save result of packWARN() for later */ | |
564 | bool unexpected_non_continuation = FALSE; | |
565 | bool overflowed = FALSE; | |
2f8f112e | 566 | bool do_overlong_test = TRUE; /* May have to skip this test */ |
a0dbb045 | 567 | |
eb83ed87 | 568 | const char* const malformed_text = "Malformed UTF-8 character"; |
7918f24d | 569 | |
de69f3af | 570 | PERL_ARGS_ASSERT_UTF8N_TO_UVCHR; |
a0dbb045 | 571 | |
eb83ed87 KW |
572 | /* The order of malformation tests here is important. We should consume as |
573 | * few bytes as possible in order to not skip any valid character. This is | |
574 | * required by the Unicode Standard (section 3.9 of Unicode 6.0); see also | |
575 | * http://unicode.org/reports/tr36 for more discussion as to why. For | |
576 | * example, once we've done a UTF8SKIP, we can tell the expected number of | |
577 | * bytes, and could fail right off the bat if the input parameters indicate | |
578 | * that there are too few available. But it could be that just that first | |
579 | * byte is garbled, and the intended character occupies fewer bytes. If we | |
580 | * blindly assumed that the first byte is correct, and skipped based on | |
581 | * that number, we could skip over a valid input character. So instead, we | |
582 | * always examine the sequence byte-by-byte. | |
583 | * | |
584 | * We also should not consume too few bytes, otherwise someone could inject | |
585 | * things. For example, an input could be deliberately designed to | |
586 | * overflow, and if this code bailed out immediately upon discovering that, | |
e2660c54 | 587 | * returning to the caller C<*retlen> pointing to the very next byte (one |
eb83ed87 KW |
588 | * which is actually part of of the overflowing sequence), that could look |
589 | * legitimate to the caller, which could discard the initial partial | |
590 | * sequence and process the rest, inappropriately */ | |
591 | ||
592 | /* Zero length strings, if allowed, of necessity are zero */ | |
b5b9af04 | 593 | if (UNLIKELY(curlen == 0)) { |
eb83ed87 KW |
594 | if (retlen) { |
595 | *retlen = 0; | |
596 | } | |
a0dbb045 | 597 | |
eb83ed87 KW |
598 | if (flags & UTF8_ALLOW_EMPTY) { |
599 | return 0; | |
600 | } | |
601 | if (! (flags & UTF8_CHECK_ONLY)) { | |
602 | sv = sv_2mortal(Perl_newSVpvf(aTHX_ "%s (empty string)", malformed_text)); | |
603 | } | |
0c443dc2 JH |
604 | goto malformed; |
605 | } | |
606 | ||
eb83ed87 KW |
607 | expectlen = UTF8SKIP(s); |
608 | ||
609 | /* A well-formed UTF-8 character, as the vast majority of calls to this | |
610 | * function will be for, has this expected length. For efficiency, set | |
611 | * things up here to return it. It will be overriden only in those rare | |
612 | * cases where a malformation is found */ | |
613 | if (retlen) { | |
614 | *retlen = expectlen; | |
615 | } | |
616 | ||
617 | /* An invariant is trivially well-formed */ | |
1d72bdf6 | 618 | if (UTF8_IS_INVARIANT(uv)) { |
de69f3af | 619 | return uv; |
a0ed51b3 | 620 | } |
67e989fb | 621 | |
eb83ed87 | 622 | /* A continuation character can't start a valid sequence */ |
b5b9af04 | 623 | if (UNLIKELY(UTF8_IS_CONTINUATION(uv))) { |
eb83ed87 KW |
624 | if (flags & UTF8_ALLOW_CONTINUATION) { |
625 | if (retlen) { | |
626 | *retlen = 1; | |
627 | } | |
628 | return UNICODE_REPLACEMENT; | |
629 | } | |
ba210ebe | 630 | |
eb83ed87 KW |
631 | if (! (flags & UTF8_CHECK_ONLY)) { |
632 | sv = sv_2mortal(Perl_newSVpvf(aTHX_ "%s (unexpected continuation byte 0x%02x, with no preceding start byte)", malformed_text, *s0)); | |
633 | } | |
634 | curlen = 1; | |
ba210ebe JH |
635 | goto malformed; |
636 | } | |
9041c2e3 | 637 | |
dcd27b3c KW |
638 | /* Here is not a continuation byte, nor an invariant. The only thing left |
639 | * is a start byte (possibly for an overlong) */ | |
640 | ||
534752c1 KW |
641 | /* Convert to I8 on EBCDIC (no-op on ASCII), then remove the leading bits |
642 | * that indicate the number of bytes in the character's whole UTF-8 | |
643 | * sequence, leaving just the bits that are part of the value. */ | |
644 | uv = NATIVE_UTF8_TO_I8(uv) & UTF_START_MASK(expectlen); | |
ba210ebe | 645 | |
eb83ed87 KW |
646 | /* Now, loop through the remaining bytes in the character's sequence, |
647 | * accumulating each into the working value as we go. Be sure to not look | |
648 | * past the end of the input string */ | |
0b8d30e8 KW |
649 | send = (U8*) s0 + ((expectlen <= curlen) ? expectlen : curlen); |
650 | ||
eb83ed87 | 651 | for (s = s0 + 1; s < send; s++) { |
b5b9af04 | 652 | if (LIKELY(UTF8_IS_CONTINUATION(*s))) { |
eb83ed87 KW |
653 | if (uv & UTF_ACCUMULATION_OVERFLOW_MASK) { |
654 | ||
655 | /* The original implementors viewed this malformation as more | |
656 | * serious than the others (though I, khw, don't understand | |
657 | * why, since other malformations also give very very wrong | |
658 | * results), so there is no way to turn off checking for it. | |
659 | * Set a flag, but keep going in the loop, so that we absorb | |
660 | * the rest of the bytes that comprise the character. */ | |
661 | overflowed = TRUE; | |
662 | overflow_byte = *s; /* Save for warning message's use */ | |
663 | } | |
8850bf83 | 664 | uv = UTF8_ACCUMULATE(uv, *s); |
eb83ed87 KW |
665 | } |
666 | else { | |
667 | /* Here, found a non-continuation before processing all expected | |
668 | * bytes. This byte begins a new character, so quit, even if | |
669 | * allowing this malformation. */ | |
670 | unexpected_non_continuation = TRUE; | |
671 | break; | |
672 | } | |
673 | } /* End of loop through the character's bytes */ | |
674 | ||
675 | /* Save how many bytes were actually in the character */ | |
676 | curlen = s - s0; | |
677 | ||
678 | /* The loop above finds two types of malformations: non-continuation and/or | |
679 | * overflow. The non-continuation malformation is really a too-short | |
680 | * malformation, as it means that the current character ended before it was | |
681 | * expected to (being terminated prematurely by the beginning of the next | |
682 | * character, whereas in the too-short malformation there just are too few | |
683 | * bytes available to hold the character. In both cases, the check below | |
684 | * that we have found the expected number of bytes would fail if executed.) | |
685 | * Thus the non-continuation malformation is really unnecessary, being a | |
686 | * subset of the too-short malformation. But there may be existing | |
687 | * applications that are expecting the non-continuation type, so we retain | |
688 | * it, and return it in preference to the too-short malformation. (If this | |
689 | * code were being written from scratch, the two types might be collapsed | |
690 | * into one.) I, khw, am also giving priority to returning the | |
691 | * non-continuation and too-short malformations over overflow when multiple | |
692 | * ones are present. I don't know of any real reason to prefer one over | |
693 | * the other, except that it seems to me that multiple-byte errors trumps | |
694 | * errors from a single byte */ | |
b5b9af04 | 695 | if (UNLIKELY(unexpected_non_continuation)) { |
eb83ed87 KW |
696 | if (!(flags & UTF8_ALLOW_NON_CONTINUATION)) { |
697 | if (! (flags & UTF8_CHECK_ONLY)) { | |
698 | if (curlen == 1) { | |
699 | sv = sv_2mortal(Perl_newSVpvf(aTHX_ "%s (unexpected non-continuation byte 0x%02x, immediately after start byte 0x%02x)", malformed_text, *s, *s0)); | |
700 | } | |
701 | else { | |
702 | sv = sv_2mortal(Perl_newSVpvf(aTHX_ "%s (unexpected non-continuation byte 0x%02x, %d bytes after start byte 0x%02x, expected %d bytes)", malformed_text, *s, (int) curlen, *s0, (int)expectlen)); | |
a0dbb045 JH |
703 | } |
704 | } | |
eb83ed87 KW |
705 | goto malformed; |
706 | } | |
707 | uv = UNICODE_REPLACEMENT; | |
2f8f112e KW |
708 | |
709 | /* Skip testing for overlongs, as the REPLACEMENT may not be the same | |
710 | * as what the original expectations were. */ | |
711 | do_overlong_test = FALSE; | |
eb83ed87 KW |
712 | if (retlen) { |
713 | *retlen = curlen; | |
714 | } | |
715 | } | |
b5b9af04 | 716 | else if (UNLIKELY(curlen < expectlen)) { |
eb83ed87 KW |
717 | if (! (flags & UTF8_ALLOW_SHORT)) { |
718 | if (! (flags & UTF8_CHECK_ONLY)) { | |
719 | sv = sv_2mortal(Perl_newSVpvf(aTHX_ "%s (%d byte%s, need %d, after start byte 0x%02x)", malformed_text, (int)curlen, curlen == 1 ? "" : "s", (int)expectlen, *s0)); | |
a0dbb045 | 720 | } |
eb83ed87 KW |
721 | goto malformed; |
722 | } | |
723 | uv = UNICODE_REPLACEMENT; | |
2f8f112e | 724 | do_overlong_test = FALSE; |
eb83ed87 KW |
725 | if (retlen) { |
726 | *retlen = curlen; | |
727 | } | |
728 | } | |
729 | ||
b5b9af04 | 730 | if (UNLIKELY(overflowed)) { |
eb83ed87 | 731 | sv = sv_2mortal(Perl_newSVpvf(aTHX_ "%s (overflow at byte 0x%02x, after start byte 0x%02x)", malformed_text, overflow_byte, *s0)); |
ba210ebe | 732 | goto malformed; |
eb83ed87 | 733 | } |
eb83ed87 | 734 | |
2f8f112e | 735 | if (do_overlong_test |
5aaebcb3 | 736 | && expectlen > (STRLEN) OFFUNISKIP(uv) |
2f8f112e KW |
737 | && ! (flags & UTF8_ALLOW_LONG)) |
738 | { | |
eb83ed87 KW |
739 | /* The overlong malformation has lower precedence than the others. |
740 | * Note that if this malformation is allowed, we return the actual | |
741 | * value, instead of the replacement character. This is because this | |
742 | * value is actually well-defined. */ | |
743 | if (! (flags & UTF8_CHECK_ONLY)) { | |
5aaebcb3 | 744 | sv = sv_2mortal(Perl_newSVpvf(aTHX_ "%s (%d byte%s, need %d, after start byte 0x%02x)", malformed_text, (int)expectlen, expectlen == 1 ? "": "s", OFFUNISKIP(uv), *s0)); |
eb83ed87 KW |
745 | } |
746 | goto malformed; | |
747 | } | |
748 | ||
1a89bb6c | 749 | /* Here, the input is considered to be well-formed, but it still could be a |
eb83ed87 KW |
750 | * problematic code point that is not allowed by the input parameters. */ |
751 | if (uv >= UNICODE_SURROGATE_FIRST /* isn't problematic if < this */ | |
760c7c2f KW |
752 | && ((flags & ( UTF8_DISALLOW_NONCHAR |
753 | |UTF8_DISALLOW_SURROGATE | |
754 | |UTF8_DISALLOW_SUPER | |
755 | |UTF8_DISALLOW_ABOVE_31_BIT | |
756 | |UTF8_WARN_NONCHAR | |
757 | |UTF8_WARN_SURROGATE | |
758 | |UTF8_WARN_SUPER | |
759 | |UTF8_WARN_ABOVE_31_BIT)) | |
760 | || ( UNLIKELY(uv > MAX_NON_DEPRECATED_CP) | |
761 | && ckWARN_d(WARN_DEPRECATED)))) | |
eb83ed87 | 762 | { |
949cf498 | 763 | if (UNICODE_IS_SURROGATE(uv)) { |
ea5ced44 KW |
764 | |
765 | /* By adding UTF8_CHECK_ONLY to the test, we avoid unnecessary | |
766 | * generation of the sv, since no warnings are raised under CHECK */ | |
eb83ed87 | 767 | if ((flags & (UTF8_WARN_SURROGATE|UTF8_CHECK_ONLY)) == UTF8_WARN_SURROGATE |
54f4afef | 768 | && ckWARN_d(WARN_SURROGATE)) |
eb83ed87 | 769 | { |
111d382d | 770 | sv = sv_2mortal(Perl_newSVpvf(aTHX_ "UTF-16 surrogate U+%04"UVXf"", uv)); |
54f4afef | 771 | pack_warn = packWARN(WARN_SURROGATE); |
949cf498 KW |
772 | } |
773 | if (flags & UTF8_DISALLOW_SURROGATE) { | |
774 | goto disallowed; | |
775 | } | |
776 | } | |
949cf498 | 777 | else if ((uv > PERL_UNICODE_MAX)) { |
eb83ed87 | 778 | if ((flags & (UTF8_WARN_SUPER|UTF8_CHECK_ONLY)) == UTF8_WARN_SUPER |
ea5ced44 | 779 | && ckWARN_d(WARN_NON_UNICODE)) |
eb83ed87 | 780 | { |
0bcdd8f6 KW |
781 | sv = sv_2mortal(Perl_newSVpvf(aTHX_ |
782 | "Code point 0x%04"UVXf" is not Unicode, may not be portable", | |
783 | uv)); | |
54f4afef | 784 | pack_warn = packWARN(WARN_NON_UNICODE); |
949cf498 | 785 | } |
c0236afe KW |
786 | |
787 | /* The maximum code point ever specified by a standard was | |
788 | * 2**31 - 1. Anything larger than that is a Perl extension that | |
789 | * very well may not be understood by other applications (including | |
790 | * earlier perl versions on EBCDIC platforms). On ASCII platforms, | |
791 | * these code points are indicated by the first UTF-8 byte being | |
d35f2ca5 KW |
792 | * 0xFE or 0xFF. We test for these after the regular SUPER ones, |
793 | * and before possibly bailing out, so that the slightly more dire | |
c0236afe KW |
794 | * warning will override the regular one. */ |
795 | if ( | |
796 | #ifndef EBCDIC | |
797 | (*s0 & 0xFE) == 0xFE /* matches both FE, FF */ | |
798 | #else | |
799 | /* The I8 for 2**31 (U+80000000) is | |
800 | * \xFF\xA0\xA0\xA0\xA0\xA0\xA0\xA2\xA0\xA0\xA0\xA0\xA0\xA0 | |
801 | * and it turns out that on all EBCDIC pages recognized that | |
802 | * the UTF-EBCDIC for that code point is | |
803 | * \xFE\x41\x41\x41\x41\x41\x41\x43\x41\x41\x41\x41\x41\x41 | |
804 | * For the next lower code point, the 1047 UTF-EBCDIC is | |
805 | * \xFE\x41\x41\x41\x41\x41\x41\x42\x73\x73\x73\x73\x73\x73 | |
806 | * The other code pages differ only in the bytes following | |
807 | * \x42. Thus the following works (the minimum continuation | |
808 | * byte is \x41). */ | |
809 | *s0 == 0xFE && send - s0 > 7 && ( s0[1] > 0x41 | |
810 | || s0[2] > 0x41 | |
811 | || s0[3] > 0x41 | |
812 | || s0[4] > 0x41 | |
813 | || s0[5] > 0x41 | |
814 | || s0[6] > 0x41 | |
815 | || s0[7] > 0x42) | |
816 | #endif | |
d35f2ca5 KW |
817 | && (flags & (UTF8_WARN_ABOVE_31_BIT|UTF8_WARN_SUPER |
818 | |UTF8_DISALLOW_ABOVE_31_BIT))) | |
ea5ced44 | 819 | { |
0bcdd8f6 | 820 | if ( ! (flags & UTF8_CHECK_ONLY) |
d35f2ca5 | 821 | && (flags & (UTF8_WARN_ABOVE_31_BIT|UTF8_WARN_SUPER)) |
0bcdd8f6 | 822 | && ckWARN_d(WARN_UTF8)) |
ea5ced44 | 823 | { |
0bcdd8f6 KW |
824 | sv = sv_2mortal(Perl_newSVpvf(aTHX_ |
825 | "Code point 0x%"UVXf" is not Unicode, and not portable", | |
826 | uv)); | |
ea5ced44 KW |
827 | pack_warn = packWARN(WARN_UTF8); |
828 | } | |
d35f2ca5 | 829 | if (flags & UTF8_DISALLOW_ABOVE_31_BIT) { |
ea5ced44 KW |
830 | goto disallowed; |
831 | } | |
832 | } | |
c0236afe | 833 | |
949cf498 KW |
834 | if (flags & UTF8_DISALLOW_SUPER) { |
835 | goto disallowed; | |
836 | } | |
760c7c2f KW |
837 | |
838 | /* The deprecated warning overrides any non-deprecated one */ | |
839 | if (UNLIKELY(uv > MAX_NON_DEPRECATED_CP) && ckWARN_d(WARN_DEPRECATED)) | |
840 | { | |
841 | sv = sv_2mortal(Perl_newSVpvf(aTHX_ cp_above_legal_max, | |
842 | uv, MAX_NON_DEPRECATED_CP)); | |
843 | pack_warn = packWARN(WARN_DEPRECATED); | |
844 | } | |
949cf498 | 845 | } |
4190d317 KW |
846 | else if (UNICODE_IS_NONCHAR(uv)) { |
847 | if ((flags & (UTF8_WARN_NONCHAR|UTF8_CHECK_ONLY)) == UTF8_WARN_NONCHAR | |
54f4afef | 848 | && ckWARN_d(WARN_NONCHAR)) |
4190d317 | 849 | { |
ba707cdc | 850 | sv = sv_2mortal(Perl_newSVpvf(aTHX_ "Unicode non-character U+%04"UVXf" is not recommended for open interchange", uv)); |
54f4afef | 851 | pack_warn = packWARN(WARN_NONCHAR); |
4190d317 KW |
852 | } |
853 | if (flags & UTF8_DISALLOW_NONCHAR) { | |
854 | goto disallowed; | |
855 | } | |
856 | } | |
949cf498 | 857 | |
eb83ed87 | 858 | if (sv) { |
de69f3af KW |
859 | outlier_ret = uv; /* Note we don't bother to convert to native, |
860 | as all the outlier code points are the same | |
861 | in both ASCII and EBCDIC */ | |
eb83ed87 KW |
862 | goto do_warn; |
863 | } | |
864 | ||
949cf498 KW |
865 | /* Here, this is not considered a malformed character, so drop through |
866 | * to return it */ | |
a0ed51b3 | 867 | } |
ba210ebe | 868 | |
de69f3af | 869 | return UNI_TO_NATIVE(uv); |
ba210ebe | 870 | |
eb83ed87 KW |
871 | /* There are three cases which get to beyond this point. In all 3 cases: |
872 | * <sv> if not null points to a string to print as a warning. | |
873 | * <curlen> is what <*retlen> should be set to if UTF8_CHECK_ONLY isn't | |
874 | * set. | |
875 | * <outlier_ret> is what return value to use if UTF8_CHECK_ONLY isn't set. | |
876 | * This is done by initializing it to 0, and changing it only | |
877 | * for case 1). | |
878 | * The 3 cases are: | |
879 | * 1) The input is valid but problematic, and to be warned about. The | |
880 | * return value is the resultant code point; <*retlen> is set to | |
881 | * <curlen>, the number of bytes that comprise the code point. | |
882 | * <pack_warn> contains the result of packWARN() for the warning | |
883 | * types. The entry point for this case is the label <do_warn>; | |
884 | * 2) The input is a valid code point but disallowed by the parameters to | |
885 | * this function. The return value is 0. If UTF8_CHECK_ONLY is set, | |
886 | * <*relen> is -1; otherwise it is <curlen>, the number of bytes that | |
887 | * comprise the code point. <pack_warn> contains the result of | |
888 | * packWARN() for the warning types. The entry point for this case is | |
889 | * the label <disallowed>. | |
890 | * 3) The input is malformed. The return value is 0. If UTF8_CHECK_ONLY | |
891 | * is set, <*relen> is -1; otherwise it is <curlen>, the number of | |
892 | * bytes that comprise the malformation. All such malformations are | |
893 | * assumed to be warning type <utf8>. The entry point for this case | |
894 | * is the label <malformed>. | |
895 | */ | |
949cf498 | 896 | |
7b52d656 | 897 | malformed: |
ba210ebe | 898 | |
eb83ed87 KW |
899 | if (sv && ckWARN_d(WARN_UTF8)) { |
900 | pack_warn = packWARN(WARN_UTF8); | |
901 | } | |
902 | ||
7b52d656 | 903 | disallowed: |
eb83ed87 | 904 | |
fcc8fcf6 | 905 | if (flags & UTF8_CHECK_ONLY) { |
ba210ebe | 906 | if (retlen) |
10edeb5d | 907 | *retlen = ((STRLEN) -1); |
ba210ebe JH |
908 | return 0; |
909 | } | |
910 | ||
7b52d656 | 911 | do_warn: |
5b311467 | 912 | |
eb83ed87 KW |
913 | if (pack_warn) { /* <pack_warn> was initialized to 0, and changed only |
914 | if warnings are to be raised. */ | |
f555bc63 | 915 | const char * const string = SvPVX_const(sv); |
a0dbb045 | 916 | |
f555bc63 KW |
917 | if (PL_op) |
918 | Perl_warner(aTHX_ pack_warn, "%s in %s", string, OP_DESC(PL_op)); | |
919 | else | |
920 | Perl_warner(aTHX_ pack_warn, "%s", string); | |
a0dbb045 JH |
921 | } |
922 | ||
eb83ed87 KW |
923 | if (retlen) { |
924 | *retlen = curlen; | |
925 | } | |
ba210ebe | 926 | |
eb83ed87 | 927 | return outlier_ret; |
a0ed51b3 LW |
928 | } |
929 | ||
8e84507e | 930 | /* |
ec5f19d0 KW |
931 | =for apidoc utf8_to_uvchr_buf |
932 | ||
933 | Returns the native code point of the first character in the string C<s> which | |
934 | is assumed to be in UTF-8 encoding; C<send> points to 1 beyond the end of C<s>. | |
524080c4 | 935 | C<*retlen> will be set to the length, in bytes, of that character. |
ec5f19d0 | 936 | |
524080c4 KW |
937 | If C<s> does not point to a well-formed UTF-8 character and UTF8 warnings are |
938 | enabled, zero is returned and C<*retlen> is set (if C<retlen> isn't | |
796b6530 | 939 | C<NULL>) to -1. If those warnings are off, the computed value, if well-defined |
173db420 | 940 | (or the Unicode REPLACEMENT CHARACTER if not), is silently returned, and |
796b6530 | 941 | C<*retlen> is set (if C<retlen> isn't C<NULL>) so that (S<C<s> + C<*retlen>>) is |
173db420 | 942 | the next possible position in C<s> that could begin a non-malformed character. |
de69f3af | 943 | See L</utf8n_to_uvchr> for details on when the REPLACEMENT CHARACTER is |
173db420 | 944 | returned. |
ec5f19d0 | 945 | |
760c7c2f KW |
946 | Code points above the platform's C<IV_MAX> will raise a deprecation warning, |
947 | unless those are turned off. | |
948 | ||
ec5f19d0 KW |
949 | =cut |
950 | */ | |
951 | ||
952 | ||
953 | UV | |
954 | Perl_utf8_to_uvchr_buf(pTHX_ const U8 *s, const U8 *send, STRLEN *retlen) | |
955 | { | |
ec5f19d0 KW |
956 | assert(s < send); |
957 | ||
958 | return utf8n_to_uvchr(s, send - s, retlen, | |
959 | ckWARN_d(WARN_UTF8) ? 0 : UTF8_ALLOW_ANY); | |
960 | } | |
961 | ||
27d6c58a | 962 | /* Like L</utf8_to_uvchr_buf>(), but should only be called when it is known that |
3986bb7c | 963 | * there are no malformations in the input UTF-8 string C<s>. surrogates, |
57b0056d | 964 | * non-character code points, and non-Unicode code points are allowed. */ |
27d6c58a KW |
965 | |
966 | UV | |
967 | Perl_valid_utf8_to_uvchr(pTHX_ const U8 *s, STRLEN *retlen) | |
968 | { | |
010ab96b KW |
969 | UV expectlen = UTF8SKIP(s); |
970 | const U8* send = s + expectlen; | |
9ff2f0f7 | 971 | UV uv = *s; |
3986bb7c | 972 | |
27d6c58a | 973 | PERL_ARGS_ASSERT_VALID_UTF8_TO_UVCHR; |
81611534 | 974 | PERL_UNUSED_CONTEXT; |
27d6c58a | 975 | |
010ab96b KW |
976 | if (retlen) { |
977 | *retlen = expectlen; | |
978 | } | |
979 | ||
980 | /* An invariant is trivially returned */ | |
981 | if (expectlen == 1) { | |
9ff2f0f7 | 982 | return uv; |
010ab96b KW |
983 | } |
984 | ||
9ff2f0f7 KW |
985 | #ifdef EBCDIC |
986 | uv = NATIVE_UTF8_TO_I8(uv); | |
987 | #endif | |
988 | ||
010ab96b KW |
989 | /* Remove the leading bits that indicate the number of bytes, leaving just |
990 | * the bits that are part of the value */ | |
991 | uv &= UTF_START_MASK(expectlen); | |
992 | ||
993 | /* Now, loop through the remaining bytes, accumulating each into the | |
994 | * working total as we go. (I khw tried unrolling the loop for up to 4 | |
995 | * bytes, but there was no performance improvement) */ | |
996 | for (++s; s < send; s++) { | |
997 | uv = UTF8_ACCUMULATE(uv, *s); | |
998 | } | |
999 | ||
3986bb7c | 1000 | return UNI_TO_NATIVE(uv); |
010ab96b | 1001 | |
27d6c58a KW |
1002 | } |
1003 | ||
ec5f19d0 | 1004 | /* |
ec5f19d0 KW |
1005 | =for apidoc utf8_to_uvuni_buf |
1006 | ||
de69f3af KW |
1007 | Only in very rare circumstances should code need to be dealing in Unicode |
1008 | (as opposed to native) code points. In those few cases, use | |
1009 | C<L<NATIVE_TO_UNI(utf8_to_uvchr_buf(...))|/utf8_to_uvchr_buf>> instead. | |
4f83cdcd KW |
1010 | |
1011 | Returns the Unicode (not-native) code point of the first character in the | |
1012 | string C<s> which | |
ec5f19d0 KW |
1013 | is assumed to be in UTF-8 encoding; C<send> points to 1 beyond the end of C<s>. |
1014 | C<retlen> will be set to the length, in bytes, of that character. | |
1015 | ||
524080c4 KW |
1016 | If C<s> does not point to a well-formed UTF-8 character and UTF8 warnings are |
1017 | enabled, zero is returned and C<*retlen> is set (if C<retlen> isn't | |
1018 | NULL) to -1. If those warnings are off, the computed value if well-defined (or | |
1019 | the Unicode REPLACEMENT CHARACTER, if not) is silently returned, and C<*retlen> | |
1020 | is set (if C<retlen> isn't NULL) so that (S<C<s> + C<*retlen>>) is the | |
1021 | next possible position in C<s> that could begin a non-malformed character. | |
de69f3af | 1022 | See L</utf8n_to_uvchr> for details on when the REPLACEMENT CHARACTER is returned. |
ec5f19d0 | 1023 | |
760c7c2f KW |
1024 | Code points above the platform's C<IV_MAX> will raise a deprecation warning, |
1025 | unless those are turned off. | |
1026 | ||
ec5f19d0 KW |
1027 | =cut |
1028 | */ | |
1029 | ||
1030 | UV | |
1031 | Perl_utf8_to_uvuni_buf(pTHX_ const U8 *s, const U8 *send, STRLEN *retlen) | |
1032 | { | |
1033 | PERL_ARGS_ASSERT_UTF8_TO_UVUNI_BUF; | |
1034 | ||
1035 | assert(send > s); | |
1036 | ||
1037 | /* Call the low level routine asking for checks */ | |
de69f3af KW |
1038 | return NATIVE_TO_UNI(Perl_utf8n_to_uvchr(aTHX_ s, send -s, retlen, |
1039 | ckWARN_d(WARN_UTF8) ? 0 : UTF8_ALLOW_ANY)); | |
ec5f19d0 KW |
1040 | } |
1041 | ||
b76347f2 | 1042 | /* |
87cea99e | 1043 | =for apidoc utf8_length |
b76347f2 JH |
1044 | |
1045 | Return the length of the UTF-8 char encoded string C<s> in characters. | |
02eb7b47 JH |
1046 | Stops at C<e> (inclusive). If C<e E<lt> s> or if the scan would end |
1047 | up past C<e>, croaks. | |
b76347f2 JH |
1048 | |
1049 | =cut | |
1050 | */ | |
1051 | ||
1052 | STRLEN | |
35a4481c | 1053 | Perl_utf8_length(pTHX_ const U8 *s, const U8 *e) |
b76347f2 JH |
1054 | { |
1055 | STRLEN len = 0; | |
1056 | ||
7918f24d NC |
1057 | PERL_ARGS_ASSERT_UTF8_LENGTH; |
1058 | ||
8850bf83 JH |
1059 | /* Note: cannot use UTF8_IS_...() too eagerly here since e.g. |
1060 | * the bitops (especially ~) can create illegal UTF-8. | |
1061 | * In other words: in Perl UTF-8 is not just for Unicode. */ | |
1062 | ||
a3b680e6 AL |
1063 | if (e < s) |
1064 | goto warn_and_return; | |
b76347f2 | 1065 | while (s < e) { |
4cbf4130 | 1066 | s += UTF8SKIP(s); |
8e91ec7f AV |
1067 | len++; |
1068 | } | |
1069 | ||
1070 | if (e != s) { | |
1071 | len--; | |
1072 | warn_and_return: | |
9b387841 NC |
1073 | if (PL_op) |
1074 | Perl_ck_warner_d(aTHX_ packWARN(WARN_UTF8), | |
1075 | "%s in %s", unees, OP_DESC(PL_op)); | |
1076 | else | |
61a12c31 | 1077 | Perl_ck_warner_d(aTHX_ packWARN(WARN_UTF8), "%s", unees); |
b76347f2 JH |
1078 | } |
1079 | ||
1080 | return len; | |
1081 | } | |
1082 | ||
b06226ff | 1083 | /* |
87cea99e | 1084 | =for apidoc utf8_distance |
b06226ff | 1085 | |
1e54db1a | 1086 | Returns the number of UTF-8 characters between the UTF-8 pointers C<a> |
b06226ff JH |
1087 | and C<b>. |
1088 | ||
1089 | WARNING: use only if you *know* that the pointers point inside the | |
1090 | same UTF-8 buffer. | |
1091 | ||
37607a96 PK |
1092 | =cut |
1093 | */ | |
a0ed51b3 | 1094 | |
02eb7b47 | 1095 | IV |
35a4481c | 1096 | Perl_utf8_distance(pTHX_ const U8 *a, const U8 *b) |
a0ed51b3 | 1097 | { |
7918f24d NC |
1098 | PERL_ARGS_ASSERT_UTF8_DISTANCE; |
1099 | ||
bf1665bc | 1100 | return (a < b) ? -1 * (IV) utf8_length(a, b) : (IV) utf8_length(b, a); |
a0ed51b3 LW |
1101 | } |
1102 | ||
b06226ff | 1103 | /* |
87cea99e | 1104 | =for apidoc utf8_hop |
b06226ff | 1105 | |
8850bf83 JH |
1106 | Return the UTF-8 pointer C<s> displaced by C<off> characters, either |
1107 | forward or backward. | |
b06226ff JH |
1108 | |
1109 | WARNING: do not use the following unless you *know* C<off> is within | |
8850bf83 JH |
1110 | the UTF-8 data pointed to by C<s> *and* that on entry C<s> is aligned |
1111 | on the first byte of character or just after the last byte of a character. | |
b06226ff | 1112 | |
37607a96 PK |
1113 | =cut |
1114 | */ | |
a0ed51b3 LW |
1115 | |
1116 | U8 * | |
4caf7d8c | 1117 | Perl_utf8_hop(const U8 *s, SSize_t off) |
a0ed51b3 | 1118 | { |
7918f24d NC |
1119 | PERL_ARGS_ASSERT_UTF8_HOP; |
1120 | ||
8850bf83 JH |
1121 | /* Note: cannot use UTF8_IS_...() too eagerly here since e.g |
1122 | * the bitops (especially ~) can create illegal UTF-8. | |
1123 | * In other words: in Perl UTF-8 is not just for Unicode. */ | |
1124 | ||
a0ed51b3 LW |
1125 | if (off >= 0) { |
1126 | while (off--) | |
1127 | s += UTF8SKIP(s); | |
1128 | } | |
1129 | else { | |
1130 | while (off++) { | |
1131 | s--; | |
8850bf83 JH |
1132 | while (UTF8_IS_CONTINUATION(*s)) |
1133 | s--; | |
a0ed51b3 LW |
1134 | } |
1135 | } | |
4373e329 | 1136 | return (U8 *)s; |
a0ed51b3 LW |
1137 | } |
1138 | ||
6940069f | 1139 | /* |
fed3ba5d NC |
1140 | =for apidoc bytes_cmp_utf8 |
1141 | ||
a1433954 | 1142 | Compares the sequence of characters (stored as octets) in C<b>, C<blen> with the |
72d33970 FC |
1143 | sequence of characters (stored as UTF-8) |
1144 | in C<u>, C<ulen>. Returns 0 if they are | |
fed3ba5d NC |
1145 | equal, -1 or -2 if the first string is less than the second string, +1 or +2 |
1146 | if the first string is greater than the second string. | |
1147 | ||
1148 | -1 or +1 is returned if the shorter string was identical to the start of the | |
72d33970 FC |
1149 | longer string. -2 or +2 is returned if |
1150 | there was a difference between characters | |
fed3ba5d NC |
1151 | within the strings. |
1152 | ||
1153 | =cut | |
1154 | */ | |
1155 | ||
1156 | int | |
1157 | Perl_bytes_cmp_utf8(pTHX_ const U8 *b, STRLEN blen, const U8 *u, STRLEN ulen) | |
1158 | { | |
1159 | const U8 *const bend = b + blen; | |
1160 | const U8 *const uend = u + ulen; | |
1161 | ||
1162 | PERL_ARGS_ASSERT_BYTES_CMP_UTF8; | |
fed3ba5d NC |
1163 | |
1164 | while (b < bend && u < uend) { | |
1165 | U8 c = *u++; | |
1166 | if (!UTF8_IS_INVARIANT(c)) { | |
1167 | if (UTF8_IS_DOWNGRADEABLE_START(c)) { | |
1168 | if (u < uend) { | |
1169 | U8 c1 = *u++; | |
1170 | if (UTF8_IS_CONTINUATION(c1)) { | |
a62b247b | 1171 | c = EIGHT_BIT_UTF8_TO_NATIVE(c, c1); |
fed3ba5d NC |
1172 | } else { |
1173 | Perl_ck_warner_d(aTHX_ packWARN(WARN_UTF8), | |
1174 | "Malformed UTF-8 character " | |
1175 | "(unexpected non-continuation byte 0x%02x" | |
1176 | ", immediately after start byte 0x%02x)" | |
1177 | /* Dear diag.t, it's in the pod. */ | |
1178 | "%s%s", c1, c, | |
1179 | PL_op ? " in " : "", | |
1180 | PL_op ? OP_DESC(PL_op) : ""); | |
1181 | return -2; | |
1182 | } | |
1183 | } else { | |
1184 | if (PL_op) | |
1185 | Perl_ck_warner_d(aTHX_ packWARN(WARN_UTF8), | |
1186 | "%s in %s", unees, OP_DESC(PL_op)); | |
1187 | else | |
61a12c31 | 1188 | Perl_ck_warner_d(aTHX_ packWARN(WARN_UTF8), "%s", unees); |
fed3ba5d NC |
1189 | return -2; /* Really want to return undef :-) */ |
1190 | } | |
1191 | } else { | |
1192 | return -2; | |
1193 | } | |
1194 | } | |
1195 | if (*b != c) { | |
1196 | return *b < c ? -2 : +2; | |
1197 | } | |
1198 | ++b; | |
1199 | } | |
1200 | ||
1201 | if (b == bend && u == uend) | |
1202 | return 0; | |
1203 | ||
1204 | return b < bend ? +1 : -1; | |
1205 | } | |
1206 | ||
1207 | /* | |
87cea99e | 1208 | =for apidoc utf8_to_bytes |
6940069f | 1209 | |
2bbc8d55 | 1210 | Converts a string C<s> of length C<len> from UTF-8 into native byte encoding. |
a1433954 KW |
1211 | Unlike L</bytes_to_utf8>, this over-writes the original string, and |
1212 | updates C<len> to contain the new length. | |
67e989fb | 1213 | Returns zero on failure, setting C<len> to -1. |
6940069f | 1214 | |
a1433954 | 1215 | If you need a copy of the string, see L</bytes_from_utf8>. |
95be277c | 1216 | |
6940069f GS |
1217 | =cut |
1218 | */ | |
1219 | ||
1220 | U8 * | |
37607a96 | 1221 | Perl_utf8_to_bytes(pTHX_ U8 *s, STRLEN *len) |
6940069f | 1222 | { |
d4c19fe8 AL |
1223 | U8 * const save = s; |
1224 | U8 * const send = s + *len; | |
6940069f | 1225 | U8 *d; |
246fae53 | 1226 | |
7918f24d | 1227 | PERL_ARGS_ASSERT_UTF8_TO_BYTES; |
81611534 | 1228 | PERL_UNUSED_CONTEXT; |
7918f24d | 1229 | |
1e54db1a | 1230 | /* ensure valid UTF-8 and chars < 256 before updating string */ |
d4c19fe8 | 1231 | while (s < send) { |
d59937ca KW |
1232 | if (! UTF8_IS_INVARIANT(*s)) { |
1233 | if (! UTF8_IS_NEXT_CHAR_DOWNGRADEABLE(s, send)) { | |
1234 | *len = ((STRLEN) -1); | |
1235 | return 0; | |
1236 | } | |
1237 | s++; | |
dcad2880 | 1238 | } |
d59937ca | 1239 | s++; |
246fae53 | 1240 | } |
dcad2880 JH |
1241 | |
1242 | d = s = save; | |
6940069f | 1243 | while (s < send) { |
80e0b38f KW |
1244 | U8 c = *s++; |
1245 | if (! UTF8_IS_INVARIANT(c)) { | |
1246 | /* Then it is two-byte encoded */ | |
a62b247b | 1247 | c = EIGHT_BIT_UTF8_TO_NATIVE(c, *s); |
80e0b38f KW |
1248 | s++; |
1249 | } | |
1250 | *d++ = c; | |
6940069f GS |
1251 | } |
1252 | *d = '\0'; | |
246fae53 | 1253 | *len = d - save; |
6940069f GS |
1254 | return save; |
1255 | } | |
1256 | ||
1257 | /* | |
87cea99e | 1258 | =for apidoc bytes_from_utf8 |
f9a63242 | 1259 | |
2bbc8d55 | 1260 | Converts a string C<s> of length C<len> from UTF-8 into native byte encoding. |
a1433954 | 1261 | Unlike L</utf8_to_bytes> but like L</bytes_to_utf8>, returns a pointer to |
ef9edfd0 JH |
1262 | the newly-created string, and updates C<len> to contain the new |
1263 | length. Returns the original string if no conversion occurs, C<len> | |
72d33970 | 1264 | is unchanged. Do nothing if C<is_utf8> points to 0. Sets C<is_utf8> to |
2bbc8d55 | 1265 | 0 if C<s> is converted or consisted entirely of characters that are invariant |
4a4088c4 | 1266 | in UTF-8 (i.e., US-ASCII on non-EBCDIC machines). |
f9a63242 | 1267 | |
37607a96 PK |
1268 | =cut |
1269 | */ | |
f9a63242 JH |
1270 | |
1271 | U8 * | |
e1ec3a88 | 1272 | Perl_bytes_from_utf8(pTHX_ const U8 *s, STRLEN *len, bool *is_utf8) |
f9a63242 | 1273 | { |
f9a63242 | 1274 | U8 *d; |
e1ec3a88 AL |
1275 | const U8 *start = s; |
1276 | const U8 *send; | |
f9a63242 JH |
1277 | I32 count = 0; |
1278 | ||
7918f24d | 1279 | PERL_ARGS_ASSERT_BYTES_FROM_UTF8; |
96a5add6 | 1280 | PERL_UNUSED_CONTEXT; |
f9a63242 | 1281 | if (!*is_utf8) |
73d840c0 | 1282 | return (U8 *)start; |
f9a63242 | 1283 | |
1e54db1a | 1284 | /* ensure valid UTF-8 and chars < 256 before converting string */ |
f9a63242 | 1285 | for (send = s + *len; s < send;) { |
d59937ca KW |
1286 | if (! UTF8_IS_INVARIANT(*s)) { |
1287 | if (! UTF8_IS_NEXT_CHAR_DOWNGRADEABLE(s, send)) { | |
73d840c0 | 1288 | return (U8 *)start; |
d59937ca KW |
1289 | } |
1290 | count++; | |
1291 | s++; | |
db42d148 | 1292 | } |
d59937ca | 1293 | s++; |
f9a63242 JH |
1294 | } |
1295 | ||
35da51f7 | 1296 | *is_utf8 = FALSE; |
f9a63242 | 1297 | |
212542aa | 1298 | Newx(d, (*len) - count + 1, U8); |
ef9edfd0 | 1299 | s = start; start = d; |
f9a63242 JH |
1300 | while (s < send) { |
1301 | U8 c = *s++; | |
1a91c45d | 1302 | if (! UTF8_IS_INVARIANT(c)) { |
c4d5f83a | 1303 | /* Then it is two-byte encoded */ |
a62b247b | 1304 | c = EIGHT_BIT_UTF8_TO_NATIVE(c, *s); |
1a91c45d | 1305 | s++; |
c4d5f83a NIS |
1306 | } |
1307 | *d++ = c; | |
f9a63242 JH |
1308 | } |
1309 | *d = '\0'; | |
1310 | *len = d - start; | |
73d840c0 | 1311 | return (U8 *)start; |
f9a63242 JH |
1312 | } |
1313 | ||
1314 | /* | |
87cea99e | 1315 | =for apidoc bytes_to_utf8 |
6940069f | 1316 | |
ff97e5cf KW |
1317 | Converts a string C<s> of length C<len> bytes from the native encoding into |
1318 | UTF-8. | |
6662521e | 1319 | Returns a pointer to the newly-created string, and sets C<len> to |
ff97e5cf | 1320 | reflect the new length in bytes. |
6940069f | 1321 | |
75200dff | 1322 | A C<NUL> character will be written after the end of the string. |
2bbc8d55 SP |
1323 | |
1324 | If you want to convert to UTF-8 from encodings other than | |
1325 | the native (Latin1 or EBCDIC), | |
a1433954 | 1326 | see L</sv_recode_to_utf8>(). |
c9ada85f | 1327 | |
497711e7 | 1328 | =cut |
6940069f GS |
1329 | */ |
1330 | ||
c682ebef FC |
1331 | /* This logic is duplicated in sv_catpvn_flags, so any bug fixes will |
1332 | likewise need duplication. */ | |
1333 | ||
6940069f | 1334 | U8* |
35a4481c | 1335 | Perl_bytes_to_utf8(pTHX_ const U8 *s, STRLEN *len) |
6940069f | 1336 | { |
35a4481c | 1337 | const U8 * const send = s + (*len); |
6940069f GS |
1338 | U8 *d; |
1339 | U8 *dst; | |
7918f24d NC |
1340 | |
1341 | PERL_ARGS_ASSERT_BYTES_TO_UTF8; | |
96a5add6 | 1342 | PERL_UNUSED_CONTEXT; |
6940069f | 1343 | |
212542aa | 1344 | Newx(d, (*len) * 2 + 1, U8); |
6940069f GS |
1345 | dst = d; |
1346 | ||
1347 | while (s < send) { | |
55d09dc8 KW |
1348 | append_utf8_from_native_byte(*s, &d); |
1349 | s++; | |
6940069f GS |
1350 | } |
1351 | *d = '\0'; | |
6662521e | 1352 | *len = d-dst; |
6940069f GS |
1353 | return dst; |
1354 | } | |
1355 | ||
a0ed51b3 | 1356 | /* |
dea0fc0b | 1357 | * Convert native (big-endian) or reversed (little-endian) UTF-16 to UTF-8. |
a0ed51b3 LW |
1358 | * |
1359 | * Destination must be pre-extended to 3/2 source. Do not use in-place. | |
1360 | * We optimize for native, for obvious reasons. */ | |
1361 | ||
1362 | U8* | |
dea0fc0b | 1363 | Perl_utf16_to_utf8(pTHX_ U8* p, U8* d, I32 bytelen, I32 *newlen) |
a0ed51b3 | 1364 | { |
dea0fc0b JH |
1365 | U8* pend; |
1366 | U8* dstart = d; | |
1367 | ||
7918f24d NC |
1368 | PERL_ARGS_ASSERT_UTF16_TO_UTF8; |
1369 | ||
dea0fc0b | 1370 | if (bytelen & 1) |
f5992bc4 | 1371 | Perl_croak(aTHX_ "panic: utf16_to_utf8: odd bytelen %"UVuf, (UV)bytelen); |
dea0fc0b JH |
1372 | |
1373 | pend = p + bytelen; | |
1374 | ||
a0ed51b3 | 1375 | while (p < pend) { |
dea0fc0b JH |
1376 | UV uv = (p[0] << 8) + p[1]; /* UTF-16BE */ |
1377 | p += 2; | |
2d1545e5 | 1378 | if (OFFUNI_IS_INVARIANT(uv)) { |
56d37426 | 1379 | *d++ = LATIN1_TO_NATIVE((U8) uv); |
a0ed51b3 LW |
1380 | continue; |
1381 | } | |
56d37426 KW |
1382 | if (uv <= MAX_UTF8_TWO_BYTE) { |
1383 | *d++ = UTF8_TWO_BYTE_HI(UNI_TO_NATIVE(uv)); | |
1384 | *d++ = UTF8_TWO_BYTE_LO(UNI_TO_NATIVE(uv)); | |
a0ed51b3 LW |
1385 | continue; |
1386 | } | |
46956fad KW |
1387 | #define FIRST_HIGH_SURROGATE UNICODE_SURROGATE_FIRST |
1388 | #define LAST_HIGH_SURROGATE 0xDBFF | |
1389 | #define FIRST_LOW_SURROGATE 0xDC00 | |
1390 | #define LAST_LOW_SURROGATE UNICODE_SURROGATE_LAST | |
e23c50db KW |
1391 | |
1392 | /* This assumes that most uses will be in the first Unicode plane, not | |
1393 | * needing surrogates */ | |
1394 | if (UNLIKELY(uv >= UNICODE_SURROGATE_FIRST | |
1395 | && uv <= UNICODE_SURROGATE_LAST)) | |
1396 | { | |
1397 | if (UNLIKELY(p >= pend) || UNLIKELY(uv > LAST_HIGH_SURROGATE)) { | |
1398 | Perl_croak(aTHX_ "Malformed UTF-16 surrogate"); | |
1399 | } | |
1400 | else { | |
01ea242b | 1401 | UV low = (p[0] << 8) + p[1]; |
e23c50db KW |
1402 | if ( UNLIKELY(low < FIRST_LOW_SURROGATE) |
1403 | || UNLIKELY(low > LAST_LOW_SURROGATE)) | |
1404 | { | |
01ea242b | 1405 | Perl_croak(aTHX_ "Malformed UTF-16 surrogate"); |
e23c50db KW |
1406 | } |
1407 | p += 2; | |
46956fad KW |
1408 | uv = ((uv - FIRST_HIGH_SURROGATE) << 10) |
1409 | + (low - FIRST_LOW_SURROGATE) + 0x10000; | |
01ea242b | 1410 | } |
a0ed51b3 | 1411 | } |
56d37426 KW |
1412 | #ifdef EBCDIC |
1413 | d = uvoffuni_to_utf8_flags(d, uv, 0); | |
1414 | #else | |
a0ed51b3 | 1415 | if (uv < 0x10000) { |
eb160463 GS |
1416 | *d++ = (U8)(( uv >> 12) | 0xe0); |
1417 | *d++ = (U8)(((uv >> 6) & 0x3f) | 0x80); | |
1418 | *d++ = (U8)(( uv & 0x3f) | 0x80); | |
a0ed51b3 LW |
1419 | continue; |
1420 | } | |
1421 | else { | |
eb160463 GS |
1422 | *d++ = (U8)(( uv >> 18) | 0xf0); |
1423 | *d++ = (U8)(((uv >> 12) & 0x3f) | 0x80); | |
1424 | *d++ = (U8)(((uv >> 6) & 0x3f) | 0x80); | |
1425 | *d++ = (U8)(( uv & 0x3f) | 0x80); | |
a0ed51b3 LW |
1426 | continue; |
1427 | } | |
56d37426 | 1428 | #endif |
a0ed51b3 | 1429 | } |
dea0fc0b | 1430 | *newlen = d - dstart; |
a0ed51b3 LW |
1431 | return d; |
1432 | } | |
1433 | ||
1434 | /* Note: this one is slightly destructive of the source. */ | |
1435 | ||
1436 | U8* | |
dea0fc0b | 1437 | Perl_utf16_to_utf8_reversed(pTHX_ U8* p, U8* d, I32 bytelen, I32 *newlen) |
a0ed51b3 LW |
1438 | { |
1439 | U8* s = (U8*)p; | |
d4c19fe8 | 1440 | U8* const send = s + bytelen; |
7918f24d NC |
1441 | |
1442 | PERL_ARGS_ASSERT_UTF16_TO_UTF8_REVERSED; | |
1443 | ||
e0ea5e2d NC |
1444 | if (bytelen & 1) |
1445 | Perl_croak(aTHX_ "panic: utf16_to_utf8_reversed: odd bytelen %"UVuf, | |
1446 | (UV)bytelen); | |
1447 | ||
a0ed51b3 | 1448 | while (s < send) { |
d4c19fe8 | 1449 | const U8 tmp = s[0]; |
a0ed51b3 LW |
1450 | s[0] = s[1]; |
1451 | s[1] = tmp; | |
1452 | s += 2; | |
1453 | } | |
dea0fc0b | 1454 | return utf16_to_utf8(p, d, bytelen, newlen); |
a0ed51b3 LW |
1455 | } |
1456 | ||
922e8cb4 KW |
1457 | bool |
1458 | Perl__is_uni_FOO(pTHX_ const U8 classnum, const UV c) | |
1459 | { | |
1460 | U8 tmpbuf[UTF8_MAXBYTES+1]; | |
1461 | uvchr_to_utf8(tmpbuf, c); | |
1462 | return _is_utf8_FOO(classnum, tmpbuf); | |
1463 | } | |
1464 | ||
f9ae8fb6 JD |
1465 | /* Internal function so we can deprecate the external one, and call |
1466 | this one from other deprecated functions in this file */ | |
1467 | ||
f2645549 KW |
1468 | bool |
1469 | Perl__is_utf8_idstart(pTHX_ const U8 *p) | |
61b19385 | 1470 | { |
f2645549 | 1471 | PERL_ARGS_ASSERT__IS_UTF8_IDSTART; |
61b19385 KW |
1472 | |
1473 | if (*p == '_') | |
1474 | return TRUE; | |
f25ce844 | 1475 | return is_utf8_common(p, &PL_utf8_idstart, "IdStart", NULL); |
61b19385 KW |
1476 | } |
1477 | ||
5092f92a | 1478 | bool |
eba68aa0 KW |
1479 | Perl__is_uni_perl_idcont(pTHX_ UV c) |
1480 | { | |
1481 | U8 tmpbuf[UTF8_MAXBYTES+1]; | |
1482 | uvchr_to_utf8(tmpbuf, c); | |
1483 | return _is_utf8_perl_idcont(tmpbuf); | |
1484 | } | |
1485 | ||
1486 | bool | |
f91dcd13 KW |
1487 | Perl__is_uni_perl_idstart(pTHX_ UV c) |
1488 | { | |
1489 | U8 tmpbuf[UTF8_MAXBYTES+1]; | |
1490 | uvchr_to_utf8(tmpbuf, c); | |
1491 | return _is_utf8_perl_idstart(tmpbuf); | |
1492 | } | |
1493 | ||
3a4c58c9 KW |
1494 | UV |
1495 | Perl__to_upper_title_latin1(pTHX_ const U8 c, U8* p, STRLEN *lenp, const char S_or_s) | |
1496 | { | |
1497 | /* We have the latin1-range values compiled into the core, so just use | |
4a4088c4 | 1498 | * those, converting the result to UTF-8. The only difference between upper |
3a4c58c9 KW |
1499 | * and title case in this range is that LATIN_SMALL_LETTER_SHARP_S is |
1500 | * either "SS" or "Ss". Which one to use is passed into the routine in | |
1501 | * 'S_or_s' to avoid a test */ | |
1502 | ||
1503 | UV converted = toUPPER_LATIN1_MOD(c); | |
1504 | ||
1505 | PERL_ARGS_ASSERT__TO_UPPER_TITLE_LATIN1; | |
1506 | ||
1507 | assert(S_or_s == 'S' || S_or_s == 's'); | |
1508 | ||
6f2d5cbc | 1509 | if (UVCHR_IS_INVARIANT(converted)) { /* No difference between the two for |
f4cd282c | 1510 | characters in this range */ |
3a4c58c9 KW |
1511 | *p = (U8) converted; |
1512 | *lenp = 1; | |
1513 | return converted; | |
1514 | } | |
1515 | ||
1516 | /* toUPPER_LATIN1_MOD gives the correct results except for three outliers, | |
1517 | * which it maps to one of them, so as to only have to have one check for | |
1518 | * it in the main case */ | |
1519 | if (UNLIKELY(converted == LATIN_SMALL_LETTER_Y_WITH_DIAERESIS)) { | |
1520 | switch (c) { | |
1521 | case LATIN_SMALL_LETTER_Y_WITH_DIAERESIS: | |
1522 | converted = LATIN_CAPITAL_LETTER_Y_WITH_DIAERESIS; | |
1523 | break; | |
1524 | case MICRO_SIGN: | |
1525 | converted = GREEK_CAPITAL_LETTER_MU; | |
1526 | break; | |
79e064b9 KW |
1527 | #if UNICODE_MAJOR_VERSION > 2 \ |
1528 | || (UNICODE_MAJOR_VERSION == 2 && UNICODE_DOT_VERSION >= 1 \ | |
1529 | && UNICODE_DOT_DOT_VERSION >= 8) | |
3a4c58c9 KW |
1530 | case LATIN_SMALL_LETTER_SHARP_S: |
1531 | *(p)++ = 'S'; | |
1532 | *p = S_or_s; | |
1533 | *lenp = 2; | |
1534 | return 'S'; | |
79e064b9 | 1535 | #endif |
3a4c58c9 KW |
1536 | default: |
1537 | Perl_croak(aTHX_ "panic: to_upper_title_latin1 did not expect '%c' to map to '%c'", c, LATIN_SMALL_LETTER_Y_WITH_DIAERESIS); | |
e5964223 | 1538 | NOT_REACHED; /* NOTREACHED */ |
3a4c58c9 KW |
1539 | } |
1540 | } | |
1541 | ||
1542 | *(p)++ = UTF8_TWO_BYTE_HI(converted); | |
1543 | *p = UTF8_TWO_BYTE_LO(converted); | |
1544 | *lenp = 2; | |
1545 | ||
1546 | return converted; | |
1547 | } | |
1548 | ||
50bda2c3 KW |
1549 | /* Call the function to convert a UTF-8 encoded character to the specified case. |
1550 | * Note that there may be more than one character in the result. | |
1551 | * INP is a pointer to the first byte of the input character | |
1552 | * OUTP will be set to the first byte of the string of changed characters. It | |
1553 | * needs to have space for UTF8_MAXBYTES_CASE+1 bytes | |
1554 | * LENP will be set to the length in bytes of the string of changed characters | |
1555 | * | |
1556 | * The functions return the ordinal of the first character in the string of OUTP */ | |
b9992569 KW |
1557 | #define CALL_UPPER_CASE(uv, s, d, lenp) _to_utf8_case(uv, s, d, lenp, &PL_utf8_toupper, "ToUc", "") |
1558 | #define CALL_TITLE_CASE(uv, s, d, lenp) _to_utf8_case(uv, s, d, lenp, &PL_utf8_totitle, "ToTc", "") | |
1559 | #define CALL_LOWER_CASE(uv, s, d, lenp) _to_utf8_case(uv, s, d, lenp, &PL_utf8_tolower, "ToLc", "") | |
50bda2c3 | 1560 | |
b9992569 KW |
1561 | /* This additionally has the input parameter 'specials', which if non-zero will |
1562 | * cause this to use the specials hash for folding (meaning get full case | |
50bda2c3 | 1563 | * folding); otherwise, when zero, this implies a simple case fold */ |
b9992569 | 1564 | #define CALL_FOLD_CASE(uv, s, d, lenp, specials) _to_utf8_case(uv, s, d, lenp, &PL_utf8_tofold, "ToCf", (specials) ? "" : NULL) |
c3fd2246 | 1565 | |
84afefe6 JH |
1566 | UV |
1567 | Perl_to_uni_upper(pTHX_ UV c, U8* p, STRLEN *lenp) | |
a0ed51b3 | 1568 | { |
a1433954 KW |
1569 | /* Convert the Unicode character whose ordinal is <c> to its uppercase |
1570 | * version and store that in UTF-8 in <p> and its length in bytes in <lenp>. | |
1571 | * Note that the <p> needs to be at least UTF8_MAXBYTES_CASE+1 bytes since | |
c3fd2246 KW |
1572 | * the changed version may be longer than the original character. |
1573 | * | |
1574 | * The ordinal of the first character of the changed version is returned | |
1575 | * (but note, as explained above, that there may be more.) */ | |
1576 | ||
7918f24d NC |
1577 | PERL_ARGS_ASSERT_TO_UNI_UPPER; |
1578 | ||
3a4c58c9 KW |
1579 | if (c < 256) { |
1580 | return _to_upper_title_latin1((U8) c, p, lenp, 'S'); | |
1581 | } | |
1582 | ||
0ebc6274 | 1583 | uvchr_to_utf8(p, c); |
b9992569 | 1584 | return CALL_UPPER_CASE(c, p, p, lenp); |
a0ed51b3 LW |
1585 | } |
1586 | ||
84afefe6 JH |
1587 | UV |
1588 | Perl_to_uni_title(pTHX_ UV c, U8* p, STRLEN *lenp) | |
a0ed51b3 | 1589 | { |
7918f24d NC |
1590 | PERL_ARGS_ASSERT_TO_UNI_TITLE; |
1591 | ||
3a4c58c9 KW |
1592 | if (c < 256) { |
1593 | return _to_upper_title_latin1((U8) c, p, lenp, 's'); | |
1594 | } | |
1595 | ||
0ebc6274 | 1596 | uvchr_to_utf8(p, c); |
b9992569 | 1597 | return CALL_TITLE_CASE(c, p, p, lenp); |
a0ed51b3 LW |
1598 | } |
1599 | ||
afc16117 | 1600 | STATIC U8 |
81611534 | 1601 | S_to_lower_latin1(const U8 c, U8* p, STRLEN *lenp) |
afc16117 KW |
1602 | { |
1603 | /* We have the latin1-range values compiled into the core, so just use | |
4a4088c4 | 1604 | * those, converting the result to UTF-8. Since the result is always just |
a1433954 | 1605 | * one character, we allow <p> to be NULL */ |
afc16117 KW |
1606 | |
1607 | U8 converted = toLOWER_LATIN1(c); | |
1608 | ||
1609 | if (p != NULL) { | |
6f2d5cbc | 1610 | if (NATIVE_BYTE_IS_INVARIANT(converted)) { |
afc16117 KW |
1611 | *p = converted; |
1612 | *lenp = 1; | |
1613 | } | |
1614 | else { | |
430c9760 KW |
1615 | /* Result is known to always be < 256, so can use the EIGHT_BIT |
1616 | * macros */ | |
1617 | *p = UTF8_EIGHT_BIT_HI(converted); | |
1618 | *(p+1) = UTF8_EIGHT_BIT_LO(converted); | |
afc16117 KW |
1619 | *lenp = 2; |
1620 | } | |
1621 | } | |
1622 | return converted; | |
1623 | } | |
1624 | ||
84afefe6 JH |
1625 | UV |
1626 | Perl_to_uni_lower(pTHX_ UV c, U8* p, STRLEN *lenp) | |
a0ed51b3 | 1627 | { |
7918f24d NC |
1628 | PERL_ARGS_ASSERT_TO_UNI_LOWER; |
1629 | ||
afc16117 KW |
1630 | if (c < 256) { |
1631 | return to_lower_latin1((U8) c, p, lenp); | |
bca00c02 KW |
1632 | } |
1633 | ||
afc16117 | 1634 | uvchr_to_utf8(p, c); |
b9992569 | 1635 | return CALL_LOWER_CASE(c, p, p, lenp); |
a0ed51b3 LW |
1636 | } |
1637 | ||
84afefe6 | 1638 | UV |
51910141 | 1639 | Perl__to_fold_latin1(pTHX_ const U8 c, U8* p, STRLEN *lenp, const unsigned int flags) |
a1dde8de | 1640 | { |
51910141 | 1641 | /* Corresponds to to_lower_latin1(); <flags> bits meanings: |
1ca267a5 | 1642 | * FOLD_FLAGS_NOMIX_ASCII iff non-ASCII to ASCII folds are prohibited |
51910141 | 1643 | * FOLD_FLAGS_FULL iff full folding is to be used; |
1ca267a5 KW |
1644 | * |
1645 | * Not to be used for locale folds | |
51910141 | 1646 | */ |
f673fad4 | 1647 | |
a1dde8de KW |
1648 | UV converted; |
1649 | ||
1650 | PERL_ARGS_ASSERT__TO_FOLD_LATIN1; | |
81611534 | 1651 | PERL_UNUSED_CONTEXT; |
a1dde8de | 1652 | |
1ca267a5 KW |
1653 | assert (! (flags & FOLD_FLAGS_LOCALE)); |
1654 | ||
659a7c2d | 1655 | if (UNLIKELY(c == MICRO_SIGN)) { |
a1dde8de KW |
1656 | converted = GREEK_SMALL_LETTER_MU; |
1657 | } | |
9b63e895 KW |
1658 | #if UNICODE_MAJOR_VERSION > 3 /* no multifolds in early Unicode */ \ |
1659 | || (UNICODE_MAJOR_VERSION == 3 && ( UNICODE_DOT_VERSION > 0) \ | |
1660 | || UNICODE_DOT_DOT_VERSION > 0) | |
659a7c2d KW |
1661 | else if ( (flags & FOLD_FLAGS_FULL) |
1662 | && UNLIKELY(c == LATIN_SMALL_LETTER_SHARP_S)) | |
1663 | { | |
1ca267a5 KW |
1664 | /* If can't cross 127/128 boundary, can't return "ss"; instead return |
1665 | * two U+017F characters, as fc("\df") should eq fc("\x{17f}\x{17f}") | |
1666 | * under those circumstances. */ | |
1667 | if (flags & FOLD_FLAGS_NOMIX_ASCII) { | |
1668 | *lenp = 2 * sizeof(LATIN_SMALL_LETTER_LONG_S_UTF8) - 2; | |
1669 | Copy(LATIN_SMALL_LETTER_LONG_S_UTF8 LATIN_SMALL_LETTER_LONG_S_UTF8, | |
1670 | p, *lenp, U8); | |
1671 | return LATIN_SMALL_LETTER_LONG_S; | |
1672 | } | |
1673 | else { | |
4f489194 KW |
1674 | *(p)++ = 's'; |
1675 | *p = 's'; | |
1676 | *lenp = 2; | |
1677 | return 's'; | |
1ca267a5 | 1678 | } |
a1dde8de | 1679 | } |
9b63e895 | 1680 | #endif |
a1dde8de KW |
1681 | else { /* In this range the fold of all other characters is their lower |
1682 | case */ | |
1683 | converted = toLOWER_LATIN1(c); | |
1684 | } | |
1685 | ||
6f2d5cbc | 1686 | if (UVCHR_IS_INVARIANT(converted)) { |
a1dde8de KW |
1687 | *p = (U8) converted; |
1688 | *lenp = 1; | |
1689 | } | |
1690 | else { | |
1691 | *(p)++ = UTF8_TWO_BYTE_HI(converted); | |
1692 | *p = UTF8_TWO_BYTE_LO(converted); | |
1693 | *lenp = 2; | |
1694 | } | |
1695 | ||
1696 | return converted; | |
1697 | } | |
1698 | ||
1699 | UV | |
31f05a37 | 1700 | Perl__to_uni_fold_flags(pTHX_ UV c, U8* p, STRLEN *lenp, U8 flags) |
84afefe6 | 1701 | { |
4b593389 | 1702 | |
a0270393 KW |
1703 | /* Not currently externally documented, and subject to change |
1704 | * <flags> bits meanings: | |
1705 | * FOLD_FLAGS_FULL iff full folding is to be used; | |
31f05a37 KW |
1706 | * FOLD_FLAGS_LOCALE is set iff the rules from the current underlying |
1707 | * locale are to be used. | |
a0270393 KW |
1708 | * FOLD_FLAGS_NOMIX_ASCII iff non-ASCII to ASCII folds are prohibited |
1709 | */ | |
4b593389 | 1710 | |
36bb2ab6 | 1711 | PERL_ARGS_ASSERT__TO_UNI_FOLD_FLAGS; |
7918f24d | 1712 | |
780fcc9f KW |
1713 | if (flags & FOLD_FLAGS_LOCALE) { |
1714 | /* Treat a UTF-8 locale as not being in locale at all */ | |
1715 | if (IN_UTF8_CTYPE_LOCALE) { | |
1716 | flags &= ~FOLD_FLAGS_LOCALE; | |
1717 | } | |
1718 | else { | |
1719 | _CHECK_AND_WARN_PROBLEMATIC_LOCALE; | |
e7b7ac46 | 1720 | goto needs_full_generality; |
780fcc9f | 1721 | } |
31f05a37 KW |
1722 | } |
1723 | ||
a1dde8de | 1724 | if (c < 256) { |
e7b7ac46 | 1725 | return _to_fold_latin1((U8) c, p, lenp, |
31f05a37 | 1726 | flags & (FOLD_FLAGS_FULL | FOLD_FLAGS_NOMIX_ASCII)); |
a1dde8de KW |
1727 | } |
1728 | ||
2f306ab9 | 1729 | /* Here, above 255. If no special needs, just use the macro */ |
a0270393 KW |
1730 | if ( ! (flags & (FOLD_FLAGS_LOCALE|FOLD_FLAGS_NOMIX_ASCII))) { |
1731 | uvchr_to_utf8(p, c); | |
b9992569 | 1732 | return CALL_FOLD_CASE(c, p, p, lenp, flags & FOLD_FLAGS_FULL); |
a0270393 KW |
1733 | } |
1734 | else { /* Otherwise, _to_utf8_fold_flags has the intelligence to deal with | |
1735 | the special flags. */ | |
1736 | U8 utf8_c[UTF8_MAXBYTES + 1]; | |
e7b7ac46 KW |
1737 | |
1738 | needs_full_generality: | |
a0270393 | 1739 | uvchr_to_utf8(utf8_c, c); |
445bf929 | 1740 | return _to_utf8_fold_flags(utf8_c, p, lenp, flags); |
a0270393 | 1741 | } |
84afefe6 JH |
1742 | } |
1743 | ||
26483009 | 1744 | PERL_STATIC_INLINE bool |
5141f98e | 1745 | S_is_utf8_common(pTHX_ const U8 *const p, SV **swash, |
f25ce844 | 1746 | const char *const swashname, SV* const invlist) |
bde6a22d | 1747 | { |
ea317ccb KW |
1748 | /* returns a boolean giving whether or not the UTF8-encoded character that |
1749 | * starts at <p> is in the swash indicated by <swashname>. <swash> | |
1750 | * contains a pointer to where the swash indicated by <swashname> | |
1751 | * is to be stored; which this routine will do, so that future calls will | |
f25ce844 KW |
1752 | * look at <*swash> and only generate a swash if it is not null. <invlist> |
1753 | * is NULL or an inversion list that defines the swash. If not null, it | |
1754 | * saves time during initialization of the swash. | |
ea317ccb KW |
1755 | * |
1756 | * Note that it is assumed that the buffer length of <p> is enough to | |
1757 | * contain all the bytes that comprise the character. Thus, <*p> should | |
1758 | * have been checked before this call for mal-formedness enough to assure | |
1759 | * that. */ | |
1760 | ||
7918f24d NC |
1761 | PERL_ARGS_ASSERT_IS_UTF8_COMMON; |
1762 | ||
492a624f | 1763 | /* The API should have included a length for the UTF-8 character in <p>, |
28123549 | 1764 | * but it doesn't. We therefore assume that p has been validated at least |
492a624f KW |
1765 | * as far as there being enough bytes available in it to accommodate the |
1766 | * character without reading beyond the end, and pass that number on to the | |
1767 | * validating routine */ | |
6302f837 | 1768 | if (! isUTF8_CHAR(p, p + UTF8SKIP(p))) { |
28123549 KW |
1769 | if (ckWARN_d(WARN_UTF8)) { |
1770 | Perl_warner(aTHX_ packWARN2(WARN_DEPRECATED,WARN_UTF8), | |
9816f121 | 1771 | "Passing malformed UTF-8 to \"%s\" is deprecated", swashname); |
28123549 KW |
1772 | if (ckWARN(WARN_UTF8)) { /* This will output details as to the |
1773 | what the malformation is */ | |
1774 | utf8_to_uvchr_buf(p, p + UTF8SKIP(p), NULL); | |
1775 | } | |
1776 | } | |
1777 | return FALSE; | |
1778 | } | |
87367d5f KW |
1779 | if (!*swash) { |
1780 | U8 flags = _CORE_SWASH_INIT_ACCEPT_INVLIST; | |
f25ce844 KW |
1781 | *swash = _core_swash_init("utf8", |
1782 | ||
1783 | /* Only use the name if there is no inversion | |
1784 | * list; otherwise will go out to disk */ | |
1785 | (invlist) ? "" : swashname, | |
1786 | ||
1787 | &PL_sv_undef, 1, 0, invlist, &flags); | |
87367d5f | 1788 | } |
28123549 | 1789 | |
bde6a22d NC |
1790 | return swash_fetch(*swash, p, TRUE) != 0; |
1791 | } | |
1792 | ||
1793 | bool | |
922e8cb4 KW |
1794 | Perl__is_utf8_FOO(pTHX_ const U8 classnum, const U8 *p) |
1795 | { | |
922e8cb4 KW |
1796 | PERL_ARGS_ASSERT__IS_UTF8_FOO; |
1797 | ||
1798 | assert(classnum < _FIRST_NON_SWASH_CC); | |
1799 | ||
f25ce844 KW |
1800 | return is_utf8_common(p, |
1801 | &PL_utf8_swash_ptrs[classnum], | |
1802 | swash_property_names[classnum], | |
1803 | PL_XPosix_ptrs[classnum]); | |
922e8cb4 KW |
1804 | } |
1805 | ||
1806 | bool | |
f2645549 | 1807 | Perl__is_utf8_perl_idstart(pTHX_ const U8 *p) |
a0ed51b3 | 1808 | { |
f2645549 | 1809 | SV* invlist = NULL; |
7918f24d | 1810 | |
f2645549 | 1811 | PERL_ARGS_ASSERT__IS_UTF8_PERL_IDSTART; |
7918f24d | 1812 | |
f2645549 KW |
1813 | if (! PL_utf8_perl_idstart) { |
1814 | invlist = _new_invlist_C_array(_Perl_IDStart_invlist); | |
1815 | } | |
60071a22 | 1816 | return is_utf8_common(p, &PL_utf8_perl_idstart, "_Perl_IDStart", invlist); |
82686b01 JH |
1817 | } |
1818 | ||
1819 | bool | |
f2645549 | 1820 | Perl__is_utf8_xidstart(pTHX_ const U8 *p) |
c11ff943 | 1821 | { |
f2645549 | 1822 | PERL_ARGS_ASSERT__IS_UTF8_XIDSTART; |
c11ff943 KW |
1823 | |
1824 | if (*p == '_') | |
1825 | return TRUE; | |
f25ce844 | 1826 | return is_utf8_common(p, &PL_utf8_xidstart, "XIdStart", NULL); |
c11ff943 KW |
1827 | } |
1828 | ||
1829 | bool | |
eba68aa0 KW |
1830 | Perl__is_utf8_perl_idcont(pTHX_ const U8 *p) |
1831 | { | |
b24b43f7 | 1832 | SV* invlist = NULL; |
eba68aa0 KW |
1833 | |
1834 | PERL_ARGS_ASSERT__IS_UTF8_PERL_IDCONT; | |
1835 | ||
b24b43f7 KW |
1836 | if (! PL_utf8_perl_idcont) { |
1837 | invlist = _new_invlist_C_array(_Perl_IDCont_invlist); | |
1838 | } | |
60071a22 | 1839 | return is_utf8_common(p, &PL_utf8_perl_idcont, "_Perl_IDCont", invlist); |
eba68aa0 KW |
1840 | } |
1841 | ||
eba68aa0 | 1842 | bool |
f2645549 | 1843 | Perl__is_utf8_idcont(pTHX_ const U8 *p) |
82686b01 | 1844 | { |
f2645549 | 1845 | PERL_ARGS_ASSERT__IS_UTF8_IDCONT; |
7918f24d | 1846 | |
f25ce844 | 1847 | return is_utf8_common(p, &PL_utf8_idcont, "IdContinue", NULL); |
a0ed51b3 LW |
1848 | } |
1849 | ||
1850 | bool | |
f2645549 | 1851 | Perl__is_utf8_xidcont(pTHX_ const U8 *p) |
c11ff943 | 1852 | { |
f2645549 | 1853 | PERL_ARGS_ASSERT__IS_UTF8_XIDCONT; |
c11ff943 | 1854 | |
f25ce844 | 1855 | return is_utf8_common(p, &PL_utf8_idcont, "XIdContinue", NULL); |
c11ff943 KW |
1856 | } |
1857 | ||
1858 | bool | |
7dbf68d2 KW |
1859 | Perl__is_utf8_mark(pTHX_ const U8 *p) |
1860 | { | |
7dbf68d2 KW |
1861 | PERL_ARGS_ASSERT__IS_UTF8_MARK; |
1862 | ||
f25ce844 | 1863 | return is_utf8_common(p, &PL_utf8_mark, "IsM", NULL); |
7dbf68d2 KW |
1864 | } |
1865 | ||
6b5c0936 | 1866 | /* |
87cea99e | 1867 | =for apidoc to_utf8_case |
6b5c0936 | 1868 | |
9da1e7cb KW |
1869 | Instead use the appropriate one of L</toUPPER_utf8>, |
1870 | L</toTITLE_utf8>, | |
1871 | L</toLOWER_utf8>, | |
1872 | or L</toFOLD_utf8>. | |
1873 | ||
6fae5207 | 1874 | C<p> contains the pointer to the UTF-8 string encoding |
a1433954 KW |
1875 | the character that is being converted. This routine assumes that the character |
1876 | at C<p> is well-formed. | |
6b5c0936 | 1877 | |
6fae5207 KW |
1878 | C<ustrp> is a pointer to the character buffer to put the |
1879 | conversion result to. C<lenp> is a pointer to the length | |
6b5c0936 JH |
1880 | of the result. |
1881 | ||
6fae5207 | 1882 | C<swashp> is a pointer to the swash to use. |
6b5c0936 | 1883 | |
a1433954 | 1884 | Both the special and normal mappings are stored in F<lib/unicore/To/Foo.pl>, |
796b6530 | 1885 | and loaded by C<SWASHNEW>, using F<lib/utf8_heavy.pl>. C<special> (usually, |
0134edef | 1886 | but not always, a multicharacter mapping), is tried first. |
6b5c0936 | 1887 | |
4a8240a3 KW |
1888 | C<special> is a string, normally C<NULL> or C<"">. C<NULL> means to not use |
1889 | any special mappings; C<""> means to use the special mappings. Values other | |
1890 | than these two are treated as the name of the hash containing the special | |
1891 | mappings, like C<"utf8::ToSpecLower">. | |
6b5c0936 | 1892 | |
796b6530 KW |
1893 | C<normal> is a string like C<"ToLower"> which means the swash |
1894 | C<%utf8::ToLower>. | |
0134edef | 1895 | |
760c7c2f KW |
1896 | Code points above the platform's C<IV_MAX> will raise a deprecation warning, |
1897 | unless those are turned off. | |
1898 | ||
0134edef | 1899 | =cut */ |
6b5c0936 | 1900 | |
2104c8d9 | 1901 | UV |
9a957fbc AL |
1902 | Perl_to_utf8_case(pTHX_ const U8 *p, U8* ustrp, STRLEN *lenp, |
1903 | SV **swashp, const char *normal, const char *special) | |
a0ed51b3 | 1904 | { |
b9992569 KW |
1905 | PERL_ARGS_ASSERT_TO_UTF8_CASE; |
1906 | ||
1907 | return _to_utf8_case(valid_utf8_to_uvchr(p, NULL), p, ustrp, lenp, swashp, normal, special); | |
1908 | } | |
1909 | ||
1910 | /* change namve uv1 to 'from' */ | |
6a4a25f4 | 1911 | STATIC UV |
b9992569 KW |
1912 | S__to_utf8_case(pTHX_ const UV uv1, const U8 *p, U8* ustrp, STRLEN *lenp, |
1913 | SV **swashp, const char *normal, const char *special) | |
1914 | { | |
0134edef | 1915 | STRLEN len = 0; |
7918f24d | 1916 | |
b9992569 | 1917 | PERL_ARGS_ASSERT__TO_UTF8_CASE; |
7918f24d | 1918 | |
36eaa811 KW |
1919 | /* For code points that don't change case, we already know that the output |
1920 | * of this function is the unchanged input, so we can skip doing look-ups | |
1921 | * for them. Unfortunately the case-changing code points are scattered | |
1922 | * around. But there are some long consecutive ranges where there are no | |
1923 | * case changing code points. By adding tests, we can eliminate the lookup | |
1924 | * for all the ones in such ranges. This is currently done here only for | |
1925 | * just a few cases where the scripts are in common use in modern commerce | |
1926 | * (and scripts adjacent to those which can be included without additional | |
1927 | * tests). */ | |
1928 | ||
1929 | if (uv1 >= 0x0590) { | |
1930 | /* This keeps from needing further processing the code points most | |
1931 | * likely to be used in the following non-cased scripts: Hebrew, | |
1932 | * Arabic, Syriac, Thaana, NKo, Samaritan, Mandaic, Devanagari, | |
1933 | * Bengali, Gurmukhi, Gujarati, Oriya, Tamil, Telugu, Kannada, | |
1934 | * Malayalam, Sinhala, Thai, Lao, Tibetan, Myanmar */ | |
1935 | if (uv1 < 0x10A0) { | |
1936 | goto cases_to_self; | |
1937 | } | |
1938 | ||
1939 | /* The following largish code point ranges also don't have case | |
1940 | * changes, but khw didn't think they warranted extra tests to speed | |
1941 | * them up (which would slightly slow down everything else above them): | |
1942 | * 1100..139F Hangul Jamo, Ethiopic | |
1943 | * 1400..1CFF Unified Canadian Aboriginal Syllabics, Ogham, Runic, | |
1944 | * Tagalog, Hanunoo, Buhid, Tagbanwa, Khmer, Mongolian, | |
1945 | * Limbu, Tai Le, New Tai Lue, Buginese, Tai Tham, | |
1946 | * Combining Diacritical Marks Extended, Balinese, | |
1947 | * Sundanese, Batak, Lepcha, Ol Chiki | |
1948 | * 2000..206F General Punctuation | |
1949 | */ | |
1950 | ||
1951 | if (uv1 >= 0x2D30) { | |
1952 | ||
1953 | /* This keeps the from needing further processing the code points | |
1954 | * most likely to be used in the following non-cased major scripts: | |
1955 | * CJK, Katakana, Hiragana, plus some less-likely scripts. | |
1956 | * | |
1957 | * (0x2D30 above might have to be changed to 2F00 in the unlikely | |
1958 | * event that Unicode eventually allocates the unused block as of | |
1959 | * v8.0 2FE0..2FEF to code points that are cased. khw has verified | |
1960 | * that the test suite will start having failures to alert you | |
1961 | * should that happen) */ | |
1962 | if (uv1 < 0xA640) { | |
1963 | goto cases_to_self; | |
1964 | } | |
1965 | ||
1966 | if (uv1 >= 0xAC00) { | |
1967 | if (UNLIKELY(UNICODE_IS_SURROGATE(uv1))) { | |
5af9bc97 KW |
1968 | if (ckWARN_d(WARN_SURROGATE)) { |
1969 | const char* desc = (PL_op) ? OP_DESC(PL_op) : normal; | |
1970 | Perl_warner(aTHX_ packWARN(WARN_SURROGATE), | |
1971 | "Operation \"%s\" returns its argument for UTF-16 surrogate U+%04"UVXf"", desc, uv1); | |
1972 | } | |
1973 | goto cases_to_self; | |
1974 | } | |
36eaa811 KW |
1975 | |
1976 | /* AC00..FAFF Catches Hangul syllables and private use, plus | |
1977 | * some others */ | |
1978 | if (uv1 < 0xFB00) { | |
1979 | goto cases_to_self; | |
1980 | ||
1981 | } | |
1982 | ||
5af9bc97 KW |
1983 | if (UNLIKELY(UNICODE_IS_SUPER(uv1))) { |
1984 | if ( UNLIKELY(uv1 > MAX_NON_DEPRECATED_CP) | |
1985 | && ckWARN_d(WARN_DEPRECATED)) | |
1986 | { | |
1987 | Perl_warner(aTHX_ packWARN(WARN_DEPRECATED), | |
1988 | cp_above_legal_max, uv1, MAX_NON_DEPRECATED_CP); | |
1989 | } | |
1990 | if (ckWARN_d(WARN_NON_UNICODE)) { | |
1991 | const char* desc = (PL_op) ? OP_DESC(PL_op) : normal; | |
1992 | Perl_warner(aTHX_ packWARN(WARN_NON_UNICODE), | |
1993 | "Operation \"%s\" returns its argument for non-Unicode code point 0x%04"UVXf"", desc, uv1); | |
1994 | } | |
1995 | goto cases_to_self; | |
1996 | } | |
3bfc1e70 KW |
1997 | #ifdef HIGHEST_CASE_CHANGING_CP_FOR_USE_ONLY_BY_UTF8_DOT_C |
1998 | if (UNLIKELY(uv1 | |
1999 | > HIGHEST_CASE_CHANGING_CP_FOR_USE_ONLY_BY_UTF8_DOT_C)) | |
2000 | { | |
2001 | ||
2002 | /* As of this writing, this means we avoid swash creation | |
2003 | * for anything beyond low Plane 1 */ | |
2004 | goto cases_to_self; | |
2005 | } | |
2006 | #endif | |
36eaa811 KW |
2007 | } |
2008 | } | |
9ae3ac1a | 2009 | |
36eaa811 KW |
2010 | /* Note that non-characters are perfectly legal, so no warning should |
2011 | * be given. There are so few of them, that it isn't worth the extra | |
2012 | * tests to avoid swash creation */ | |
9ae3ac1a KW |
2013 | } |
2014 | ||
0134edef | 2015 | if (!*swashp) /* load on-demand */ |
5ab9d2ef | 2016 | *swashp = _core_swash_init("utf8", normal, &PL_sv_undef, 4, 0, NULL, NULL); |
0134edef | 2017 | |
a6f87d8c | 2018 | if (special) { |
0134edef | 2019 | /* It might be "special" (sometimes, but not always, |
2a37f04d | 2020 | * a multicharacter mapping) */ |
4a8240a3 | 2021 | HV *hv = NULL; |
b08cf34e JH |
2022 | SV **svp; |
2023 | ||
4a8240a3 KW |
2024 | /* If passed in the specials name, use that; otherwise use any |
2025 | * given in the swash */ | |
2026 | if (*special != '\0') { | |
2027 | hv = get_hv(special, 0); | |
2028 | } | |
2029 | else { | |
2030 | svp = hv_fetchs(MUTABLE_HV(SvRV(*swashp)), "SPECIALS", 0); | |
2031 | if (svp) { | |
2032 | hv = MUTABLE_HV(SvRV(*svp)); | |
2033 | } | |
2034 | } | |
2035 | ||
176fe009 | 2036 | if (hv |
5f560d8a | 2037 | && (svp = hv_fetch(hv, (const char*)p, UVCHR_SKIP(uv1), FALSE)) |
176fe009 KW |
2038 | && (*svp)) |
2039 | { | |
cfd0369c | 2040 | const char *s; |
47654450 | 2041 | |
cfd0369c | 2042 | s = SvPV_const(*svp, len); |
47654450 | 2043 | if (len == 1) |
f4cd282c | 2044 | /* EIGHTBIT */ |
c80e42f3 | 2045 | len = uvchr_to_utf8(ustrp, *(U8*)s) - ustrp; |
2a37f04d | 2046 | else { |
d2dcd0fb | 2047 | Copy(s, ustrp, len, U8); |
29e98929 | 2048 | } |
983ffd37 | 2049 | } |
0134edef JH |
2050 | } |
2051 | ||
2052 | if (!len && *swashp) { | |
4a4088c4 | 2053 | const UV uv2 = swash_fetch(*swashp, p, TRUE /* => is UTF-8 */); |
d4c19fe8 | 2054 | |
0134edef JH |
2055 | if (uv2) { |
2056 | /* It was "normal" (a single character mapping). */ | |
f4cd282c | 2057 | len = uvchr_to_utf8(ustrp, uv2) - ustrp; |
2a37f04d JH |
2058 | } |
2059 | } | |
1feea2c7 | 2060 | |
cbe07460 KW |
2061 | if (len) { |
2062 | if (lenp) { | |
2063 | *lenp = len; | |
2064 | } | |
2065 | return valid_utf8_to_uvchr(ustrp, 0); | |
2066 | } | |
2067 | ||
2068 | /* Here, there was no mapping defined, which means that the code point maps | |
2069 | * to itself. Return the inputs */ | |
e24dfe9c | 2070 | cases_to_self: |
bfdf22ec | 2071 | len = UTF8SKIP(p); |
ca9fab46 KW |
2072 | if (p != ustrp) { /* Don't copy onto itself */ |
2073 | Copy(p, ustrp, len, U8); | |
2074 | } | |
0134edef | 2075 | |
2a37f04d JH |
2076 | if (lenp) |
2077 | *lenp = len; | |
2078 | ||
f4cd282c | 2079 | return uv1; |
cbe07460 | 2080 | |
a0ed51b3 LW |
2081 | } |
2082 | ||
051a06d4 | 2083 | STATIC UV |
357aadde | 2084 | S_check_locale_boundary_crossing(pTHX_ const U8* const p, const UV result, U8* const ustrp, STRLEN *lenp) |
051a06d4 | 2085 | { |
4a4088c4 | 2086 | /* This is called when changing the case of a UTF-8-encoded character above |
31f05a37 KW |
2087 | * the Latin1 range, and the operation is in a non-UTF-8 locale. If the |
2088 | * result contains a character that crosses the 255/256 boundary, disallow | |
2089 | * the change, and return the original code point. See L<perlfunc/lc> for | |
2090 | * why; | |
051a06d4 | 2091 | * |
a1433954 KW |
2092 | * p points to the original string whose case was changed; assumed |
2093 | * by this routine to be well-formed | |
051a06d4 KW |
2094 | * result the code point of the first character in the changed-case string |
2095 | * ustrp points to the changed-case string (<result> represents its first char) | |
2096 | * lenp points to the length of <ustrp> */ | |
2097 | ||
2098 | UV original; /* To store the first code point of <p> */ | |
2099 | ||
2100 | PERL_ARGS_ASSERT_CHECK_LOCALE_BOUNDARY_CROSSING; | |
2101 | ||
a4f12ed7 | 2102 | assert(UTF8_IS_ABOVE_LATIN1(*p)); |
051a06d4 KW |
2103 | |
2104 | /* We know immediately if the first character in the string crosses the | |
2105 | * boundary, so can skip */ | |
2106 | if (result > 255) { | |
2107 | ||
2108 | /* Look at every character in the result; if any cross the | |
2109 | * boundary, the whole thing is disallowed */ | |
2110 | U8* s = ustrp + UTF8SKIP(ustrp); | |
2111 | U8* e = ustrp + *lenp; | |
2112 | while (s < e) { | |
a4f12ed7 | 2113 | if (! UTF8_IS_ABOVE_LATIN1(*s)) { |
051a06d4 KW |
2114 | goto bad_crossing; |
2115 | } | |
2116 | s += UTF8SKIP(s); | |
2117 | } | |
2118 | ||
613abc6d KW |
2119 | /* Here, no characters crossed, result is ok as-is, but we warn. */ |
2120 | _CHECK_AND_OUTPUT_WIDE_LOCALE_UTF8_MSG(p, p + UTF8SKIP(p)); | |
051a06d4 KW |
2121 | return result; |
2122 | } | |
2123 | ||
7b52d656 | 2124 | bad_crossing: |
051a06d4 KW |
2125 | |
2126 | /* Failed, have to return the original */ | |
4b88fb76 | 2127 | original = valid_utf8_to_uvchr(p, lenp); |
ab0b796c KW |
2128 | |
2129 | /* diag_listed_as: Can't do %s("%s") on non-UTF-8 locale; resolved to "%s". */ | |
2130 | Perl_ck_warner(aTHX_ packWARN(WARN_LOCALE), | |
2131 | "Can't do %s(\"\\x{%"UVXf"}\") on non-UTF-8 locale; " | |
2132 | "resolved to \"\\x{%"UVXf"}\".", | |
357aadde | 2133 | OP_DESC(PL_op), |
ab0b796c KW |
2134 | original, |
2135 | original); | |
051a06d4 KW |
2136 | Copy(p, ustrp, *lenp, char); |
2137 | return original; | |
2138 | } | |
2139 | ||
d3e79532 | 2140 | /* |
87cea99e | 2141 | =for apidoc to_utf8_upper |
d3e79532 | 2142 | |
1f607577 | 2143 | Instead use L</toUPPER_utf8>. |
a1433954 | 2144 | |
d3e79532 JH |
2145 | =cut */ |
2146 | ||
051a06d4 | 2147 | /* Not currently externally documented, and subject to change: |
31f05a37 KW |
2148 | * <flags> is set iff iff the rules from the current underlying locale are to |
2149 | * be used. */ | |
051a06d4 | 2150 | |
2104c8d9 | 2151 | UV |
31f05a37 | 2152 | Perl__to_utf8_upper_flags(pTHX_ const U8 *p, U8* ustrp, STRLEN *lenp, bool flags) |
a0ed51b3 | 2153 | { |
051a06d4 KW |
2154 | UV result; |
2155 | ||
2156 | PERL_ARGS_ASSERT__TO_UTF8_UPPER_FLAGS; | |
7918f24d | 2157 | |
780fcc9f KW |
2158 | if (flags) { |
2159 | /* Treat a UTF-8 locale as not being in locale at all */ | |
2160 | if (IN_UTF8_CTYPE_LOCALE) { | |
2161 | flags = FALSE; | |
2162 | } | |
2163 | else { | |
2164 | _CHECK_AND_WARN_PROBLEMATIC_LOCALE; | |
2165 | } | |
31f05a37 KW |
2166 | } |
2167 | ||
3a4c58c9 | 2168 | if (UTF8_IS_INVARIANT(*p)) { |
051a06d4 KW |
2169 | if (flags) { |
2170 | result = toUPPER_LC(*p); | |
2171 | } | |
2172 | else { | |
81c6c7ce | 2173 | return _to_upper_title_latin1(*p, ustrp, lenp, 'S'); |
051a06d4 | 2174 | } |
3a4c58c9 KW |
2175 | } |
2176 | else if UTF8_IS_DOWNGRADEABLE_START(*p) { | |
051a06d4 | 2177 | if (flags) { |
a62b247b | 2178 | U8 c = EIGHT_BIT_UTF8_TO_NATIVE(*p, *(p+1)); |
68067e4e | 2179 | result = toUPPER_LC(c); |
051a06d4 KW |
2180 | } |
2181 | else { | |
a62b247b | 2182 | return _to_upper_title_latin1(EIGHT_BIT_UTF8_TO_NATIVE(*p, *(p+1)), |
81c6c7ce | 2183 | ustrp, lenp, 'S'); |
051a06d4 KW |
2184 | } |
2185 | } | |
4a4088c4 | 2186 | else { /* UTF-8, ord above 255 */ |
b9992569 | 2187 | result = CALL_UPPER_CASE(valid_utf8_to_uvchr(p, NULL), p, ustrp, lenp); |
051a06d4 KW |
2188 | |
2189 | if (flags) { | |
357aadde | 2190 | result = check_locale_boundary_crossing(p, result, ustrp, lenp); |
051a06d4 KW |
2191 | } |
2192 | return result; | |
2193 | } | |
2194 | ||
4a4088c4 | 2195 | /* Here, used locale rules. Convert back to UTF-8 */ |
051a06d4 KW |
2196 | if (UTF8_IS_INVARIANT(result)) { |
2197 | *ustrp = (U8) result; | |
2198 | *lenp = 1; | |
2199 | } | |
2200 | else { | |
62cb07ea KW |
2201 | *ustrp = UTF8_EIGHT_BIT_HI((U8) result); |
2202 | *(ustrp + 1) = UTF8_EIGHT_BIT_LO((U8) result); | |
051a06d4 | 2203 | *lenp = 2; |
3a4c58c9 | 2204 | } |
baa60164 | 2205 | |
051a06d4 | 2206 | return result; |
983ffd37 | 2207 | } |
a0ed51b3 | 2208 | |
d3e79532 | 2209 | /* |
87cea99e | 2210 | =for apidoc to_utf8_title |
d3e79532 | 2211 | |
1f607577 | 2212 | Instead use L</toTITLE_utf8>. |
a1433954 | 2213 | |
d3e79532 JH |
2214 | =cut */ |
2215 | ||
051a06d4 | 2216 | /* Not currently externally documented, and subject to change: |
31f05a37 KW |
2217 | * <flags> is set iff the rules from the current underlying locale are to be |
2218 | * used. Since titlecase is not defined in POSIX, for other than a | |
2219 | * UTF-8 locale, uppercase is used instead for code points < 256. | |
445bf929 | 2220 | */ |
051a06d4 | 2221 | |
983ffd37 | 2222 | UV |
31f05a37 | 2223 | Perl__to_utf8_title_flags(pTHX_ const U8 *p, U8* ustrp, STRLEN *lenp, bool flags) |
983ffd37 | 2224 | { |
051a06d4 KW |
2225 | UV result; |
2226 | ||
2227 | PERL_ARGS_ASSERT__TO_UTF8_TITLE_FLAGS; | |
7918f24d | 2228 | |
780fcc9f KW |
2229 | if (flags) { |
2230 | /* Treat a UTF-8 locale as not being in locale at all */ | |
2231 | if (IN_UTF8_CTYPE_LOCALE) { | |
2232 | flags = FALSE; | |
2233 | } | |
2234 | else { | |
2235 | _CHECK_AND_WARN_PROBLEMATIC_LOCALE; | |
2236 | } | |
31f05a37 KW |
2237 | } |
2238 | ||
3a4c58c9 | 2239 | if (UTF8_IS_INVARIANT(*p)) { |
051a06d4 KW |
2240 | if (flags) { |
2241 | result = toUPPER_LC(*p); | |
2242 | } | |
2243 | else { | |
81c6c7ce | 2244 | return _to_upper_title_latin1(*p, ustrp, lenp, 's'); |
051a06d4 | 2245 | } |
3a4c58c9 KW |
2246 | } |
2247 | else if UTF8_IS_DOWNGRADEABLE_START(*p) { | |
051a06d4 | 2248 | if (flags) { |
a62b247b | 2249 | U8 c = EIGHT_BIT_UTF8_TO_NATIVE(*p, *(p+1)); |
68067e4e | 2250 | result = toUPPER_LC(c); |
051a06d4 KW |
2251 | } |
2252 | else { | |
a62b247b | 2253 | return _to_upper_title_latin1(EIGHT_BIT_UTF8_TO_NATIVE(*p, *(p+1)), |
81c6c7ce | 2254 | ustrp, lenp, 's'); |
051a06d4 KW |
2255 | } |
2256 | } | |
4a4088c4 | 2257 | else { /* UTF-8, ord above 255 */ |
b9992569 | 2258 | result = CALL_TITLE_CASE(valid_utf8_to_uvchr(p, NULL), p, ustrp, lenp); |
051a06d4 KW |
2259 | |
2260 | if (flags) { | |
357aadde | 2261 | result = check_locale_boundary_crossing(p, result, ustrp, lenp); |
051a06d4 KW |
2262 | } |
2263 | return result; | |
2264 | } | |
2265 | ||
4a4088c4 | 2266 | /* Here, used locale rules. Convert back to UTF-8 */ |
051a06d4 KW |
2267 | if (UTF8_IS_INVARIANT(result)) { |
2268 | *ustrp = (U8) result; | |
2269 | *lenp = 1; | |
2270 | } | |
2271 | else { | |
62cb07ea KW |
2272 | *ustrp = UTF8_EIGHT_BIT_HI((U8) result); |
2273 | *(ustrp + 1) = UTF8_EIGHT_BIT_LO((U8) result); | |
051a06d4 | 2274 | *lenp = 2; |
3a4c58c9 KW |
2275 | } |
2276 | ||
051a06d4 | 2277 | return result; |
a0ed51b3 LW |
2278 | } |
2279 | ||
d3e79532 | 2280 | /* |
87cea99e | 2281 | =for apidoc to_utf8_lower |
d3e79532 | 2282 | |
1f607577 | 2283 | Instead use L</toLOWER_utf8>. |
a1433954 | 2284 | |
d3e79532 JH |
2285 | =cut */ |
2286 | ||
051a06d4 | 2287 | /* Not currently externally documented, and subject to change: |
31f05a37 KW |
2288 | * <flags> is set iff iff the rules from the current underlying locale are to |
2289 | * be used. | |
2290 | */ | |
051a06d4 | 2291 | |
2104c8d9 | 2292 | UV |
31f05a37 | 2293 | Perl__to_utf8_lower_flags(pTHX_ const U8 *p, U8* ustrp, STRLEN *lenp, bool flags) |
a0ed51b3 | 2294 | { |
051a06d4 KW |
2295 | UV result; |
2296 | ||
051a06d4 | 2297 | PERL_ARGS_ASSERT__TO_UTF8_LOWER_FLAGS; |
7918f24d | 2298 | |
780fcc9f KW |
2299 | if (flags) { |
2300 | /* Treat a UTF-8 locale as not being in locale at all */ | |
2301 | if (IN_UTF8_CTYPE_LOCALE) { | |
2302 | flags = FALSE; | |
2303 | } | |
2304 | else { | |
2305 | _CHECK_AND_WARN_PROBLEMATIC_LOCALE; | |
2306 | } | |
31f05a37 KW |
2307 | } |
2308 | ||
968c5e6a | 2309 | if (UTF8_IS_INVARIANT(*p)) { |
051a06d4 KW |
2310 | if (flags) { |
2311 | result = toLOWER_LC(*p); | |
2312 | } | |
2313 | else { | |
81c6c7ce | 2314 | return to_lower_latin1(*p, ustrp, lenp); |
051a06d4 | 2315 | } |
968c5e6a KW |
2316 | } |
2317 | else if UTF8_IS_DOWNGRADEABLE_START(*p) { | |
051a06d4 | 2318 | if (flags) { |
a62b247b | 2319 | U8 c = EIGHT_BIT_UTF8_TO_NATIVE(*p, *(p+1)); |
68067e4e | 2320 | result = toLOWER_LC(c); |
051a06d4 KW |
2321 | } |
2322 | else { | |
a62b247b | 2323 | return to_lower_latin1(EIGHT_BIT_UTF8_TO_NATIVE(*p, *(p+1)), |
81c6c7ce | 2324 | ustrp, lenp); |
051a06d4 | 2325 | } |
968c5e6a | 2326 | } |
4a4088c4 | 2327 | else { /* UTF-8, ord above 255 */ |
b9992569 | 2328 | result = CALL_LOWER_CASE(valid_utf8_to_uvchr(p, NULL), p, ustrp, lenp); |
051a06d4 KW |
2329 | |
2330 | if (flags) { | |
357aadde | 2331 | result = check_locale_boundary_crossing(p, result, ustrp, lenp); |
051a06d4 | 2332 | } |
968c5e6a | 2333 | |
051a06d4 KW |
2334 | return result; |
2335 | } | |
2336 | ||
4a4088c4 | 2337 | /* Here, used locale rules. Convert back to UTF-8 */ |
051a06d4 KW |
2338 | if (UTF8_IS_INVARIANT(result)) { |
2339 | *ustrp = (U8) result; | |
2340 | *lenp = 1; | |
2341 | } | |
2342 | else { | |
62cb07ea KW |
2343 | *ustrp = UTF8_EIGHT_BIT_HI((U8) result); |
2344 | *(ustrp + 1) = UTF8_EIGHT_BIT_LO((U8) result); | |
051a06d4 KW |
2345 | *lenp = 2; |
2346 | } | |
2347 | ||
051a06d4 | 2348 | return result; |
b4e400f9 JH |
2349 | } |
2350 | ||
d3e79532 | 2351 | /* |
87cea99e | 2352 | =for apidoc to_utf8_fold |
d3e79532 | 2353 | |
1f607577 | 2354 | Instead use L</toFOLD_utf8>. |
a1433954 | 2355 | |
d3e79532 JH |
2356 | =cut */ |
2357 | ||
051a06d4 KW |
2358 | /* Not currently externally documented, and subject to change, |
2359 | * in <flags> | |
31f05a37 KW |
2360 | * bit FOLD_FLAGS_LOCALE is set iff the rules from the current underlying |
2361 | * locale are to be used. | |
051a06d4 KW |
2362 | * bit FOLD_FLAGS_FULL is set iff full case folds are to be used; |
2363 | * otherwise simple folds | |
a0270393 KW |
2364 | * bit FOLD_FLAGS_NOMIX_ASCII is set iff folds of non-ASCII to ASCII are |
2365 | * prohibited | |
445bf929 | 2366 | */ |
36bb2ab6 | 2367 | |
b4e400f9 | 2368 | UV |
445bf929 | 2369 | Perl__to_utf8_fold_flags(pTHX_ const U8 *p, U8* ustrp, STRLEN *lenp, U8 flags) |
b4e400f9 | 2370 | { |
051a06d4 KW |
2371 | UV result; |
2372 | ||
36bb2ab6 | 2373 | PERL_ARGS_ASSERT__TO_UTF8_FOLD_FLAGS; |
7918f24d | 2374 | |
a0270393 KW |
2375 | /* These are mutually exclusive */ |
2376 | assert (! ((flags & FOLD_FLAGS_LOCALE) && (flags & FOLD_FLAGS_NOMIX_ASCII))); | |
2377 | ||
50ba90ff KW |
2378 | assert(p != ustrp); /* Otherwise overwrites */ |
2379 | ||
780fcc9f KW |
2380 | if (flags & FOLD_FLAGS_LOCALE) { |
2381 | /* Treat a UTF-8 locale as not being in locale at all */ | |
2382 | if (IN_UTF8_CTYPE_LOCALE) { | |
2383 | flags &= ~FOLD_FLAGS_LOCALE; | |
2384 | } | |
2385 | else { | |
2386 | _CHECK_AND_WARN_PROBLEMATIC_LOCALE; | |
2387 | } | |
31f05a37 KW |
2388 | } |
2389 | ||
a1dde8de | 2390 | if (UTF8_IS_INVARIANT(*p)) { |
051a06d4 | 2391 | if (flags & FOLD_FLAGS_LOCALE) { |
d22b930b | 2392 | result = toFOLD_LC(*p); |
051a06d4 KW |
2393 | } |
2394 | else { | |
81c6c7ce | 2395 | return _to_fold_latin1(*p, ustrp, lenp, |
1ca267a5 | 2396 | flags & (FOLD_FLAGS_FULL | FOLD_FLAGS_NOMIX_ASCII)); |
051a06d4 | 2397 | } |
a1dde8de KW |
2398 | } |
2399 | else if UTF8_IS_DOWNGRADEABLE_START(*p) { | |
051a06d4 | 2400 | if (flags & FOLD_FLAGS_LOCALE) { |
a62b247b | 2401 | U8 c = EIGHT_BIT_UTF8_TO_NATIVE(*p, *(p+1)); |
68067e4e | 2402 | result = toFOLD_LC(c); |
051a06d4 KW |
2403 | } |
2404 | else { | |
a62b247b | 2405 | return _to_fold_latin1(EIGHT_BIT_UTF8_TO_NATIVE(*p, *(p+1)), |
51910141 | 2406 | ustrp, lenp, |
1ca267a5 | 2407 | flags & (FOLD_FLAGS_FULL | FOLD_FLAGS_NOMIX_ASCII)); |
051a06d4 | 2408 | } |
a1dde8de | 2409 | } |
4a4088c4 | 2410 | else { /* UTF-8, ord above 255 */ |
b9992569 | 2411 | result = CALL_FOLD_CASE(valid_utf8_to_uvchr(p, NULL), p, ustrp, lenp, flags & FOLD_FLAGS_FULL); |
a1dde8de | 2412 | |
1ca267a5 KW |
2413 | if (flags & FOLD_FLAGS_LOCALE) { |
2414 | ||
76f2ffcd | 2415 | # define LONG_S_T LATIN_SMALL_LIGATURE_LONG_S_T_UTF8 |
0766489e KW |
2416 | const unsigned int long_s_t_len = sizeof(LONG_S_T) - 1; |
2417 | ||
2418 | # ifdef LATIN_CAPITAL_LETTER_SHARP_S_UTF8 | |
2419 | # define CAP_SHARP_S LATIN_CAPITAL_LETTER_SHARP_S_UTF8 | |
76f2ffcd KW |
2420 | |
2421 | const unsigned int cap_sharp_s_len = sizeof(CAP_SHARP_S) - 1; | |
76f2ffcd | 2422 | |
538e84ed KW |
2423 | /* Special case these two characters, as what normally gets |
2424 | * returned under locale doesn't work */ | |
76f2ffcd KW |
2425 | if (UTF8SKIP(p) == cap_sharp_s_len |
2426 | && memEQ((char *) p, CAP_SHARP_S, cap_sharp_s_len)) | |
1ca267a5 | 2427 | { |
ab0b796c KW |
2428 | /* diag_listed_as: Can't do %s("%s") on non-UTF-8 locale; resolved to "%s". */ |
2429 | Perl_ck_warner(aTHX_ packWARN(WARN_LOCALE), | |
2430 | "Can't do fc(\"\\x{1E9E}\") on non-UTF-8 locale; " | |
2431 | "resolved to \"\\x{17F}\\x{17F}\"."); | |
1ca267a5 KW |
2432 | goto return_long_s; |
2433 | } | |
0766489e KW |
2434 | else |
2435 | #endif | |
2436 | if (UTF8SKIP(p) == long_s_t_len | |
76f2ffcd | 2437 | && memEQ((char *) p, LONG_S_T, long_s_t_len)) |
9fc2026f | 2438 | { |
ab0b796c KW |
2439 | /* diag_listed_as: Can't do %s("%s") on non-UTF-8 locale; resolved to "%s". */ |
2440 | Perl_ck_warner(aTHX_ packWARN(WARN_LOCALE), | |
2441 | "Can't do fc(\"\\x{FB05}\") on non-UTF-8 locale; " | |
2442 | "resolved to \"\\x{FB06}\"."); | |
9fc2026f KW |
2443 | goto return_ligature_st; |
2444 | } | |
74894415 KW |
2445 | |
2446 | #if UNICODE_MAJOR_VERSION == 3 \ | |
2447 | && UNICODE_DOT_VERSION == 0 \ | |
2448 | && UNICODE_DOT_DOT_VERSION == 1 | |
2449 | # define DOTTED_I LATIN_CAPITAL_LETTER_I_WITH_DOT_ABOVE_UTF8 | |
2450 | ||
2451 | /* And special case this on this Unicode version only, for the same | |
2452 | * reaons the other two are special cased. They would cross the | |
2453 | * 255/256 boundary which is forbidden under /l, and so the code | |
2454 | * wouldn't catch that they are equivalent (which they are only in | |
2455 | * this release) */ | |
2456 | else if (UTF8SKIP(p) == sizeof(DOTTED_I) - 1 | |
2457 | && memEQ((char *) p, DOTTED_I, sizeof(DOTTED_I) - 1)) | |
2458 | { | |
2459 | /* diag_listed_as: Can't do %s("%s") on non-UTF-8 locale; resolved to "%s". */ | |
2460 | Perl_ck_warner(aTHX_ packWARN(WARN_LOCALE), | |
2461 | "Can't do fc(\"\\x{0130}\") on non-UTF-8 locale; " | |
2462 | "resolved to \"\\x{0131}\"."); | |
2463 | goto return_dotless_i; | |
2464 | } | |
2465 | #endif | |
2466 | ||
357aadde | 2467 | return check_locale_boundary_crossing(p, result, ustrp, lenp); |
051a06d4 | 2468 | } |
a0270393 KW |
2469 | else if (! (flags & FOLD_FLAGS_NOMIX_ASCII)) { |
2470 | return result; | |
2471 | } | |
2472 | else { | |
4a4088c4 | 2473 | /* This is called when changing the case of a UTF-8-encoded |
9fc2026f KW |
2474 | * character above the ASCII range, and the result should not |
2475 | * contain an ASCII character. */ | |
a0270393 KW |
2476 | |
2477 | UV original; /* To store the first code point of <p> */ | |
2478 | ||
2479 | /* Look at every character in the result; if any cross the | |
2480 | * boundary, the whole thing is disallowed */ | |
2481 | U8* s = ustrp; | |
2482 | U8* e = ustrp + *lenp; | |
2483 | while (s < e) { | |
2484 | if (isASCII(*s)) { | |
2485 | /* Crossed, have to return the original */ | |
2486 | original = valid_utf8_to_uvchr(p, lenp); | |
1ca267a5 | 2487 | |
9fc2026f | 2488 | /* But in these instances, there is an alternative we can |
1ca267a5 | 2489 | * return that is valid */ |
0766489e KW |
2490 | if (original == LATIN_SMALL_LETTER_SHARP_S |
2491 | #ifdef LATIN_CAPITAL_LETTER_SHARP_S /* not defined in early Unicode releases */ | |
2492 | || original == LATIN_CAPITAL_LETTER_SHARP_S | |
2493 | #endif | |
2494 | ) { | |
1ca267a5 KW |
2495 | goto return_long_s; |
2496 | } | |
9fc2026f KW |
2497 | else if (original == LATIN_SMALL_LIGATURE_LONG_S_T) { |
2498 | goto return_ligature_st; | |
2499 | } | |
74894415 KW |
2500 | #if UNICODE_MAJOR_VERSION == 3 \ |
2501 | && UNICODE_DOT_VERSION == 0 \ | |
2502 | && UNICODE_DOT_DOT_VERSION == 1 | |
2503 | ||
2504 | else if (original == LATIN_CAPITAL_LETTER_I_WITH_DOT_ABOVE) { | |
2505 | goto return_dotless_i; | |
2506 | } | |
2507 | #endif | |
a0270393 KW |
2508 | Copy(p, ustrp, *lenp, char); |
2509 | return original; | |
2510 | } | |
2511 | s += UTF8SKIP(s); | |
2512 | } | |
051a06d4 | 2513 | |
a0270393 KW |
2514 | /* Here, no characters crossed, result is ok as-is */ |
2515 | return result; | |
2516 | } | |
051a06d4 KW |
2517 | } |
2518 | ||
4a4088c4 | 2519 | /* Here, used locale rules. Convert back to UTF-8 */ |
051a06d4 KW |
2520 | if (UTF8_IS_INVARIANT(result)) { |
2521 | *ustrp = (U8) result; | |
2522 | *lenp = 1; | |
2523 | } | |
2524 | else { | |
62cb07ea KW |
2525 | *ustrp = UTF8_EIGHT_BIT_HI((U8) result); |
2526 | *(ustrp + 1) = UTF8_EIGHT_BIT_LO((U8) result); | |
051a06d4 KW |
2527 | *lenp = 2; |
2528 | } | |
2529 | ||
051a06d4 | 2530 | return result; |
1ca267a5 KW |
2531 | |
2532 | return_long_s: | |
2533 | /* Certain folds to 'ss' are prohibited by the options, but they do allow | |
2534 | * folds to a string of two of these characters. By returning this | |
2535 | * instead, then, e.g., | |
2536 | * fc("\x{1E9E}") eq fc("\x{17F}\x{17F}") | |
2537 | * works. */ | |
2538 | ||
2539 | *lenp = 2 * sizeof(LATIN_SMALL_LETTER_LONG_S_UTF8) - 2; | |
2540 | Copy(LATIN_SMALL_LETTER_LONG_S_UTF8 LATIN_SMALL_LETTER_LONG_S_UTF8, | |
2541 | ustrp, *lenp, U8); | |
2542 | return LATIN_SMALL_LETTER_LONG_S; | |
9fc2026f KW |
2543 | |
2544 | return_ligature_st: | |
2545 | /* Two folds to 'st' are prohibited by the options; instead we pick one and | |
2546 | * have the other one fold to it */ | |
2547 | ||
2548 | *lenp = sizeof(LATIN_SMALL_LIGATURE_ST_UTF8) - 1; | |
2549 | Copy(LATIN_SMALL_LIGATURE_ST_UTF8, ustrp, *lenp, U8); | |
2550 | return LATIN_SMALL_LIGATURE_ST; | |
74894415 KW |
2551 | |
2552 | #if UNICODE_MAJOR_VERSION == 3 \ | |
2553 | && UNICODE_DOT_VERSION == 0 \ | |
2554 | && UNICODE_DOT_DOT_VERSION == 1 | |
2555 | ||
2556 | return_dotless_i: | |
2557 | *lenp = sizeof(LATIN_SMALL_LETTER_DOTLESS_I_UTF8) - 1; | |
2558 | Copy(LATIN_SMALL_LETTER_DOTLESS_I_UTF8, ustrp, *lenp, U8); | |
2559 | return LATIN_SMALL_LETTER_DOTLESS_I; | |
2560 | ||
2561 | #endif | |
2562 | ||
a0ed51b3 LW |
2563 | } |
2564 | ||
711a919c | 2565 | /* Note: |
f90a9a02 | 2566 | * Returns a "swash" which is a hash described in utf8.c:Perl_swash_fetch(). |
711a919c TS |
2567 | * C<pkg> is a pointer to a package name for SWASHNEW, should be "utf8". |
2568 | * For other parameters, see utf8::SWASHNEW in lib/utf8_heavy.pl. | |
2569 | */ | |
c4a5db0c | 2570 | |
a0ed51b3 | 2571 | SV* |
7fc63493 | 2572 | Perl_swash_init(pTHX_ const char* pkg, const char* name, SV *listsv, I32 minbits, I32 none) |
a0ed51b3 | 2573 | { |
c4a5db0c KW |
2574 | PERL_ARGS_ASSERT_SWASH_INIT; |
2575 | ||
2576 | /* Returns a copy of a swash initiated by the called function. This is the | |
2577 | * public interface, and returning a copy prevents others from doing | |
2578 | * mischief on the original */ | |
2579 | ||
5d3d13d1 | 2580 | return newSVsv(_core_swash_init(pkg, name, listsv, minbits, none, NULL, NULL)); |
c4a5db0c KW |
2581 | } |
2582 | ||
2583 | SV* | |
5d3d13d1 | 2584 | Perl__core_swash_init(pTHX_ const char* pkg, const char* name, SV *listsv, I32 minbits, I32 none, SV* invlist, U8* const flags_p) |
c4a5db0c | 2585 | { |
2c1f00b9 YO |
2586 | |
2587 | /*NOTE NOTE NOTE - If you want to use "return" in this routine you MUST | |
2588 | * use the following define */ | |
2589 | ||
2590 | #define CORE_SWASH_INIT_RETURN(x) \ | |
2591 | PL_curpm= old_PL_curpm; \ | |
2592 | return x | |
2593 | ||
c4a5db0c | 2594 | /* Initialize and return a swash, creating it if necessary. It does this |
87367d5f KW |
2595 | * by calling utf8_heavy.pl in the general case. The returned value may be |
2596 | * the swash's inversion list instead if the input parameters allow it. | |
2597 | * Which is returned should be immaterial to callers, as the only | |
923b6d4e KW |
2598 | * operations permitted on a swash, swash_fetch(), _get_swash_invlist(), |
2599 | * and swash_to_invlist() handle both these transparently. | |
c4a5db0c KW |
2600 | * |
2601 | * This interface should only be used by functions that won't destroy or | |
2602 | * adversely change the swash, as doing so affects all other uses of the | |
2603 | * swash in the program; the general public should use 'Perl_swash_init' | |
2604 | * instead. | |
2605 | * | |
2606 | * pkg is the name of the package that <name> should be in. | |
2607 | * name is the name of the swash to find. Typically it is a Unicode | |
2608 | * property name, including user-defined ones | |
2609 | * listsv is a string to initialize the swash with. It must be of the form | |
2610 | * documented as the subroutine return value in | |
2611 | * L<perlunicode/User-Defined Character Properties> | |
2612 | * minbits is the number of bits required to represent each data element. | |
2613 | * It is '1' for binary properties. | |
2614 | * none I (khw) do not understand this one, but it is used only in tr///. | |
9a53f6cf | 2615 | * invlist is an inversion list to initialize the swash with (or NULL) |
83199d38 KW |
2616 | * flags_p if non-NULL is the address of various input and output flag bits |
2617 | * to the routine, as follows: ('I' means is input to the routine; | |
2618 | * 'O' means output from the routine. Only flags marked O are | |
2619 | * meaningful on return.) | |
2620 | * _CORE_SWASH_INIT_USER_DEFINED_PROPERTY indicates if the swash | |
2621 | * came from a user-defined property. (I O) | |
5d3d13d1 KW |
2622 | * _CORE_SWASH_INIT_RETURN_IF_UNDEF indicates that instead of croaking |
2623 | * when the swash cannot be located, to simply return NULL. (I) | |
87367d5f KW |
2624 | * _CORE_SWASH_INIT_ACCEPT_INVLIST indicates that the caller will accept a |
2625 | * return of an inversion list instead of a swash hash if this routine | |
2626 | * thinks that would result in faster execution of swash_fetch() later | |
2627 | * on. (I) | |
9a53f6cf KW |
2628 | * |
2629 | * Thus there are three possible inputs to find the swash: <name>, | |
2630 | * <listsv>, and <invlist>. At least one must be specified. The result | |
2631 | * will be the union of the specified ones, although <listsv>'s various | |
aabbdbda KW |
2632 | * actions can intersect, etc. what <name> gives. To avoid going out to |
2633 | * disk at all, <invlist> should specify completely what the swash should | |
2634 | * have, and <listsv> should be &PL_sv_undef and <name> should be "". | |
9a53f6cf KW |
2635 | * |
2636 | * <invlist> is only valid for binary properties */ | |
c4a5db0c | 2637 | |
2c1f00b9 YO |
2638 | PMOP *old_PL_curpm= PL_curpm; /* save away the old PL_curpm */ |
2639 | ||
c4a5db0c | 2640 | SV* retval = &PL_sv_undef; |
83199d38 | 2641 | HV* swash_hv = NULL; |
87367d5f KW |
2642 | const int invlist_swash_boundary = |
2643 | (flags_p && *flags_p & _CORE_SWASH_INIT_ACCEPT_INVLIST) | |
2644 | ? 512 /* Based on some benchmarking, but not extensive, see commit | |
2645 | message */ | |
2646 | : -1; /* Never return just an inversion list */ | |
9a53f6cf KW |
2647 | |
2648 | assert(listsv != &PL_sv_undef || strNE(name, "") || invlist); | |
2649 | assert(! invlist || minbits == 1); | |
2650 | ||
2c1f00b9 YO |
2651 | PL_curpm= NULL; /* reset PL_curpm so that we dont get confused between the regex |
2652 | that triggered the swash init and the swash init perl logic itself. | |
2653 | See perl #122747 */ | |
2654 | ||
9a53f6cf KW |
2655 | /* If data was passed in to go out to utf8_heavy to find the swash of, do |
2656 | * so */ | |
2657 | if (listsv != &PL_sv_undef || strNE(name, "")) { | |
69794297 KW |
2658 | dSP; |
2659 | const size_t pkg_len = strlen(pkg); | |
2660 | const size_t name_len = strlen(name); | |
2661 | HV * const stash = gv_stashpvn(pkg, pkg_len, 0); | |
2662 | SV* errsv_save; | |
2663 | GV *method; | |
2664 | ||
2665 | PERL_ARGS_ASSERT__CORE_SWASH_INIT; | |
2666 | ||
2667 | PUSHSTACKi(PERLSI_MAGIC); | |
ce3b816e | 2668 | ENTER; |
69794297 | 2669 | SAVEHINTS(); |
2782061f | 2670 | save_re_context(); |
650f067c JL |
2671 | /* We might get here via a subroutine signature which uses a utf8 |
2672 | * parameter name, at which point PL_subname will have been set | |
2673 | * but not yet used. */ | |
2674 | save_item(PL_subname); | |
69794297 KW |
2675 | if (PL_parser && PL_parser->error_count) |
2676 | SAVEI8(PL_parser->error_count), PL_parser->error_count = 0; | |
2677 | method = gv_fetchmeth(stash, "SWASHNEW", 8, -1); | |
4a4088c4 | 2678 | if (!method) { /* demand load UTF-8 */ |
69794297 | 2679 | ENTER; |
db2c6cb3 FC |
2680 | if ((errsv_save = GvSV(PL_errgv))) SAVEFREESV(errsv_save); |
2681 | GvSV(PL_errgv) = NULL; | |
1a419e6b | 2682 | #ifndef NO_TAINT_SUPPORT |
69794297 KW |
2683 | /* It is assumed that callers of this routine are not passing in |
2684 | * any user derived data. */ | |
2782061f DM |
2685 | /* Need to do this after save_re_context() as it will set |
2686 | * PL_tainted to 1 while saving $1 etc (see the code after getrx: | |
2687 | * in Perl_magic_get). Even line to create errsv_save can turn on | |
2688 | * PL_tainted. */ | |
284167a5 S |
2689 | SAVEBOOL(TAINT_get); |
2690 | TAINT_NOT; | |
2691 | #endif | |
69794297 KW |
2692 | Perl_load_module(aTHX_ PERL_LOADMOD_NOIMPORT, newSVpvn(pkg,pkg_len), |
2693 | NULL); | |
eed484f9 | 2694 | { |
db2c6cb3 FC |
2695 | /* Not ERRSV, as there is no need to vivify a scalar we are |
2696 | about to discard. */ | |
2697 | SV * const errsv = GvSV(PL_errgv); | |
2698 | if (!SvTRUE(errsv)) { | |
2699 | GvSV(PL_errgv) = SvREFCNT_inc_simple(errsv_save); | |
2700 | SvREFCNT_dec(errsv); | |
2701 | } | |
eed484f9 | 2702 | } |
69794297 KW |
2703 | LEAVE; |
2704 | } | |
2705 | SPAGAIN; | |
2706 | PUSHMARK(SP); | |
2707 | EXTEND(SP,5); | |
2708 | mPUSHp(pkg, pkg_len); | |
2709 | mPUSHp(name, name_len); | |
2710 | PUSHs(listsv); | |
2711 | mPUSHi(minbits); | |
2712 | mPUSHi(none); | |
2713 | PUTBACK; | |
db2c6cb3 FC |
2714 | if ((errsv_save = GvSV(PL_errgv))) SAVEFREESV(errsv_save); |
2715 | GvSV(PL_errgv) = NULL; | |
69794297 KW |
2716 | /* If we already have a pointer to the method, no need to use |
2717 | * call_method() to repeat the lookup. */ | |
c41800a8 KW |
2718 | if (method |
2719 | ? call_sv(MUTABLE_SV(method), G_SCALAR) | |
69794297 KW |
2720 | : call_sv(newSVpvs_flags("SWASHNEW", SVs_TEMP), G_SCALAR | G_METHOD)) |
2721 | { | |
2722 | retval = *PL_stack_sp--; | |
2723 | SvREFCNT_inc(retval); | |
2724 | } | |
eed484f9 | 2725 | { |
db2c6cb3 FC |
2726 | /* Not ERRSV. See above. */ |
2727 | SV * const errsv = GvSV(PL_errgv); | |
2728 | if (!SvTRUE(errsv)) { | |
2729 | GvSV(PL_errgv) = SvREFCNT_inc_simple(errsv_save); | |
2730 | SvREFCNT_dec(errsv); | |
2731 | } | |
eed484f9 | 2732 | } |
ce3b816e | 2733 | LEAVE; |
69794297 KW |
2734 | POPSTACK; |
2735 | if (IN_PERL_COMPILETIME) { | |
2736 | CopHINTS_set(PL_curcop, PL_hints); | |
2737 | } | |
2738 | if (!SvROK(retval) || SvTYPE(SvRV(retval)) != SVt_PVHV) { | |
d95e4a00 | 2739 | if (SvPOK(retval)) { |
69794297 KW |
2740 | |
2741 | /* If caller wants to handle missing properties, let them */ | |
5d3d13d1 | 2742 | if (flags_p && *flags_p & _CORE_SWASH_INIT_RETURN_IF_UNDEF) { |
2c1f00b9 | 2743 | CORE_SWASH_INIT_RETURN(NULL); |
69794297 KW |
2744 | } |
2745 | Perl_croak(aTHX_ | |
2746 | "Can't find Unicode property definition \"%"SVf"\"", | |
2747 | SVfARG(retval)); | |
a25b5927 | 2748 | NOT_REACHED; /* NOTREACHED */ |
d95e4a00 | 2749 | } |
69794297 | 2750 | } |
9a53f6cf | 2751 | } /* End of calling the module to find the swash */ |
36eb48b4 | 2752 | |
83199d38 KW |
2753 | /* If this operation fetched a swash, and we will need it later, get it */ |
2754 | if (retval != &PL_sv_undef | |
2755 | && (minbits == 1 || (flags_p | |
2756 | && ! (*flags_p | |
2757 | & _CORE_SWASH_INIT_USER_DEFINED_PROPERTY)))) | |
2758 | { | |
2759 | swash_hv = MUTABLE_HV(SvRV(retval)); | |
2760 | ||
2761 | /* If we don't already know that there is a user-defined component to | |
2762 | * this swash, and the user has indicated they wish to know if there is | |
2763 | * one (by passing <flags_p>), find out */ | |
2764 | if (flags_p && ! (*flags_p & _CORE_SWASH_INIT_USER_DEFINED_PROPERTY)) { | |
2765 | SV** user_defined = hv_fetchs(swash_hv, "USER_DEFINED", FALSE); | |
2766 | if (user_defined && SvUV(*user_defined)) { | |
2767 | *flags_p |= _CORE_SWASH_INIT_USER_DEFINED_PROPERTY; | |
2768 | } | |
2769 | } | |
2770 | } | |
2771 | ||
36eb48b4 KW |
2772 | /* Make sure there is an inversion list for binary properties */ |
2773 | if (minbits == 1) { | |
2774 | SV** swash_invlistsvp = NULL; | |
2775 | SV* swash_invlist = NULL; | |
9a53f6cf | 2776 | bool invlist_in_swash_is_valid = FALSE; |
02c85471 FC |
2777 | bool swash_invlist_unclaimed = FALSE; /* whether swash_invlist has |
2778 | an unclaimed reference count */ | |
36eb48b4 | 2779 | |
9a53f6cf | 2780 | /* If this operation fetched a swash, get its already existing |
83199d38 | 2781 | * inversion list, or create one for it */ |
36eb48b4 | 2782 | |
83199d38 | 2783 | if (swash_hv) { |
5c9f4bd2 | 2784 | swash_invlistsvp = hv_fetchs(swash_hv, "V", FALSE); |
9a53f6cf KW |
2785 | if (swash_invlistsvp) { |
2786 | swash_invlist = *swash_invlistsvp; | |
2787 | invlist_in_swash_is_valid = TRUE; | |
2788 | } | |
2789 | else { | |
36eb48b4 | 2790 | swash_invlist = _swash_to_invlist(retval); |
02c85471 | 2791 | swash_invlist_unclaimed = TRUE; |
9a53f6cf KW |
2792 | } |
2793 | } | |
2794 | ||
2795 | /* If an inversion list was passed in, have to include it */ | |
2796 | if (invlist) { | |
2797 | ||
2798 | /* Any fetched swash will by now have an inversion list in it; | |
2799 | * otherwise <swash_invlist> will be NULL, indicating that we | |
2800 | * didn't fetch a swash */ | |
2801 | if (swash_invlist) { | |
2802 | ||
2803 | /* Add the passed-in inversion list, which invalidates the one | |
2804 | * already stored in the swash */ | |
2805 | invlist_in_swash_is_valid = FALSE; | |
2806 | _invlist_union(invlist, swash_invlist, &swash_invlist); | |
2807 | } | |
2808 | else { | |
2809 | ||
87367d5f KW |
2810 | /* Here, there is no swash already. Set up a minimal one, if |
2811 | * we are going to return a swash */ | |
2812 | if ((int) _invlist_len(invlist) > invlist_swash_boundary) { | |
971d486f | 2813 | swash_hv = newHV(); |
4aca0fe6 | 2814 | retval = newRV_noinc(MUTABLE_SV(swash_hv)); |
87367d5f | 2815 | } |
9a53f6cf KW |
2816 | swash_invlist = invlist; |
2817 | } | |
9a53f6cf KW |
2818 | } |
2819 | ||
2820 | /* Here, we have computed the union of all the passed-in data. It may | |
2821 | * be that there was an inversion list in the swash which didn't get | |
538e84ed | 2822 | * touched; otherwise save the computed one */ |
87367d5f KW |
2823 | if (! invlist_in_swash_is_valid |
2824 | && (int) _invlist_len(swash_invlist) > invlist_swash_boundary) | |
2825 | { | |
5c9f4bd2 | 2826 | if (! hv_stores(MUTABLE_HV(SvRV(retval)), "V", swash_invlist)) |
69794297 KW |
2827 | { |
2828 | Perl_croak(aTHX_ "panic: hv_store() unexpectedly failed"); | |
2829 | } | |
cc34d8c5 FC |
2830 | /* We just stole a reference count. */ |
2831 | if (swash_invlist_unclaimed) swash_invlist_unclaimed = FALSE; | |
2832 | else SvREFCNT_inc_simple_void_NN(swash_invlist); | |
9a53f6cf | 2833 | } |
87367d5f | 2834 | |
dbfdbd26 KW |
2835 | SvREADONLY_on(swash_invlist); |
2836 | ||
c41800a8 | 2837 | /* Use the inversion list stand-alone if small enough */ |
87367d5f KW |
2838 | if ((int) _invlist_len(swash_invlist) <= invlist_swash_boundary) { |
2839 | SvREFCNT_dec(retval); | |
02c85471 FC |
2840 | if (!swash_invlist_unclaimed) |
2841 | SvREFCNT_inc_simple_void_NN(swash_invlist); | |
2842 | retval = newRV_noinc(swash_invlist); | |
87367d5f | 2843 | } |
36eb48b4 KW |
2844 | } |
2845 | ||
2c1f00b9 YO |
2846 | CORE_SWASH_INIT_RETURN(retval); |
2847 | #undef CORE_SWASH_INIT_RETURN | |
a0ed51b3 LW |
2848 | } |
2849 | ||
035d37be JH |
2850 | |
2851 | /* This API is wrong for special case conversions since we may need to | |
2852 | * return several Unicode characters for a single Unicode character | |
2853 | * (see lib/unicore/SpecCase.txt) The SWASHGET in lib/utf8_heavy.pl is | |
2854 | * the lower-level routine, and it is similarly broken for returning | |
38684baa | 2855 | * multiple values. --jhi |
b9992569 | 2856 | * For those, you should use S__to_utf8_case() instead */ |
b0e3252e | 2857 | /* Now SWASHGET is recasted into S_swatch_get in this file. */ |
680c470c TS |
2858 | |
2859 | /* Note: | |
2860 | * Returns the value of property/mapping C<swash> for the first character | |
2861 | * of the string C<ptr>. If C<do_utf8> is true, the string C<ptr> is | |
4a4088c4 | 2862 | * assumed to be in well-formed UTF-8. If C<do_utf8> is false, the string C<ptr> |
3d0f8846 | 2863 | * is assumed to be in native 8-bit encoding. Caches the swatch in C<swash>. |
af2af982 KW |
2864 | * |
2865 | * A "swash" is a hash which contains initially the keys/values set up by | |
2866 | * SWASHNEW. The purpose is to be able to completely represent a Unicode | |
2867 | * property for all possible code points. Things are stored in a compact form | |
2868 | * (see utf8_heavy.pl) so that calculation is required to find the actual | |
2869 | * property value for a given code point. As code points are looked up, new | |
2870 | * key/value pairs are added to the hash, so that the calculation doesn't have | |
2871 | * to ever be re-done. Further, each calculation is done, not just for the | |
2872 | * desired one, but for a whole block of code points adjacent to that one. | |
2873 | * For binary properties on ASCII machines, the block is usually for 64 code | |
2874 | * points, starting with a code point evenly divisible by 64. Thus if the | |
2875 | * property value for code point 257 is requested, the code goes out and | |
2876 | * calculates the property values for all 64 code points between 256 and 319, | |
2877 | * and stores these as a single 64-bit long bit vector, called a "swatch", | |
2878 | * under the key for code point 256. The key is the UTF-8 encoding for code | |
2879 | * point 256, minus the final byte. Thus, if the length of the UTF-8 encoding | |
2880 | * for a code point is 13 bytes, the key will be 12 bytes long. If the value | |
2881 | * for code point 258 is then requested, this code realizes that it would be | |
2882 | * stored under the key for 256, and would find that value and extract the | |
2883 | * relevant bit, offset from 256. | |
2884 | * | |
2885 | * Non-binary properties are stored in as many bits as necessary to represent | |
2886 | * their values (32 currently, though the code is more general than that), not | |
fc273927 | 2887 | * as single bits, but the principle is the same: the value for each key is a |
af2af982 KW |
2888 | * vector that encompasses the property values for all code points whose UTF-8 |
2889 | * representations are represented by the key. That is, for all code points | |
2890 | * whose UTF-8 representations are length N bytes, and the key is the first N-1 | |
2891 | * bytes of that. | |
680c470c | 2892 | */ |
a0ed51b3 | 2893 | UV |
680c470c | 2894 | Perl_swash_fetch(pTHX_ SV *swash, const U8 *ptr, bool do_utf8) |
a0ed51b3 | 2895 | { |
ef8f7699 | 2896 | HV *const hv = MUTABLE_HV(SvRV(swash)); |
3568d838 JH |
2897 | U32 klen; |
2898 | U32 off; | |
9b56a019 | 2899 | STRLEN slen = 0; |
7d85a32c | 2900 | STRLEN needents; |
cfd0369c | 2901 | const U8 *tmps = NULL; |
979f2922 | 2902 | SV *swatch; |
08fb1ac5 | 2903 | const U8 c = *ptr; |
3568d838 | 2904 | |
7918f24d NC |
2905 | PERL_ARGS_ASSERT_SWASH_FETCH; |
2906 | ||
87367d5f KW |
2907 | /* If it really isn't a hash, it isn't really swash; must be an inversion |
2908 | * list */ | |
2909 | if (SvTYPE(hv) != SVt_PVHV) { | |
2910 | return _invlist_contains_cp((SV*)hv, | |
2911 | (do_utf8) | |
2912 | ? valid_utf8_to_uvchr(ptr, NULL) | |
2913 | : c); | |
2914 | } | |
2915 | ||
08fb1ac5 KW |
2916 | /* We store the values in a "swatch" which is a vec() value in a swash |
2917 | * hash. Code points 0-255 are a single vec() stored with key length | |
2918 | * (klen) 0. All other code points have a UTF-8 representation | |
2919 | * 0xAA..0xYY,0xZZ. A vec() is constructed containing all of them which | |
2920 | * share 0xAA..0xYY, which is the key in the hash to that vec. So the key | |
2921 | * length for them is the length of the encoded char - 1. ptr[klen] is the | |
2922 | * final byte in the sequence representing the character */ | |
2923 | if (!do_utf8 || UTF8_IS_INVARIANT(c)) { | |
2924 | klen = 0; | |
2925 | needents = 256; | |
2926 | off = c; | |
3568d838 | 2927 | } |
08fb1ac5 KW |
2928 | else if (UTF8_IS_DOWNGRADEABLE_START(c)) { |
2929 | klen = 0; | |
2930 | needents = 256; | |
a62b247b | 2931 | off = EIGHT_BIT_UTF8_TO_NATIVE(c, *(ptr + 1)); |
979f2922 TS |
2932 | } |
2933 | else { | |
08fb1ac5 KW |
2934 | klen = UTF8SKIP(ptr) - 1; |
2935 | ||
2936 | /* Each vec() stores 2**UTF_ACCUMULATION_SHIFT values. The offset into | |
2937 | * the vec is the final byte in the sequence. (In EBCDIC this is | |
2938 | * converted to I8 to get consecutive values.) To help you visualize | |
2939 | * all this: | |
2940 | * Straight 1047 After final byte | |
2941 | * UTF-8 UTF-EBCDIC I8 transform | |
2942 | * U+0400: \xD0\x80 \xB8\x41\x41 \xB8\x41\xA0 | |
2943 | * U+0401: \xD0\x81 \xB8\x41\x42 \xB8\x41\xA1 | |
2944 | * ... | |
2945 | * U+0409: \xD0\x89 \xB8\x41\x4A \xB8\x41\xA9 | |
2946 | * U+040A: \xD0\x8A \xB8\x41\x51 \xB8\x41\xAA | |
2947 | * ... | |
2948 | * U+0412: \xD0\x92 \xB8\x41\x59 \xB8\x41\xB2 | |
2949 | * U+0413: \xD0\x93 \xB8\x41\x62 \xB8\x41\xB3 | |
2950 | * ... | |
2951 | * U+041B: \xD0\x9B \xB8\x41\x6A \xB8\x41\xBB | |
2952 | * U+041C: \xD0\x9C \xB8\x41\x70 \xB8\x41\xBC | |
2953 | * ... | |
2954 | * U+041F: \xD0\x9F \xB8\x41\x73 \xB8\x41\xBF | |
2955 | * U+0420: \xD0\xA0 \xB8\x42\x41 \xB8\x42\x41 | |
2956 | * | |
2957 | * (There are no discontinuities in the elided (...) entries.) | |
2958 | * The UTF-8 key for these 33 code points is '\xD0' (which also is the | |
2959 | * key for the next 31, up through U+043F, whose UTF-8 final byte is | |
2960 | * \xBF). Thus in UTF-8, each key is for a vec() for 64 code points. | |
2961 | * The final UTF-8 byte, which ranges between \x80 and \xBF, is an | |
2962 | * index into the vec() swatch (after subtracting 0x80, which we | |
2963 | * actually do with an '&'). | |
2964 | * In UTF-EBCDIC, each key is for a 32 code point vec(). The first 32 | |
2965 | * code points above have key '\xB8\x41'. The final UTF-EBCDIC byte has | |
2966 | * dicontinuities which go away by transforming it into I8, and we | |
2967 | * effectively subtract 0xA0 to get the index. */ | |
979f2922 | 2968 | needents = (1 << UTF_ACCUMULATION_SHIFT); |
bc3632a8 | 2969 | off = NATIVE_UTF8_TO_I8(ptr[klen]) & UTF_CONTINUATION_MASK; |
979f2922 | 2970 | } |
7d85a32c | 2971 | |
a0ed51b3 | 2972 | /* |
4a4088c4 | 2973 | * This single-entry cache saves about 1/3 of the UTF-8 overhead in test |
a0ed51b3 LW |
2974 | * suite. (That is, only 7-8% overall over just a hash cache. Still, |
2975 | * it's nothing to sniff at.) Pity we usually come through at least | |
2976 | * two function calls to get here... | |
2977 | * | |
2978 | * NB: this code assumes that swatches are never modified, once generated! | |
2979 | */ | |
2980 | ||
3568d838 | 2981 | if (hv == PL_last_swash_hv && |
a0ed51b3 | 2982 | klen == PL_last_swash_klen && |
27da23d5 | 2983 | (!klen || memEQ((char *)ptr, (char *)PL_last_swash_key, klen)) ) |
a0ed51b3 LW |
2984 | { |
2985 | tmps = PL_last_swash_tmps; | |
2986 | slen = PL_last_swash_slen; | |
2987 | } | |
2988 | else { | |
2989 | /* Try our second-level swatch cache, kept in a hash. */ | |
e1ec3a88 | 2990 | SV** svp = hv_fetch(hv, (const char*)ptr, klen, FALSE); |
a0ed51b3 | 2991 | |
b0e3252e | 2992 | /* If not cached, generate it via swatch_get */ |
979f2922 | 2993 | if (!svp || !SvPOK(*svp) |
08fb1ac5 KW |
2994 | || !(tmps = (const U8*)SvPV_const(*svp, slen))) |
2995 | { | |
2996 | if (klen) { | |
2997 | const UV code_point = valid_utf8_to_uvchr(ptr, NULL); | |
2998 | swatch = swatch_get(swash, | |
2999 | code_point & ~((UV)needents - 1), | |
3000 | needents); | |
3001 | } | |
3002 | else { /* For the first 256 code points, the swatch has a key of | |
3003 | length 0 */ | |
3004 | swatch = swatch_get(swash, 0, needents); | |
3005 | } | |
979f2922 | 3006 | |
923e4eb5 | 3007 | if (IN_PERL_COMPILETIME) |
623e6609 | 3008 | CopHINTS_set(PL_curcop, PL_hints); |
a0ed51b3 | 3009 | |
979f2922 | 3010 | svp = hv_store(hv, (const char *)ptr, klen, swatch, 0); |
a0ed51b3 | 3011 | |
979f2922 TS |
3012 | if (!svp || !(tmps = (U8*)SvPV(*svp, slen)) |
3013 | || (slen << 3) < needents) | |
5637ef5b NC |
3014 | Perl_croak(aTHX_ "panic: swash_fetch got improper swatch, " |
3015 | "svp=%p, tmps=%p, slen=%"UVuf", needents=%"UVuf, | |
3016 | svp, tmps, (UV)slen, (UV)needents); | |
a0ed51b3 LW |
3017 | } |
3018 | ||
3019 | PL_last_swash_hv = hv; | |
16d8f38a | 3020 | assert(klen <= sizeof(PL_last_swash_key)); |
eac04b2e | 3021 | PL_last_swash_klen = (U8)klen; |
cfd0369c NC |
3022 | /* FIXME change interpvar.h? */ |
3023 | PL_last_swash_tmps = (U8 *) tmps; | |
a0ed51b3 LW |
3024 | PL_last_swash_slen = slen; |
3025 | if (klen) | |
3026 | Copy(ptr, PL_last_swash_key, klen, U8); | |
3027 | } | |
3028 | ||
9faf8d75 | 3029 | switch ((int)((slen << 3) / needents)) { |
a0ed51b3 | 3030 | case 1: |
e7aca353 | 3031 | return ((UV) tmps[off >> 3] & (1 << (off & 7))) != 0; |
a0ed51b3 | 3032 | case 8: |
e7aca353 | 3033 | return ((UV) tmps[off]); |
a0ed51b3 LW |
3034 | case 16: |
3035 | off <<= 1; | |
e7aca353 JH |
3036 | return |
3037 | ((UV) tmps[off ] << 8) + | |
3038 | ((UV) tmps[off + 1]); | |
a0ed51b3 LW |
3039 | case 32: |
3040 | off <<= 2; | |
e7aca353 JH |
3041 | return |
3042 | ((UV) tmps[off ] << 24) + | |
3043 | ((UV) tmps[off + 1] << 16) + | |
3044 | ((UV) tmps[off + 2] << 8) + | |
3045 | ((UV) tmps[off + 3]); | |
a0ed51b3 | 3046 | } |
5637ef5b NC |
3047 | Perl_croak(aTHX_ "panic: swash_fetch got swatch of unexpected bit width, " |
3048 | "slen=%"UVuf", needents=%"UVuf, (UV)slen, (UV)needents); | |
670f1322 | 3049 | NORETURN_FUNCTION_END; |
a0ed51b3 | 3050 | } |
2b9d42f0 | 3051 | |
319009ee KW |
3052 | /* Read a single line of the main body of the swash input text. These are of |
3053 | * the form: | |
3054 | * 0053 0056 0073 | |
3055 | * where each number is hex. The first two numbers form the minimum and | |
3056 | * maximum of a range, and the third is the value associated with the range. | |
3057 | * Not all swashes should have a third number | |
3058 | * | |
3059 | * On input: l points to the beginning of the line to be examined; it points | |
3060 | * to somewhere in the string of the whole input text, and is | |
3061 | * terminated by a \n or the null string terminator. | |
3062 | * lend points to the null terminator of that string | |
3063 | * wants_value is non-zero if the swash expects a third number | |
3064 | * typestr is the name of the swash's mapping, like 'ToLower' | |
3065 | * On output: *min, *max, and *val are set to the values read from the line. | |
3066 | * returns a pointer just beyond the line examined. If there was no | |
3067 | * valid min number on the line, returns lend+1 | |
3068 | */ | |
3069 | ||
3070 | STATIC U8* | |
3071 | S_swash_scan_list_line(pTHX_ U8* l, U8* const lend, UV* min, UV* max, UV* val, | |
3072 | const bool wants_value, const U8* const typestr) | |
3073 | { | |
3074 | const int typeto = typestr[0] == 'T' && typestr[1] == 'o'; | |
3075 | STRLEN numlen; /* Length of the number */ | |
02470786 KW |
3076 | I32 flags = PERL_SCAN_SILENT_ILLDIGIT |
3077 | | PERL_SCAN_DISALLOW_PREFIX | |
3078 | | PERL_SCAN_SILENT_NON_PORTABLE; | |
319009ee KW |
3079 | |
3080 | /* nl points to the next \n in the scan */ | |
3081 | U8* const nl = (U8*)memchr(l, '\n', lend - l); | |
3082 | ||
95543e92 KW |
3083 | PERL_ARGS_ASSERT_SWASH_SCAN_LIST_LINE; |
3084 | ||
319009ee KW |
3085 | /* Get the first number on the line: the range minimum */ |
3086 | numlen = lend - l; | |
3087 | *min = grok_hex((char *)l, &numlen, &flags, NULL); | |
c88850db | 3088 | *max = *min; /* So can never return without setting max */ |
319009ee KW |
3089 | if (numlen) /* If found a hex number, position past it */ |
3090 | l += numlen; | |
3091 | else if (nl) { /* Else, go handle next line, if any */ | |
3092 | return nl + 1; /* 1 is length of "\n" */ | |
3093 | } | |
3094 | else { /* Else, no next line */ | |
3095 | return lend + 1; /* to LIST's end at which \n is not found */ | |
3096 | } | |
3097 | ||
3098 | /* The max range value follows, separated by a BLANK */ | |
3099 | if (isBLANK(*l)) { | |
3100 | ++l; | |
02470786 KW |
3101 | flags = PERL_SCAN_SILENT_ILLDIGIT |
3102 | | PERL_SCAN_DISALLOW_PREFIX | |
3103 | | PERL_SCAN_SILENT_NON_PORTABLE; | |
319009ee KW |
3104 | numlen = lend - l; |
3105 | *max = grok_hex((char *)l, &numlen, &flags, NULL); | |
3106 | if (numlen) | |
3107 | l += numlen; | |
3108 | else /* If no value here, it is a single element range */ | |
3109 | *max = *min; | |
3110 | ||
3111 | /* Non-binary tables have a third entry: what the first element of the | |
24303724 | 3112 | * range maps to. The map for those currently read here is in hex */ |
319009ee KW |
3113 | if (wants_value) { |
3114 | if (isBLANK(*l)) { | |
3115 | ++l; | |
f2a7d0fc KW |
3116 | flags = PERL_SCAN_SILENT_ILLDIGIT |
3117 | | PERL_SCAN_DISALLOW_PREFIX | |
3118 | | PERL_SCAN_SILENT_NON_PORTABLE; | |
3119 | numlen = lend - l; | |
3120 | *val = grok_hex((char *)l, &numlen, &flags, NULL); | |
3121 | if (numlen) | |
3122 | l += numlen; | |
3123 | else | |
3124 | *val = 0; | |
319009ee KW |
3125 | } |
3126 | else { | |
3127 | *val = 0; | |
3128 | if (typeto) { | |
dcbac5bb | 3129 | /* diag_listed_as: To%s: illegal mapping '%s' */ |
319009ee KW |
3130 | Perl_croak(aTHX_ "%s: illegal mapping '%s'", |
3131 | typestr, l); | |
3132 | } | |
3133 | } | |
3134 | } | |
3135 | else | |
3136 | *val = 0; /* bits == 1, then any val should be ignored */ | |
3137 | } | |
3138 | else { /* Nothing following range min, should be single element with no | |
3139 | mapping expected */ | |
319009ee KW |
3140 | if (wants_value) { |
3141 | *val = 0; | |
3142 | if (typeto) { | |
dcbac5bb | 3143 | /* diag_listed_as: To%s: illegal mapping '%s' */ |
319009ee KW |
3144 | Perl_croak(aTHX_ "%s: illegal mapping '%s'", typestr, l); |
3145 | } | |
3146 | } | |
3147 | else | |
3148 | *val = 0; /* bits == 1, then val should be ignored */ | |
3149 | } | |
3150 | ||
3151 | /* Position to next line if any, or EOF */ | |
3152 | if (nl) | |
3153 | l = nl + 1; | |
3154 | else | |
3155 | l = lend; | |
3156 | ||
3157 | return l; | |
3158 | } | |
3159 | ||
979f2922 TS |
3160 | /* Note: |
3161 | * Returns a swatch (a bit vector string) for a code point sequence | |
3162 | * that starts from the value C<start> and comprises the number C<span>. | |
3163 | * A C<swash> must be an object created by SWASHNEW (see lib/utf8_heavy.pl). | |
3164 | * Should be used via swash_fetch, which will cache the swatch in C<swash>. | |
3165 | */ | |
3166 | STATIC SV* | |
b0e3252e | 3167 | S_swatch_get(pTHX_ SV* swash, UV start, UV span) |
979f2922 TS |
3168 | { |
3169 | SV *swatch; | |
77f9f126 | 3170 | U8 *l, *lend, *x, *xend, *s, *send; |
979f2922 | 3171 | STRLEN lcur, xcur, scur; |
ef8f7699 | 3172 | HV *const hv = MUTABLE_HV(SvRV(swash)); |
5c9f4bd2 | 3173 | SV** const invlistsvp = hv_fetchs(hv, "V", FALSE); |
36eb48b4 | 3174 | |
88d45d28 KW |
3175 | SV** listsvp = NULL; /* The string containing the main body of the table */ |
3176 | SV** extssvp = NULL; | |
3177 | SV** invert_it_svp = NULL; | |
3178 | U8* typestr = NULL; | |
786861f5 KW |
3179 | STRLEN bits; |
3180 | STRLEN octets; /* if bits == 1, then octets == 0 */ | |
3181 | UV none; | |
3182 | UV end = start + span; | |
972dd592 | 3183 | |
36eb48b4 | 3184 | if (invlistsvp == NULL) { |
786861f5 KW |
3185 | SV** const bitssvp = hv_fetchs(hv, "BITS", FALSE); |
3186 | SV** const nonesvp = hv_fetchs(hv, "NONE", FALSE); | |
3187 | SV** const typesvp = hv_fetchs(hv, "TYPE", FALSE); | |
3188 | extssvp = hv_fetchs(hv, "EXTRAS", FALSE); | |
3189 | listsvp = hv_fetchs(hv, "LIST", FALSE); | |
3190 | invert_it_svp = hv_fetchs(hv, "INVERT_IT", FALSE); | |
3191 | ||
3192 | bits = SvUV(*bitssvp); | |
3193 | none = SvUV(*nonesvp); | |
3194 | typestr = (U8*)SvPV_nolen(*typesvp); | |
3195 | } | |
36eb48b4 KW |
3196 | else { |
3197 | bits = 1; | |
3198 | none = 0; | |
3199 | } | |
786861f5 | 3200 | octets = bits >> 3; /* if bits == 1, then octets == 0 */ |
979f2922 | 3201 | |
b0e3252e | 3202 | PERL_ARGS_ASSERT_SWATCH_GET; |
7918f24d | 3203 | |
979f2922 | 3204 | if (bits != 1 && bits != 8 && bits != 16 && bits != 32) { |
b0e3252e | 3205 | Perl_croak(aTHX_ "panic: swatch_get doesn't expect bits %"UVuf, |
660a4616 | 3206 | (UV)bits); |
979f2922 TS |
3207 | } |
3208 | ||
84ea5ef6 KW |
3209 | /* If overflowed, use the max possible */ |
3210 | if (end < start) { | |
3211 | end = UV_MAX; | |
3212 | span = end - start; | |
3213 | } | |
3214 | ||
979f2922 | 3215 | /* create and initialize $swatch */ |
979f2922 | 3216 | scur = octets ? (span * octets) : (span + 7) / 8; |
e524fe40 NC |
3217 | swatch = newSV(scur); |
3218 | SvPOK_on(swatch); | |
979f2922 TS |
3219 | s = (U8*)SvPVX(swatch); |
3220 | if (octets && none) { | |
0bd48802 | 3221 | const U8* const e = s + scur; |
979f2922 TS |
3222 | while (s < e) { |
3223 | if (bits == 8) | |
3224 | *s++ = (U8)(none & 0xff); | |
3225 | else if (bits == 16) { | |
3226 | *s++ = (U8)((none >> 8) & 0xff); | |
3227 | *s++ = (U8)( none & 0xff); | |
3228 | } | |
3229 | else if (bits == 32) { | |
3230 | *s++ = (U8)((none >> 24) & 0xff); | |
3231 | *s++ = (U8)((none >> 16) & 0xff); | |
3232 | *s++ = (U8)((none >> 8) & 0xff); | |
3233 | *s++ = (U8)( none & 0xff); | |
3234 | } | |
3235 | } | |
3236 | *s = '\0'; | |
3237 | } | |
3238 | else { | |
3239 | (void)memzero((U8*)s, scur + 1); | |
3240 | } | |
3241 | SvCUR_set(swatch, scur); | |
3242 | s = (U8*)SvPVX(swatch); | |
3243 | ||
36eb48b4 KW |
3244 | if (invlistsvp) { /* If has an inversion list set up use that */ |
3245 | _invlist_populate_swatch(*invlistsvp, start, end, s); | |
3246 | return swatch; | |
3247 | } | |
3248 | ||
3249 | /* read $swash->{LIST} */ | |
979f2922 TS |
3250 | l = (U8*)SvPV(*listsvp, lcur); |
3251 | lend = l + lcur; | |
3252 | while (l < lend) { | |
8ed25d53 | 3253 | UV min, max, val, upper; |
95543e92 KW |
3254 | l = swash_scan_list_line(l, lend, &min, &max, &val, |
3255 | cBOOL(octets), typestr); | |
319009ee | 3256 | if (l > lend) { |
979f2922 TS |
3257 | break; |
3258 | } | |
3259 | ||
972dd592 | 3260 | /* If looking for something beyond this range, go try the next one */ |
979f2922 TS |
3261 | if (max < start) |
3262 | continue; | |
3263 | ||
8ed25d53 KW |
3264 | /* <end> is generally 1 beyond where we want to set things, but at the |
3265 | * platform's infinity, where we can't go any higher, we want to | |
3266 | * include the code point at <end> */ | |
3267 | upper = (max < end) | |
3268 | ? max | |
3269 | : (max != UV_MAX || end != UV_MAX) | |
3270 | ? end - 1 | |
3271 | : end; | |
3272 | ||
979f2922 | 3273 | if (octets) { |
35da51f7 | 3274 | UV key; |
979f2922 TS |
3275 | if (min < start) { |
3276 | if (!none || val < none) { | |
3277 | val += start - min; | |
3278 | } | |
3279 | min = start; | |
3280 | } | |
8ed25d53 | 3281 | for (key = min; key <= upper; key++) { |
979f2922 | 3282 | STRLEN offset; |
979f2922 TS |
3283 | /* offset must be non-negative (start <= min <= key < end) */ |
3284 | offset = octets * (key - start); | |
3285 | if (bits == 8) | |
3286 | s[offset] = (U8)(val & 0xff); | |
3287 | else if (bits == 16) { | |
3288 | s[offset ] = (U8)((val >> 8) & 0xff); | |
3289 | s[offset + 1] = (U8)( val & 0xff); | |
3290 | } | |
3291 | else if (bits == 32) { | |
3292 | s[offset ] = (U8)((val >> 24) & 0xff); | |
3293 | s[offset + 1] = (U8)((val >> 16) & 0xff); | |
3294 | s[offset + 2] = (U8)((val >> 8) & 0xff); | |
3295 | s[offset + 3] = (U8)( val & 0xff); | |
3296 | } | |
3297 | ||
3298 | if (!none || val < none) | |
3299 | ++val; | |
3300 | } | |
3301 | } | |
711a919c | 3302 | else { /* bits == 1, then val should be ignored */ |
35da51f7 | 3303 | UV key; |
979f2922 TS |
3304 | if (min < start) |
3305 | min = start; | |
6cb05c12 | 3306 | |
8ed25d53 | 3307 | for (key = min; key <= upper; key++) { |
0bd48802 | 3308 | const STRLEN offset = (STRLEN)(key - start); |
979f2922 TS |
3309 | s[offset >> 3] |= 1 << (offset & 7); |
3310 | } | |
3311 | } | |
3312 | } /* while */ | |
979f2922 | 3313 | |
9479a769 | 3314 | /* Invert if the data says it should be. Assumes that bits == 1 */ |
77f9f126 | 3315 | if (invert_it_svp && SvUV(*invert_it_svp)) { |
0bda3001 KW |
3316 | |
3317 | /* Unicode properties should come with all bits above PERL_UNICODE_MAX | |
3318 | * be 0, and their inversion should also be 0, as we don't succeed any | |
3319 | * Unicode property matches for non-Unicode code points */ | |
3320 | if (start <= PERL_UNICODE_MAX) { | |
3321 | ||
3322 | /* The code below assumes that we never cross the | |
3323 | * Unicode/above-Unicode boundary in a range, as otherwise we would | |
3324 | * have to figure out where to stop flipping the bits. Since this | |
3325 | * boundary is divisible by a large power of 2, and swatches comes | |
3326 | * in small powers of 2, this should be a valid assumption */ | |
3327 | assert(start + span - 1 <= PERL_UNICODE_MAX); | |
3328 | ||
507a8485 KW |
3329 | send = s + scur; |
3330 | while (s < send) { | |
3331 | *s = ~(*s); | |
3332 | s++; | |
3333 | } | |
0bda3001 | 3334 | } |
77f9f126 KW |
3335 | } |
3336 | ||
d73c39c5 KW |
3337 | /* read $swash->{EXTRAS} |
3338 | * This code also copied to swash_to_invlist() below */ | |
979f2922 TS |
3339 | x = (U8*)SvPV(*extssvp, xcur); |
3340 | xend = x + xcur; | |
3341 | while (x < xend) { | |
3342 | STRLEN namelen; | |
3343 | U8 *namestr; | |
3344 | SV** othersvp; | |
3345 | HV* otherhv; | |
3346 | STRLEN otherbits; | |
3347 | SV **otherbitssvp, *other; | |
711a919c | 3348 | U8 *s, *o, *nl; |
979f2922 TS |
3349 | STRLEN slen, olen; |
3350 | ||
35da51f7 | 3351 | const U8 opc = *x++; |
979f2922 TS |
3352 | if (opc == '\n') |
3353 | continue; | |
3354 | ||
3355 | nl = (U8*)memchr(x, '\n', xend - x); | |
3356 | ||
3357 | if (opc != '-' && opc != '+' && opc != '!' && opc != '&') { | |
3358 | if (nl) { | |
3359 | x = nl + 1; /* 1 is length of "\n" */ | |
3360 | continue; | |
3361 | } | |
3362 | else { | |
3363 | x = xend; /* to EXTRAS' end at which \n is not found */ | |
3364 | break; | |
3365 | } | |
3366 | } | |
3367 | ||
3368 | namestr = x; | |
3369 | if (nl) { | |
3370 | namelen = nl - namestr; | |
3371 | x = nl + 1; | |
3372 | } | |
3373 | else { | |
3374 | namelen = xend - namestr; | |
3375 | x = xend; | |
3376 | } | |
3377 | ||
3378 | othersvp = hv_fetch(hv, (char *)namestr, namelen, FALSE); | |
ef8f7699 | 3379 | otherhv = MUTABLE_HV(SvRV(*othersvp)); |
017a3ce5 | 3380 | otherbitssvp = hv_fetchs(otherhv, "BITS", FALSE); |
979f2922 TS |
3381 | otherbits = (STRLEN)SvUV(*otherbitssvp); |
3382 | if (bits < otherbits) | |
5637ef5b NC |
3383 | Perl_croak(aTHX_ "panic: swatch_get found swatch size mismatch, " |
3384 | "bits=%"UVuf", otherbits=%"UVuf, (UV)bits, (UV)otherbits); | |
979f2922 TS |
3385 | |
3386 | /* The "other" swatch must be destroyed after. */ | |
b0e3252e | 3387 | other = swatch_get(*othersvp, start, span); |
979f2922 TS |
3388 | o = (U8*)SvPV(other, olen); |
3389 | ||
3390 | if (!olen) | |
b0e3252e | 3391 | Perl_croak(aTHX_ "panic: swatch_get got improper swatch"); |
979f2922 TS |
3392 | |
3393 | s = (U8*)SvPV(swatch, slen); | |
3394 | if (bits == 1 && otherbits == 1) { | |
3395 | if (slen != olen) | |
5637ef5b NC |
3396 | Perl_croak(aTHX_ "panic: swatch_get found swatch length " |
3397 | "mismatch, slen=%"UVuf", olen=%"UVuf, | |
3398 | (UV)slen, (UV)olen); | |
979f2922 TS |
3399 | |
3400 | switch (opc) { | |
3401 | case '+': | |
3402 | while (slen--) | |
3403 | *s++ |= *o++; | |
3404 | break; | |
3405 | case '!': | |
3406 | while (slen--) | |
3407 | *s++ |= ~*o++; | |
3408 | break; | |
3409 | case '-': | |
3410 | while (slen--) | |
3411 | *s++ &= ~*o++; | |
3412 | break; | |
3413 | case '&': | |
3414 | while (slen--) | |
3415 | *s++ &= *o++; | |
3416 | break; | |
3417 | default: | |
3418 | break; | |
3419 | } | |
3420 | } | |
711a919c | 3421 | else { |
979f2922 TS |
3422 | STRLEN otheroctets = otherbits >> 3; |
3423 | STRLEN offset = 0; | |
35da51f7 | 3424 | U8* const send = s + slen; |
979f2922 TS |
3425 | |
3426 | while (s < send) { | |
3427 | UV otherval = 0; | |
3428 | ||
3429 | if (otherbits == 1) { | |
3430 | otherval = (o[offset >> 3] >> (offset & 7)) & 1; | |
3431 | ++offset; | |
3432 | } | |
3433 | else { | |
3434 | STRLEN vlen = otheroctets; | |
3435 | otherval = *o++; | |
3436 | while (--vlen) { | |
3437 | otherval <<= 8; | |
3438 | otherval |= *o++; | |
3439 | } | |
3440 | } | |
3441 | ||
711a919c | 3442 | if (opc == '+' && otherval) |
6f207bd3 | 3443 | NOOP; /* replace with otherval */ |
979f2922 TS |
3444 | else if (opc == '!' && !otherval) |
3445 | otherval = 1; | |
3446 | else if (opc == '-' && otherval) | |
3447 | otherval = 0; | |
3448 | else if (opc == '&' && !otherval) | |
3449 | otherval = 0; | |
3450 | else { | |
711a919c | 3451 | s += octets; /* no replacement */ |
979f2922 TS |
3452 | continue; |
3453 | } | |
3454 | ||
3455 | if (bits == 8) | |
3456 | *s++ = (U8)( otherval & 0xff); | |
3457 | else if (bits == 16) { | |
3458 | *s++ = (U8)((otherval >> 8) & 0xff); | |
3459 | *s++ = (U8)( otherval & 0xff); | |
3460 | } | |
3461 | else if (bits == 32) { | |
3462 | *s++ = (U8)((otherval >> 24) & 0xff); | |
3463 | *s++ = (U8)((otherval >> 16) & 0xff); | |
3464 | *s++ = (U8)((otherval >> 8) & 0xff); | |
3465 | *s++ = (U8)( otherval & 0xff); | |
3466 | } | |
3467 | } | |
3468 | } | |
3469 | sv_free(other); /* through with it! */ | |
3470 | } /* while */ | |
3471 | return swatch; | |
3472 | } | |
3473 | ||
064c021d | 3474 | HV* |
4c2e1131 | 3475 | Perl__swash_inversion_hash(pTHX_ SV* const swash) |
064c021d KW |
3476 | { |
3477 | ||
79a2a0e8 | 3478 | /* Subject to change or removal. For use only in regcomp.c and regexec.c |
5662e334 KW |
3479 | * Can't be used on a property that is subject to user override, as it |
3480 | * relies on the value of SPECIALS in the swash which would be set by | |
3481 | * utf8_heavy.pl to the hash in the non-overriden file, and hence is not set | |
3482 | * for overridden properties | |
064c021d KW |
3483 | * |
3484 | * Returns a hash which is the inversion and closure of a swash mapping. | |
3485 | * For example, consider the input lines: | |
3486 | * 004B 006B | |
3487 | * 004C 006C | |
3488 | * 212A 006B | |
3489 | * | |
4a4088c4 | 3490 | * The returned hash would have two keys, the UTF-8 for 006B and the UTF-8 for |
064c021d | 3491 | * 006C. The value for each key is an array. For 006C, the array would |
4a4088c4 KW |
3492 | * have two elements, the UTF-8 for itself, and for 004C. For 006B, there |
3493 | * would be three elements in its array, the UTF-8 for 006B, 004B and 212A. | |
064c021d | 3494 | * |
538e84ed KW |
3495 | * Note that there are no elements in the hash for 004B, 004C, 212A. The |
3496 | * keys are only code points that are folded-to, so it isn't a full closure. | |
3497 | * | |
064c021d KW |
3498 | * Essentially, for any code point, it gives all the code points that map to |
3499 | * it, or the list of 'froms' for that point. | |
3500 | * | |
5662e334 KW |
3501 | * Currently it ignores any additions or deletions from other swashes, |
3502 | * looking at just the main body of the swash, and if there are SPECIALS | |
3503 | * in the swash, at that hash | |
3504 | * | |
3505 | * The specials hash can be extra code points, and most likely consists of | |
3506 | * maps from single code points to multiple ones (each expressed as a string | |
4a4088c4 | 3507 | * of UTF-8 characters). This function currently returns only 1-1 mappings. |
5662e334 KW |
3508 | * However consider this possible input in the specials hash: |
3509 | * "\xEF\xAC\x85" => "\x{0073}\x{0074}", # U+FB05 => 0073 0074 | |
3510 | * "\xEF\xAC\x86" => "\x{0073}\x{0074}", # U+FB06 => 0073 0074 | |
3511 | * | |
3512 | * Both FB05 and FB06 map to the same multi-char sequence, which we don't | |
3513 | * currently handle. But it also means that FB05 and FB06 are equivalent in | |
3514 | * a 1-1 mapping which we should handle, and this relationship may not be in | |
3515 | * the main table. Therefore this function examines all the multi-char | |
74894415 KW |
3516 | * sequences and adds the 1-1 mappings that come out of that. |
3517 | * | |
3518 | * XXX This function was originally intended to be multipurpose, but its | |
3519 | * only use is quite likely to remain for constructing the inversion of | |
3520 | * the CaseFolding (//i) property. If it were more general purpose for | |
3521 | * regex patterns, it would have to do the FB05/FB06 game for simple folds, | |
3522 | * because certain folds are prohibited under /iaa and /il. As an example, | |
3523 | * in Unicode 3.0.1 both U+0130 and U+0131 fold to 'i', and hence are both | |
3524 | * equivalent under /i. But under /iaa and /il, the folds to 'i' are | |
3525 | * prohibited, so we would not figure out that they fold to each other. | |
3526 | * Code could be written to automatically figure this out, similar to the | |
3527 | * code that does this for multi-character folds, but this is the only case | |
3528 | * where something like this is ever likely to happen, as all the single | |
7ee537e6 | 3529 | * char folds to the 0-255 range are now quite settled. Instead there is a |
74894415 KW |
3530 | * little special code that is compiled only for this Unicode version. This |
3531 | * is smaller and didn't require much coding time to do. But this makes | |
3532 | * this routine strongly tied to being used just for CaseFolding. If ever | |
3533 | * it should be generalized, this would have to be fixed */ | |
064c021d KW |
3534 | |
3535 | U8 *l, *lend; | |
3536 | STRLEN lcur; | |
3537 | HV *const hv = MUTABLE_HV(SvRV(swash)); | |
3538 | ||
923b6d4e KW |
3539 | /* The string containing the main body of the table. This will have its |
3540 | * assertion fail if the swash has been converted to its inversion list */ | |
064c021d KW |
3541 | SV** const listsvp = hv_fetchs(hv, "LIST", FALSE); |
3542 | ||
3543 | SV** const typesvp = hv_fetchs(hv, "TYPE", FALSE); | |
3544 | SV** const bitssvp = hv_fetchs(hv, "BITS", FALSE); | |
3545 | SV** const nonesvp = hv_fetchs(hv, "NONE", FALSE); | |
3546 | /*SV** const extssvp = hv_fetchs(hv, "EXTRAS", FALSE);*/ | |
3547 | const U8* const typestr = (U8*)SvPV_nolen(*typesvp); | |
3548 | const STRLEN bits = SvUV(*bitssvp); | |
3549 | const STRLEN octets = bits >> 3; /* if bits == 1, then octets == 0 */ | |
3550 | const UV none = SvUV(*nonesvp); | |
5662e334 | 3551 | SV **specials_p = hv_fetchs(hv, "SPECIALS", 0); |
064c021d KW |
3552 | |
3553 | HV* ret = newHV(); | |
3554 | ||
3555 | PERL_ARGS_ASSERT__SWASH_INVERSION_HASH; | |
3556 | ||
3557 | /* Must have at least 8 bits to get the mappings */ | |
3558 | if (bits != 8 && bits != 16 && bits != 32) { | |
3559 | Perl_croak(aTHX_ "panic: swash_inversion_hash doesn't expect bits %"UVuf, | |
3560 | (UV)bits); | |
3561 | } | |
3562 | ||
5662e334 KW |
3563 | if (specials_p) { /* It might be "special" (sometimes, but not always, a |
3564 | mapping to more than one character */ | |
3565 | ||
3566 | /* Construct an inverse mapping hash for the specials */ | |
3567 | HV * const specials_hv = MUTABLE_HV(SvRV(*specials_p)); | |
3568 | HV * specials_inverse = newHV(); | |
3569 | char *char_from; /* the lhs of the map */ | |
3570 | I32 from_len; /* its byte length */ | |
3571 | char *char_to; /* the rhs of the map */ | |
3572 | I32 to_len; /* its byte length */ | |
3573 | SV *sv_to; /* and in a sv */ | |
3574 | AV* from_list; /* list of things that map to each 'to' */ | |
3575 | ||
3576 | hv_iterinit(specials_hv); | |
3577 | ||
4a4088c4 KW |
3578 | /* The keys are the characters (in UTF-8) that map to the corresponding |
3579 | * UTF-8 string value. Iterate through the list creating the inverse | |
5662e334 KW |
3580 | * list. */ |
3581 | while ((sv_to = hv_iternextsv(specials_hv, &char_from, &from_len))) { | |
3582 | SV** listp; | |
3583 | if (! SvPOK(sv_to)) { | |
5637ef5b NC |
3584 | Perl_croak(aTHX_ "panic: value returned from hv_iternextsv() " |
3585 | "unexpectedly is not a string, flags=%lu", | |
3586 | (unsigned long)SvFLAGS(sv_to)); | |
5662e334 | 3587 | } |
4b88fb76 | 3588 | /*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 |
3589 | |
3590 | /* Each key in the inverse list is a mapped-to value, and the key's | |
4a4088c4 | 3591 | * hash value is a list of the strings (each in UTF-8) that map to |
5662e334 KW |
3592 | * it. Those strings are all one character long */ |
3593 | if ((listp = hv_fetch(specials_inverse, | |
3594 | SvPVX(sv_to), | |
3595 | SvCUR(sv_to), 0))) | |
3596 | { | |
3597 | from_list = (AV*) *listp; | |
3598 | } | |
3599 | else { /* No entry yet for it: create one */ | |
3600 | from_list = newAV(); | |
3601 | if (! hv_store(specials_inverse, | |
3602 | SvPVX(sv_to), | |
3603 | SvCUR(sv_to), | |
3604 | (SV*) from_list, 0)) | |
3605 | { | |
3606 | Perl_croak(aTHX_ "panic: hv_store() unexpectedly failed"); | |
3607 | } | |
3608 | } | |
3609 | ||
3610 | /* Here have the list associated with this 'to' (perhaps newly | |
3611 | * created and empty). Just add to it. Note that we ASSUME that | |
3612 | * the input is guaranteed to not have duplications, so we don't | |
3613 | * check for that. Duplications just slow down execution time. */ | |
3614 | av_push(from_list, newSVpvn_utf8(char_from, from_len, TRUE)); | |
3615 | } | |
3616 | ||
3617 | /* Here, 'specials_inverse' contains the inverse mapping. Go through | |
3618 | * it looking for cases like the FB05/FB06 examples above. There would | |
3619 | * be an entry in the hash like | |
3620 | * 'st' => [ FB05, FB06 ] | |
3621 | * In this example we will create two lists that get stored in the | |
3622 | * returned hash, 'ret': | |
3623 | * FB05 => [ FB05, FB06 ] | |
3624 | * FB06 => [ FB05, FB06 ] | |
3625 | * | |
3626 | * Note that there is nothing to do if the array only has one element. | |
3627 | * (In the normal 1-1 case handled below, we don't have to worry about | |
3628 | * two lists, as everything gets tied to the single list that is | |
3629 | * generated for the single character 'to'. But here, we are omitting | |
3630 | * that list, ('st' in the example), so must have multiple lists.) */ | |
3631 | while ((from_list = (AV *) hv_iternextsv(specials_inverse, | |
3632 | &char_to, &to_len))) | |
3633 | { | |
7e13d567 | 3634 | if (av_tindex_nomg(from_list) > 0) { |
c70927a6 | 3635 | SSize_t i; |
5662e334 KW |
3636 | |
3637 | /* We iterate over all combinations of i,j to place each code | |
3638 | * point on each list */ | |
7e13d567 | 3639 | for (i = 0; i <= av_tindex_nomg(from_list); i++) { |
c70927a6 | 3640 | SSize_t j; |
5662e334 KW |
3641 | AV* i_list = newAV(); |
3642 | SV** entryp = av_fetch(from_list, i, FALSE); | |
3643 | if (entryp == NULL) { | |
3644 | Perl_croak(aTHX_ "panic: av_fetch() unexpectedly failed"); | |
3645 | } | |
3646 | if (hv_fetch(ret, SvPVX(*entryp), SvCUR(*entryp), FALSE)) { | |
3647 | Perl_croak(aTHX_ "panic: unexpected entry for %s", SvPVX(*entryp)); | |
3648 | } | |
3649 | if (! hv_store(ret, SvPVX(*entryp), SvCUR(*entryp), | |
3650 | (SV*) i_list, FALSE)) | |
3651 | { | |
3652 | Perl_croak(aTHX_ "panic: hv_store() unexpectedly failed"); | |
3653 | } | |
3654 | ||
538e84ed | 3655 | /* For DEBUG_U: UV u = valid_utf8_to_uvchr((U8*) SvPVX(*entryp), 0);*/ |
7e13d567 | 3656 | for (j = 0; j <= av_tindex_nomg(from_list); j++) { |
5662e334 KW |
3657 | entryp = av_fetch(from_list, j, FALSE); |
3658 | if (entryp == NULL) { | |
3659 | Perl_croak(aTHX_ "panic: av_fetch() unexpectedly failed"); | |
3660 | } | |
3661 | ||
3662 | /* When i==j this adds itself to the list */ | |
4b88fb76 KW |
3663 | av_push(i_list, newSVuv(utf8_to_uvchr_buf( |
3664 | (U8*) SvPVX(*entryp), | |
3665 | (U8*) SvPVX(*entryp) + SvCUR(*entryp), | |
3666 | 0))); | |
4637d003 | 3667 | /*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 |
3668 | } |
3669 | } | |
3670 | } | |
3671 | } | |
3672 | SvREFCNT_dec(specials_inverse); /* done with it */ | |
3673 | } /* End of specials */ | |
3674 | ||
064c021d | 3675 | /* read $swash->{LIST} */ |
74894415 KW |
3676 | |
3677 | #if UNICODE_MAJOR_VERSION == 3 \ | |
3678 | && UNICODE_DOT_VERSION == 0 \ | |
3679 | && UNICODE_DOT_DOT_VERSION == 1 | |
3680 | ||
3681 | /* For this version only U+130 and U+131 are equivalent under qr//i. Add a | |
3682 | * rule so that things work under /iaa and /il */ | |
3683 | ||
3684 | SV * mod_listsv = sv_mortalcopy(*listsvp); | |
3685 | sv_catpv(mod_listsv, "130\t130\t131\n"); | |
3686 | l = (U8*)SvPV(mod_listsv, lcur); | |
3687 | ||
3688 | #else | |
3689 | ||
064c021d | 3690 | l = (U8*)SvPV(*listsvp, lcur); |
74894415 KW |
3691 | |
3692 | #endif | |
3693 | ||
064c021d KW |
3694 | lend = l + lcur; |
3695 | ||
3696 | /* Go through each input line */ | |
3697 | while (l < lend) { | |
3698 | UV min, max, val; | |
3699 | UV inverse; | |
95543e92 KW |
3700 | l = swash_scan_list_line(l, lend, &min, &max, &val, |
3701 | cBOOL(octets), typestr); | |
064c021d KW |
3702 | if (l > lend) { |
3703 | break; | |
3704 | } | |
3705 | ||
3706 | /* Each element in the range is to be inverted */ | |
3707 | for (inverse = min; inverse <= max; inverse++) { | |
3708 | AV* list; | |
064c021d KW |
3709 | SV** listp; |
3710 | IV i; | |
3711 | bool found_key = FALSE; | |
5662e334 | 3712 | bool found_inverse = FALSE; |
064c021d KW |
3713 | |
3714 | /* The key is the inverse mapping */ | |
3715 | char key[UTF8_MAXBYTES+1]; | |
c80e42f3 | 3716 | char* key_end = (char *) uvchr_to_utf8((U8*) key, val); |
064c021d KW |
3717 | STRLEN key_len = key_end - key; |
3718 | ||
064c021d KW |
3719 | /* Get the list for the map */ |
3720 | if ((listp = hv_fetch(ret, key, key_len, FALSE))) { | |
3721 | list = (AV*) *listp; | |
3722 | } | |
3723 | else { /* No entry yet for it: create one */ | |
3724 | list = newAV(); | |
3725 | if (! hv_store(ret, key, key_len, (SV*) list, FALSE)) { | |
3726 | Perl_croak(aTHX_ "panic: hv_store() unexpectedly failed"); | |
3727 | } | |
3728 | } | |
3729 | ||
5662e334 KW |
3730 | /* Look through list to see if this inverse mapping already is |
3731 | * listed, or if there is a mapping to itself already */ | |
7e13d567 | 3732 | for (i = 0; i <= av_tindex_nomg(list); i++) { |
064c021d KW |
3733 | SV** entryp = av_fetch(list, i, FALSE); |
3734 | SV* entry; | |
414db8a1 | 3735 | UV uv; |
064c021d KW |
3736 | if (entryp == NULL) { |
3737 | Perl_croak(aTHX_ "panic: av_fetch() unexpectedly failed"); | |
3738 | } | |
3739 | entry = *entryp; | |
414db8a1 DM |
3740 | uv = SvUV(entry); |
3741 | /*DEBUG_U(PerlIO_printf(Perl_debug_log, "list for %"UVXf" contains %"UVXf"\n", val, uv));*/ | |
3742 | if (uv == val) { | |
064c021d | 3743 | found_key = TRUE; |
5662e334 | 3744 | } |
414db8a1 | 3745 | if (uv == inverse) { |
5662e334 KW |
3746 | found_inverse = TRUE; |
3747 | } | |
3748 | ||
3749 | /* No need to continue searching if found everything we are | |
3750 | * looking for */ | |
3751 | if (found_key && found_inverse) { | |
064c021d KW |
3752 | break; |
3753 | } | |
3754 | } | |
56ca34ca KW |
3755 | |
3756 | /* Make sure there is a mapping to itself on the list */ | |
064c021d | 3757 | if (! found_key) { |
d397ff6a | 3758 | av_push(list, newSVuv(val)); |
4637d003 | 3759 | /*DEBUG_U(PerlIO_printf(Perl_debug_log, "%s: %d: Adding %"UVXf" to list for %"UVXf"\n", __FILE__, __LINE__, val, val));*/ |
064c021d KW |
3760 | } |
3761 | ||
3762 | ||
3763 | /* Simply add the value to the list */ | |
5662e334 KW |
3764 | if (! found_inverse) { |
3765 | av_push(list, newSVuv(inverse)); | |
4637d003 | 3766 | /*DEBUG_U(PerlIO_printf(Perl_debug_log, "%s: %d: Adding %"UVXf" to list for %"UVXf"\n", __FILE__, __LINE__, inverse, val));*/ |
5662e334 | 3767 | } |
064c021d | 3768 | |
b0e3252e | 3769 | /* swatch_get() increments the value of val for each element in the |
064c021d KW |
3770 | * range. That makes more compact tables possible. You can |
3771 | * express the capitalization, for example, of all consecutive | |
3772 | * letters with a single line: 0061\t007A\t0041 This maps 0061 to | |
3773 | * 0041, 0062 to 0042, etc. I (khw) have never understood 'none', | |
bd3f2f94 | 3774 | * and it's not documented; it appears to be used only in |
b0e3252e | 3775 | * implementing tr//; I copied the semantics from swatch_get(), just |
bd3f2f94 | 3776 | * in case */ |
064c021d KW |
3777 | if (!none || val < none) { |
3778 | ++val; | |
3779 | } | |
3780 | } | |
3781 | } | |
3782 | ||
3783 | return ret; | |
3784 | } | |
3785 | ||
a25abddc | 3786 | SV* |
d764b54e KW |
3787 | Perl__swash_to_invlist(pTHX_ SV* const swash) |
3788 | { | |
3789 | ||
ed92f1b3 KW |
3790 | /* Subject to change or removal. For use only in one place in regcomp.c. |
3791 | * Ownership is given to one reference count in the returned SV* */ | |
d764b54e KW |
3792 | |
3793 | U8 *l, *lend; | |
3794 | char *loc; | |
3795 | STRLEN lcur; | |
3796 | HV *const hv = MUTABLE_HV(SvRV(swash)); | |
3797 | UV elements = 0; /* Number of elements in the inversion list */ | |
b443038a | 3798 | U8 empty[] = ""; |
923b6d4e KW |
3799 | SV** listsvp; |
3800 | SV** typesvp; | |
3801 | SV** bitssvp; | |
3802 | SV** extssvp; | |
3803 | SV** invert_it_svp; | |
d764b54e | 3804 | |
923b6d4e KW |
3805 | U8* typestr; |
3806 | STRLEN bits; | |
3807 | STRLEN octets; /* if bits == 1, then octets == 0 */ | |
d73c39c5 KW |
3808 | U8 *x, *xend; |
3809 | STRLEN xcur; | |
d764b54e | 3810 | |
a25abddc | 3811 | SV* invlist; |
d764b54e | 3812 | |
b81740c0 KW |
3813 | PERL_ARGS_ASSERT__SWASH_TO_INVLIST; |
3814 | ||
923b6d4e KW |
3815 | /* If not a hash, it must be the swash's inversion list instead */ |
3816 | if (SvTYPE(hv) != SVt_PVHV) { | |
ed92f1b3 | 3817 | return SvREFCNT_inc_simple_NN((SV*) hv); |
923b6d4e KW |
3818 | } |
3819 | ||
3820 | /* The string containing the main body of the table */ | |
3821 | listsvp = hv_fetchs(hv, "LIST", FALSE); | |
3822 | typesvp = hv_fetchs(hv, "TYPE", FALSE); | |
3823 | bitssvp = hv_fetchs(hv, "BITS", FALSE); | |
3824 | extssvp = hv_fetchs(hv, "EXTRAS", FALSE); | |
3825 | invert_it_svp = hv_fetchs(hv, "INVERT_IT", FALSE); | |
3826 | ||
3827 | typestr = (U8*)SvPV_nolen(*typesvp); | |
3828 | bits = SvUV(*bitssvp); | |
3829 | octets = bits >> 3; /* if bits == 1, then octets == 0 */ | |
3830 | ||
d764b54e | 3831 | /* read $swash->{LIST} */ |
b443038a KW |
3832 | if (SvPOK(*listsvp)) { |
3833 | l = (U8*)SvPV(*listsvp, lcur); | |
3834 | } | |
3835 | else { | |
3836 | /* LIST legitimately doesn't contain a string during compilation phases | |
3837 | * of Perl itself, before the Unicode tables are generated. In this | |
3838 | * case, just fake things up by creating an empty list */ | |
3839 | l = empty; | |
3840 | lcur = 0; | |
3841 | } | |
d764b54e KW |
3842 | loc = (char *) l; |
3843 | lend = l + lcur; | |
3844 | ||
31aa6e0b | 3845 | if (*l == 'V') { /* Inversion list format */ |
99944878 | 3846 | const char *after_atou = (char *) lend; |
31aa6e0b KW |
3847 | UV element0; |
3848 | UV* other_elements_ptr; | |
3849 | ||
3850 | /* The first number is a count of the rest */ | |
3851 | l++; | |
22ff3130 HS |
3852 | if (!grok_atoUV((const char *)l, &elements, &after_atou)) { |
3853 | Perl_croak(aTHX_ "panic: Expecting a valid count of elements at start of inversion list"); | |
3854 | } | |
eb092534 KW |
3855 | if (elements == 0) { |
3856 | invlist = _new_invlist(0); | |
3857 | } | |
3858 | else { | |
99944878 JH |
3859 | while (isSPACE(*l)) l++; |
3860 | l = (U8 *) after_atou; | |
1f9f7d4c KW |
3861 | |
3862 | /* Get the 0th element, which is needed to setup the inversion list */ | |
99944878 | 3863 | while (isSPACE(*l)) l++; |
22ff3130 HS |
3864 | if (!grok_atoUV((const char *)l, &element0, &after_atou)) { |
3865 | Perl_croak(aTHX_ "panic: Expecting a valid 0th element for inversion list"); | |
3866 | } | |
99944878 | 3867 | l = (U8 *) after_atou; |
1f9f7d4c KW |
3868 | invlist = _setup_canned_invlist(elements, element0, &other_elements_ptr); |
3869 | elements--; | |
3870 | ||
3871 | /* Then just populate the rest of the input */ | |
3872 | while (elements-- > 0) { | |
3873 | if (l > lend) { | |
3874 | Perl_croak(aTHX_ "panic: Expecting %"UVuf" more elements than available", elements); | |
3875 | } | |
99944878 | 3876 | while (isSPACE(*l)) l++; |
22ff3130 HS |
3877 | if (!grok_atoUV((const char *)l, other_elements_ptr++, &after_atou)) { |
3878 | Perl_croak(aTHX_ "panic: Expecting a valid element in inversion list"); | |
3879 | } | |
99944878 | 3880 | l = (U8 *) after_atou; |
1f9f7d4c | 3881 | } |
eb092534 | 3882 | } |
31aa6e0b KW |
3883 | } |
3884 | else { | |
3885 | ||
1784d2f9 KW |
3886 | /* Scan the input to count the number of lines to preallocate array |
3887 | * size based on worst possible case, which is each line in the input | |
3888 | * creates 2 elements in the inversion list: 1) the beginning of a | |
3889 | * range in the list; 2) the beginning of a range not in the list. */ | |
3890 | while ((loc = (strchr(loc, '\n'))) != NULL) { | |
3891 | elements += 2; | |
3892 | loc++; | |
3893 | } | |
d764b54e | 3894 | |
1784d2f9 KW |
3895 | /* If the ending is somehow corrupt and isn't a new line, add another |
3896 | * element for the final range that isn't in the inversion list */ | |
3897 | if (! (*lend == '\n' | |
3898 | || (*lend == '\0' && (lcur == 0 || *(lend - 1) == '\n')))) | |
3899 | { | |
3900 | elements++; | |
3901 | } | |
d764b54e | 3902 | |
1784d2f9 | 3903 | invlist = _new_invlist(elements); |
d764b54e | 3904 | |
1784d2f9 KW |
3905 | /* Now go through the input again, adding each range to the list */ |
3906 | while (l < lend) { | |
3907 | UV start, end; | |
3908 | UV val; /* Not used by this function */ | |
d764b54e | 3909 | |
95543e92 KW |
3910 | l = swash_scan_list_line(l, lend, &start, &end, &val, |
3911 | cBOOL(octets), typestr); | |
d764b54e | 3912 | |
1784d2f9 KW |
3913 | if (l > lend) { |
3914 | break; | |
3915 | } | |
3916 | ||
3917 | invlist = _add_range_to_invlist(invlist, start, end); | |
3918 | } | |
31aa6e0b | 3919 | } |
d764b54e | 3920 | |
77f9f126 KW |
3921 | /* Invert if the data says it should be */ |
3922 | if (invert_it_svp && SvUV(*invert_it_svp)) { | |
25151030 | 3923 | _invlist_invert(invlist); |
77f9f126 KW |
3924 | } |
3925 | ||
b0e3252e | 3926 | /* This code is copied from swatch_get() |
d73c39c5 KW |
3927 | * read $swash->{EXTRAS} */ |
3928 | x = (U8*)SvPV(*extssvp, xcur); | |
3929 | xend = x + xcur; | |
3930 | while (x < xend) { | |
3931 | STRLEN namelen; | |
3932 | U8 *namestr; | |
3933 | SV** othersvp; | |
3934 | HV* otherhv; | |
3935 | STRLEN otherbits; | |
3936 | SV **otherbitssvp, *other; | |
3937 | U8 *nl; | |
3938 | ||
3939 | const U8 opc = *x++; | |
3940 | if (opc == '\n') | |
3941 | continue; | |
3942 | ||
3943 | nl = (U8*)memchr(x, '\n', xend - x); | |
3944 | ||
3945 | if (opc != '-' && opc != '+' && opc != '!' && opc != '&') { | |
3946 | if (nl) { | |
3947 | x = nl + 1; /* 1 is length of "\n" */ | |
3948 | continue; | |
3949 | } | |
3950 | else { | |
3951 | x = xend; /* to EXTRAS' end at which \n is not found */ | |
3952 | break; | |
3953 | } | |
3954 | } | |
3955 | ||
3956 | namestr = x; | |
3957 | if (nl) { | |
3958 | namelen = nl - namestr; | |
3959 | x = nl + 1; | |
3960 | } | |
3961 | else { | |
3962 | namelen = xend - namestr; | |
3963 | x = xend; | |
3964 | } | |
3965 | ||
3966 | othersvp = hv_fetch(hv, (char *)namestr, namelen, FALSE); | |
3967 | otherhv = MUTABLE_HV(SvRV(*othersvp)); | |
3968 | otherbitssvp = hv_fetchs(otherhv, "BITS", FALSE); | |
3969 | otherbits = (STRLEN)SvUV(*otherbitssvp); | |
3970 | ||
3971 | if (bits != otherbits || bits != 1) { | |
5637ef5b NC |
3972 | Perl_croak(aTHX_ "panic: _swash_to_invlist only operates on boolean " |
3973 | "properties, bits=%"UVuf", otherbits=%"UVuf, | |
3974 | (UV)bits, (UV)otherbits); | |
d73c39c5 KW |
3975 | } |
3976 | ||
3977 | /* The "other" swatch must be destroyed after. */ | |
3978 | other = _swash_to_invlist((SV *)*othersvp); | |
3979 | ||
b0e3252e | 3980 | /* End of code copied from swatch_get() */ |
d73c39c5 KW |
3981 | switch (opc) { |
3982 | case '+': | |
3983 | _invlist_union(invlist, other, &invlist); | |
3984 | break; | |
3985 | case '!': | |
6c46377d | 3986 | _invlist_union_maybe_complement_2nd(invlist, other, TRUE, &invlist); |
d73c39c5 KW |
3987 | break; |
3988 | case '-': | |
3989 | _invlist_subtract(invlist, other, &invlist); | |
3990 | break; | |
3991 | case '&': | |
3992 | _invlist_intersection(invlist, other, &invlist); | |
3993 | break; | |
3994 | default: | |
3995 | break; | |
3996 | } | |
3997 | sv_free(other); /* through with it! */ | |
3998 | } | |
3999 | ||
dbfdbd26 | 4000 | SvREADONLY_on(invlist); |
d764b54e KW |
4001 | return invlist; |
4002 | } | |
4003 | ||
3fdfee00 KW |
4004 | SV* |
4005 | Perl__get_swash_invlist(pTHX_ SV* const swash) | |
4006 | { | |
872dd7e0 | 4007 | SV** ptr; |
3fdfee00 KW |
4008 | |
4009 | PERL_ARGS_ASSERT__GET_SWASH_INVLIST; | |
4010 | ||
87367d5f | 4011 | if (! SvROK(swash)) { |
872dd7e0 KW |
4012 | return NULL; |
4013 | } | |
4014 | ||
87367d5f KW |
4015 | /* If it really isn't a hash, it isn't really swash; must be an inversion |
4016 | * list */ | |
4017 | if (SvTYPE(SvRV(swash)) != SVt_PVHV) { | |
4018 | return SvRV(swash); | |
4019 | } | |
872dd7e0 | 4020 | |
87367d5f | 4021 | ptr = hv_fetchs(MUTABLE_HV(SvRV(swash)), "V", FALSE); |
3fdfee00 KW |
4022 | if (! ptr) { |
4023 | return NULL; | |
4024 | } | |
4025 | ||
4026 | return *ptr; | |
4027 | } | |
4028 | ||
0876b9a0 | 4029 | bool |
5aaab254 | 4030 | Perl_check_utf8_print(pTHX_ const U8* s, const STRLEN len) |
0876b9a0 KW |
4031 | { |
4032 | /* May change: warns if surrogates, non-character code points, or | |
af2af982 KW |
4033 | * non-Unicode code points are in s which has length len bytes. Returns |
4034 | * TRUE if none found; FALSE otherwise. The only other validity check is | |
760c7c2f KW |
4035 | * to make sure that this won't exceed the string's length. |
4036 | * | |
4037 | * Code points above the platform's C<IV_MAX> will raise a deprecation | |
4038 | * warning, unless those are turned off. */ | |
0876b9a0 KW |
4039 | |
4040 | const U8* const e = s + len; | |
4041 | bool ok = TRUE; | |
4042 | ||
4043 | PERL_ARGS_ASSERT_CHECK_UTF8_PRINT; | |
4044 | ||
4045 | while (s < e) { | |
4046 | if (UTF8SKIP(s) > len) { | |
4047 | Perl_ck_warner_d(aTHX_ packWARN(WARN_UTF8), | |
4048 | "%s in %s", unees, PL_op ? OP_DESC(PL_op) : "print"); | |
4049 | return FALSE; | |
4050 | } | |
ac6f1fbe | 4051 | if (UNLIKELY(isUTF8_POSSIBLY_PROBLEMATIC(*s))) { |
0876b9a0 | 4052 | STRLEN char_len; |
0c58a72b | 4053 | if (UTF8_IS_SUPER(s, e)) { |
760c7c2f KW |
4054 | if ( ckWARN_d(WARN_NON_UNICODE) |
4055 | || ( ckWARN_d(WARN_DEPRECATED) | |
4056 | #if defined(UV_IS_QUAD) | |
4057 | /* 2**63 and up meet these conditions provided we have | |
4058 | * a 64-bit word. */ | |
4059 | # ifdef EBCDIC | |
4060 | && *s == 0xFE && e - s >= UTF8_MAXBYTES | |
4061 | && s[1] >= 0x49 | |
4062 | # else | |
4063 | && *s == 0xFF && e -s >= UTF8_MAXBYTES | |
4064 | && s[2] >= 0x88 | |
4065 | # endif | |
4066 | #else /* Below is 32-bit words */ | |
4067 | /* 2**31 and above meet these conditions on all EBCDIC | |
4068 | * pages recognized for 32-bit platforms */ | |
4069 | # ifdef EBCDIC | |
4070 | && *s == 0xFE && e - s >= UTF8_MAXBYTES | |
4071 | && s[6] >= 0x43 | |
4072 | # else | |
4073 | && *s >= 0xFE | |
4074 | # endif | |
4075 | #endif | |
4076 | )) { | |
15ca5930 KW |
4077 | /* A side effect of this function will be to warn */ |
4078 | (void) utf8n_to_uvchr(s, e - s, &char_len, UTF8_WARN_SUPER); | |
7ee537e6 KW |
4079 | ok = FALSE; |
4080 | } | |
0876b9a0 | 4081 | } |
0c58a72b | 4082 | else if (UTF8_IS_SURROGATE(s, e)) { |
8457b38f | 4083 | if (ckWARN_d(WARN_SURROGATE)) { |
15ca5930 KW |
4084 | /* This has a different warning than the one the called |
4085 | * function would output, so can't just call it, unlike we | |
4086 | * do for the non-chars and above-unicodes */ | |
4b88fb76 | 4087 | UV uv = utf8_to_uvchr_buf(s, e, &char_len); |
8457b38f KW |
4088 | Perl_warner(aTHX_ packWARN(WARN_SURROGATE), |
4089 | "Unicode surrogate U+%04"UVXf" is illegal in UTF-8", uv); | |
4090 | ok = FALSE; | |
4091 | } | |
0876b9a0 | 4092 | } |
0c58a72b | 4093 | else if ((UTF8_IS_NONCHAR(s, e)) && (ckWARN_d(WARN_NONCHAR))) { |
15ca5930 KW |
4094 | /* A side effect of this function will be to warn */ |
4095 | (void) utf8n_to_uvchr(s, e - s, &char_len, UTF8_WARN_NONCHAR); | |
0876b9a0 KW |
4096 | ok = FALSE; |
4097 | } | |
4098 | } | |
4099 | s += UTF8SKIP(s); | |
4100 | } | |
4101 | ||
4102 | return ok; | |
4103 | } | |
4104 | ||
0f830e0b | 4105 | /* |
87cea99e | 4106 | =for apidoc pv_uni_display |
d2cc3551 | 4107 | |
a1433954 KW |
4108 | Build to the scalar C<dsv> a displayable version of the string C<spv>, |
4109 | length C<len>, the displayable version being at most C<pvlim> bytes long | |
796b6530 | 4110 | (if longer, the rest is truncated and C<"..."> will be appended). |
0a2ef054 | 4111 | |
796b6530 KW |
4112 | The C<flags> argument can have C<UNI_DISPLAY_ISPRINT> set to display |
4113 | C<isPRINT()>able characters as themselves, C<UNI_DISPLAY_BACKSLASH> | |
4114 | to display the C<\\[nrfta\\]> as the backslashed versions (like C<"\n">) | |
4115 | (C<UNI_DISPLAY_BACKSLASH> is preferred over C<UNI_DISPLAY_ISPRINT> for C<"\\">). | |
4116 | C<UNI_DISPLAY_QQ> (and its alias C<UNI_DISPLAY_REGEX>) have both | |
4117 | C<UNI_DISPLAY_BACKSLASH> and C<UNI_DISPLAY_ISPRINT> turned on. | |
0a2ef054 | 4118 | |
a1433954 | 4119 | The pointer to the PV of the C<dsv> is returned. |
d2cc3551 | 4120 | |
119bc988 KW |
4121 | See also L</sv_uni_display>. |
4122 | ||
d2cc3551 | 4123 | =cut */ |
e6b2e755 | 4124 | char * |
e1ec3a88 | 4125 | Perl_pv_uni_display(pTHX_ SV *dsv, const U8 *spv, STRLEN len, STRLEN pvlim, UV flags) |
e6b2e755 JH |
4126 | { |
4127 | int truncated = 0; | |
e1ec3a88 | 4128 | const char *s, *e; |
e6b2e755 | 4129 | |
7918f24d NC |
4130 | PERL_ARGS_ASSERT_PV_UNI_DISPLAY; |
4131 | ||
76f68e9b | 4132 | sv_setpvs(dsv, ""); |
7fddd944 | 4133 | SvUTF8_off(dsv); |
e1ec3a88 | 4134 | for (s = (const char *)spv, e = s + len; s < e; s += UTF8SKIP(s)) { |
e6b2e755 | 4135 | UV u; |
a49f32c6 NC |
4136 | /* This serves double duty as a flag and a character to print after |
4137 | a \ when flags & UNI_DISPLAY_BACKSLASH is true. | |
4138 | */ | |
4139 | char ok = 0; | |
c728cb41 | 4140 | |
e6b2e755 JH |
4141 | if (pvlim && SvCUR(dsv) >= pvlim) { |
4142 | truncated++; | |
4143 | break; | |
4144 | } | |
4b88fb76 | 4145 | u = utf8_to_uvchr_buf((U8*)s, (U8*)e, 0); |
c728cb41 | 4146 | if (u < 256) { |
a3b680e6 | 4147 | const unsigned char c = (unsigned char)u & 0xFF; |
0bd48802 | 4148 | if (flags & UNI_DISPLAY_BACKSLASH) { |
a49f32c6 | 4149 | switch (c) { |
c728cb41 | 4150 | case '\n': |
a49f32c6 | 4151 | ok = 'n'; break; |
c728cb41 | 4152 | case '\r': |
a49f32c6 | 4153 | ok = 'r'; break; |
c728cb41 | 4154 | case '\t': |
a49f32c6 | 4155 | ok = 't'; break; |
c728cb41 | 4156 | case '\f': |
a49f32c6 | 4157 | ok = 'f'; break; |
c728cb41 | 4158 | case '\a': |
a49f32c6 | 4159 | ok = 'a'; break; |
c728cb41 | 4160 | case '\\': |
a49f32c6 | 4161 | ok = '\\'; break; |
c728cb41 JH |
4162 | default: break; |
4163 | } | |
a49f32c6 | 4164 | if (ok) { |
88c9ea1e | 4165 | const char string = ok; |
76f68e9b | 4166 | sv_catpvs(dsv, "\\"); |
5e7aa789 | 4167 | sv_catpvn(dsv, &string, 1); |
a49f32c6 | 4168 | } |
c728cb41 | 4169 | } |
00e86452 | 4170 | /* isPRINT() is the locale-blind version. */ |
a49f32c6 | 4171 | if (!ok && (flags & UNI_DISPLAY_ISPRINT) && isPRINT(c)) { |
88c9ea1e | 4172 | const char string = c; |
5e7aa789 | 4173 | sv_catpvn(dsv, &string, 1); |
a49f32c6 | 4174 | ok = 1; |
0a2ef054 | 4175 | } |
c728cb41 JH |
4176 | } |
4177 | if (!ok) | |
9e55ce06 | 4178 | Perl_sv_catpvf(aTHX_ dsv, "\\x{%"UVxf"}", u); |
e6b2e755 JH |
4179 | } |
4180 | if (truncated) | |
396482e1 | 4181 | sv_catpvs(dsv, "..."); |
48ef279e | 4182 | |
e6b2e755 JH |
4183 | return SvPVX(dsv); |
4184 | } | |
2b9d42f0 | 4185 | |
d2cc3551 | 4186 | /* |
87cea99e | 4187 | =for apidoc sv_uni_display |
d2cc3551 | 4188 | |
a1433954 KW |
4189 | Build to the scalar C<dsv> a displayable version of the scalar C<sv>, |
4190 | the displayable version being at most C<pvlim> bytes long | |
d2cc3551 | 4191 | (if longer, the rest is truncated and "..." will be appended). |
0a2ef054 | 4192 | |
a1433954 | 4193 | The C<flags> argument is as in L</pv_uni_display>(). |
0a2ef054 | 4194 | |
a1433954 | 4195 | The pointer to the PV of the C<dsv> is returned. |
d2cc3551 | 4196 | |
d4c19fe8 AL |
4197 | =cut |
4198 | */ | |
e6b2e755 JH |
4199 | char * |
4200 | Perl_sv_uni_display(pTHX_ SV *dsv, SV *ssv, STRLEN pvlim, UV flags) | |
4201 | { | |
8cdde9f8 NC |
4202 | const char * const ptr = |
4203 | isREGEXP(ssv) ? RX_WRAPPED((REGEXP*)ssv) : SvPVX_const(ssv); | |
4204 | ||
7918f24d NC |
4205 | PERL_ARGS_ASSERT_SV_UNI_DISPLAY; |
4206 | ||
8cdde9f8 | 4207 | return Perl_pv_uni_display(aTHX_ dsv, (const U8*)ptr, |
cfd0369c | 4208 | SvCUR(ssv), pvlim, flags); |
701a277b JH |
4209 | } |
4210 | ||
d2cc3551 | 4211 | /* |
e6226b18 | 4212 | =for apidoc foldEQ_utf8 |
d2cc3551 | 4213 | |
a1433954 | 4214 | Returns true if the leading portions of the strings C<s1> and C<s2> (either or both |
e6226b18 | 4215 | of which may be in UTF-8) are the same case-insensitively; false otherwise. |
d51c1b21 | 4216 | How far into the strings to compare is determined by other input parameters. |
8b35872c | 4217 | |
a1433954 KW |
4218 | If C<u1> is true, the string C<s1> is assumed to be in UTF-8-encoded Unicode; |
4219 | otherwise it is assumed to be in native 8-bit encoding. Correspondingly for C<u2> | |
4220 | with respect to C<s2>. | |
8b35872c | 4221 | |
a1433954 KW |
4222 | If the byte length C<l1> is non-zero, it says how far into C<s1> to check for fold |
4223 | equality. In other words, C<s1>+C<l1> will be used as a goal to reach. The | |
8b35872c | 4224 | scan will not be considered to be a match unless the goal is reached, and |
a1433954 KW |
4225 | scanning won't continue past that goal. Correspondingly for C<l2> with respect to |
4226 | C<s2>. | |
4227 | ||
796b6530 | 4228 | If C<pe1> is non-C<NULL> and the pointer it points to is not C<NULL>, that pointer is |
03bb5c85 KW |
4229 | considered an end pointer to the position 1 byte past the maximum point |
4230 | in C<s1> beyond which scanning will not continue under any circumstances. | |
4231 | (This routine assumes that UTF-8 encoded input strings are not malformed; | |
4232 | malformed input can cause it to read past C<pe1>). | |
4233 | This means that if both C<l1> and C<pe1> are specified, and C<pe1> | |
a1433954 KW |
4234 | is less than C<s1>+C<l1>, the match will never be successful because it can |
4235 | never | |
d51c1b21 | 4236 | get as far as its goal (and in fact is asserted against). Correspondingly for |
a1433954 | 4237 | C<pe2> with respect to C<s2>. |
8b35872c | 4238 | |
a1433954 KW |
4239 | At least one of C<s1> and C<s2> must have a goal (at least one of C<l1> and |
4240 | C<l2> must be non-zero), and if both do, both have to be | |
8b35872c KW |
4241 | reached for a successful match. Also, if the fold of a character is multiple |
4242 | characters, all of them must be matched (see tr21 reference below for | |
4243 | 'folding'). | |
4244 | ||
796b6530 | 4245 | Upon a successful match, if C<pe1> is non-C<NULL>, |
a1433954 KW |
4246 | it will be set to point to the beginning of the I<next> character of C<s1> |
4247 | beyond what was matched. Correspondingly for C<pe2> and C<s2>. | |
d2cc3551 JH |
4248 | |
4249 | For case-insensitiveness, the "casefolding" of Unicode is used | |
4250 | instead of upper/lowercasing both the characters, see | |
a1433954 | 4251 | L<http://www.unicode.org/unicode/reports/tr21/> (Case Mappings). |
d2cc3551 JH |
4252 | |
4253 | =cut */ | |
a33c29bc KW |
4254 | |
4255 | /* A flags parameter has been added which may change, and hence isn't | |
4256 | * externally documented. Currently it is: | |
4257 | * 0 for as-documented above | |
4258 | * FOLDEQ_UTF8_NOMIX_ASCII meaning that if a non-ASCII character folds to an | |
4259 | ASCII one, to not match | |
31f05a37 KW |
4260 | * FOLDEQ_LOCALE is set iff the rules from the current underlying |
4261 | * locale are to be used. | |
4262 | * FOLDEQ_S1_ALREADY_FOLDED s1 has already been folded before calling this | |
aa8ebe62 KW |
4263 | * routine. This allows that step to be skipped. |
4264 | * Currently, this requires s1 to be encoded as UTF-8 | |
4265 | * (u1 must be true), which is asserted for. | |
d635b710 KW |
4266 | * FOLDEQ_S1_FOLDS_SANE With either NOMIX_ASCII or LOCALE, no folds may |
4267 | * cross certain boundaries. Hence, the caller should | |
4268 | * let this function do the folding instead of | |
4269 | * pre-folding. This code contains an assertion to | |
4270 | * that effect. However, if the caller knows what | |
4271 | * it's doing, it can pass this flag to indicate that, | |
4272 | * and the assertion is skipped. | |
31f05a37 | 4273 | * FOLDEQ_S2_ALREADY_FOLDED Similarly. |
d635b710 | 4274 | * FOLDEQ_S2_FOLDS_SANE |
a33c29bc | 4275 | */ |
701a277b | 4276 | I32 |
5aaab254 | 4277 | Perl_foldEQ_utf8_flags(pTHX_ const char *s1, char **pe1, UV l1, bool u1, const char *s2, char **pe2, UV l2, bool u2, U32 flags) |
332ddc25 | 4278 | { |
eb578fdb KW |
4279 | const U8 *p1 = (const U8*)s1; /* Point to current char */ |
4280 | const U8 *p2 = (const U8*)s2; | |
4281 | const U8 *g1 = NULL; /* goal for s1 */ | |
4282 | const U8 *g2 = NULL; | |
4283 | const U8 *e1 = NULL; /* Don't scan s1 past this */ | |
4284 | U8 *f1 = NULL; /* Point to current folded */ | |
4285 | const U8 *e2 = NULL; | |
4286 | U8 *f2 = NULL; | |
48ef279e | 4287 | STRLEN n1 = 0, n2 = 0; /* Number of bytes in current char */ |
8b35872c KW |
4288 | U8 foldbuf1[UTF8_MAXBYTES_CASE+1]; |
4289 | U8 foldbuf2[UTF8_MAXBYTES_CASE+1]; | |
1d39b2cd | 4290 | U8 flags_for_folder = FOLD_FLAGS_FULL; |
8b35872c | 4291 | |
eda9cac1 | 4292 | PERL_ARGS_ASSERT_FOLDEQ_UTF8_FLAGS; |
8b35872c | 4293 | |
cea315b6 | 4294 | assert( ! ((flags & (FOLDEQ_UTF8_NOMIX_ASCII | FOLDEQ_LOCALE)) |
d635b710 KW |
4295 | && (((flags & FOLDEQ_S1_ALREADY_FOLDED) |
4296 | && !(flags & FOLDEQ_S1_FOLDS_SANE)) | |
4297 | || ((flags & FOLDEQ_S2_ALREADY_FOLDED) | |
4298 | && !(flags & FOLDEQ_S2_FOLDS_SANE))))); | |
b08f1bd5 KW |
4299 | /* The algorithm is to trial the folds without regard to the flags on |
4300 | * the first line of the above assert(), and then see if the result | |
4301 | * violates them. This means that the inputs can't be pre-folded to a | |
4302 | * violating result, hence the assert. This could be changed, with the | |
4303 | * addition of extra tests here for the already-folded case, which would | |
4304 | * slow it down. That cost is more than any possible gain for when these | |
4305 | * flags are specified, as the flags indicate /il or /iaa matching which | |
4306 | * is less common than /iu, and I (khw) also believe that real-world /il | |
4307 | * and /iaa matches are most likely to involve code points 0-255, and this | |
4308 | * function only under rare conditions gets called for 0-255. */ | |
18f762c3 | 4309 | |
1d39b2cd KW |
4310 | if (flags & FOLDEQ_LOCALE) { |
4311 | if (IN_UTF8_CTYPE_LOCALE) { | |
4312 | flags &= ~FOLDEQ_LOCALE; | |
4313 | } | |
4314 | else { | |
4315 | flags_for_folder |= FOLD_FLAGS_LOCALE; | |
4316 | } | |
31f05a37 KW |
4317 | } |
4318 | ||
8b35872c | 4319 | if (pe1) { |
48ef279e | 4320 | e1 = *(U8**)pe1; |
8b35872c KW |
4321 | } |
4322 | ||
4323 | if (l1) { | |
48ef279e | 4324 | g1 = (const U8*)s1 + l1; |
8b35872c KW |
4325 | } |
4326 | ||
4327 | if (pe2) { | |
48ef279e | 4328 | e2 = *(U8**)pe2; |
8b35872c KW |
4329 | } |
4330 | ||
4331 | if (l2) { | |
48ef279e | 4332 | g2 = (const U8*)s2 + l2; |
8b35872c KW |
4333 | } |
4334 | ||
4335 | /* Must have at least one goal */ | |
4336 | assert(g1 || g2); | |
4337 | ||
4338 | if (g1) { | |
4339 | ||
48ef279e KW |
4340 | /* Will never match if goal is out-of-bounds */ |
4341 | assert(! e1 || e1 >= g1); | |
8b35872c | 4342 | |
48ef279e KW |
4343 | /* Here, there isn't an end pointer, or it is beyond the goal. We |
4344 | * only go as far as the goal */ | |
4345 | e1 = g1; | |
8b35872c | 4346 | } |
313b38e5 NC |
4347 | else { |
4348 | assert(e1); /* Must have an end for looking at s1 */ | |
4349 | } | |
8b35872c KW |
4350 | |
4351 | /* Same for goal for s2 */ | |
4352 | if (g2) { | |
48ef279e KW |
4353 | assert(! e2 || e2 >= g2); |
4354 | e2 = g2; | |
8b35872c | 4355 | } |
313b38e5 NC |
4356 | else { |
4357 | assert(e2); | |
4358 | } | |
8b35872c | 4359 | |
18f762c3 KW |
4360 | /* If both operands are already folded, we could just do a memEQ on the |
4361 | * whole strings at once, but it would be better if the caller realized | |
4362 | * this and didn't even call us */ | |
4363 | ||
8b35872c KW |
4364 | /* Look through both strings, a character at a time */ |
4365 | while (p1 < e1 && p2 < e2) { | |
4366 | ||
d51c1b21 | 4367 | /* If at the beginning of a new character in s1, get its fold to use |
1d39b2cd | 4368 | * and the length of the fold. */ |
48ef279e | 4369 | if (n1 == 0) { |
18f762c3 KW |
4370 | if (flags & FOLDEQ_S1_ALREADY_FOLDED) { |
4371 | f1 = (U8 *) p1; | |
aa8ebe62 | 4372 | assert(u1); |
18f762c3 | 4373 | n1 = UTF8SKIP(f1); |
18f762c3 KW |
4374 | } |
4375 | else { | |
1d39b2cd KW |
4376 | if (isASCII(*p1) && ! (flags & FOLDEQ_LOCALE)) { |
4377 | ||
4378 | /* We have to forbid mixing ASCII with non-ASCII if the | |
4379 | * flags so indicate. And, we can short circuit having to | |
4380 | * call the general functions for this common ASCII case, | |
4381 | * all of whose non-locale folds are also ASCII, and hence | |
4382 | * UTF-8 invariants, so the UTF8ness of the strings is not | |
4383 | * relevant. */ | |
4384 | if ((flags & FOLDEQ_UTF8_NOMIX_ASCII) && ! isASCII(*p2)) { | |
4385 | return 0; | |
4386 | } | |
4387 | n1 = 1; | |
4388 | *foldbuf1 = toFOLD(*p1); | |
4389 | } | |
4390 | else if (u1) { | |
4391 | _to_utf8_fold_flags(p1, foldbuf1, &n1, flags_for_folder); | |
4392 | } | |
4a4088c4 | 4393 | else { /* Not UTF-8, get UTF-8 fold */ |
1d39b2cd KW |
4394 | _to_uni_fold_flags(*p1, foldbuf1, &n1, flags_for_folder); |
4395 | } | |
4396 | f1 = foldbuf1; | |
4397 | } | |
48ef279e | 4398 | } |
8b35872c | 4399 | |
48ef279e | 4400 | if (n2 == 0) { /* Same for s2 */ |
18f762c3 KW |
4401 | if (flags & FOLDEQ_S2_ALREADY_FOLDED) { |
4402 | f2 = (U8 *) p2; | |
aa8ebe62 | 4403 | assert(u2); |
18f762c3 KW |
4404 | n2 = UTF8SKIP(f2); |
4405 | } | |
4406 | else { | |
1d39b2cd KW |
4407 | if (isASCII(*p2) && ! (flags & FOLDEQ_LOCALE)) { |
4408 | if ((flags & FOLDEQ_UTF8_NOMIX_ASCII) && ! isASCII(*p1)) { | |
4409 | return 0; | |
4410 | } | |
4411 | n2 = 1; | |
4412 | *foldbuf2 = toFOLD(*p2); | |
4413 | } | |
4414 | else if (u2) { | |
4415 | _to_utf8_fold_flags(p2, foldbuf2, &n2, flags_for_folder); | |
4416 | } | |
4417 | else { | |
4418 | _to_uni_fold_flags(*p2, foldbuf2, &n2, flags_for_folder); | |
4419 | } | |
4420 | f2 = foldbuf2; | |
18f762c3 | 4421 | } |
48ef279e | 4422 | } |
8b35872c | 4423 | |
5001101e | 4424 | /* Here f1 and f2 point to the beginning of the strings to compare. |
227968da | 4425 | * These strings are the folds of the next character from each input |
4a4088c4 | 4426 | * string, stored in UTF-8. */ |
5e64d0fa | 4427 | |
48ef279e KW |
4428 | /* While there is more to look for in both folds, see if they |
4429 | * continue to match */ | |
4430 | while (n1 && n2) { | |
4431 | U8 fold_length = UTF8SKIP(f1); | |
4432 | if (fold_length != UTF8SKIP(f2) | |
4433 | || (fold_length == 1 && *f1 != *f2) /* Short circuit memNE | |
4434 | function call for single | |
a6d5f321 | 4435 | byte */ |
48ef279e KW |
4436 | || memNE((char*)f1, (char*)f2, fold_length)) |
4437 | { | |
e6226b18 | 4438 | return 0; /* mismatch */ |
48ef279e KW |
4439 | } |
4440 | ||
4441 | /* Here, they matched, advance past them */ | |
4442 | n1 -= fold_length; | |
4443 | f1 += fold_length; | |
4444 | n2 -= fold_length; | |
4445 | f2 += fold_length; | |
4446 | } | |
8b35872c | 4447 | |
48ef279e KW |
4448 | /* When reach the end of any fold, advance the input past it */ |
4449 | if (n1 == 0) { | |
4450 | p1 += u1 ? UTF8SKIP(p1) : 1; | |
4451 | } | |
4452 | if (n2 == 0) { | |
4453 | p2 += u2 ? UTF8SKIP(p2) : 1; | |
4454 | } | |
8b35872c KW |
4455 | } /* End of loop through both strings */ |
4456 | ||
4457 | /* A match is defined by each scan that specified an explicit length | |
4458 | * reaching its final goal, and the other not having matched a partial | |
4459 | * character (which can happen when the fold of a character is more than one | |
4460 | * character). */ | |
4461 | if (! ((g1 == 0 || p1 == g1) && (g2 == 0 || p2 == g2)) || n1 || n2) { | |
e6226b18 | 4462 | return 0; |
8b35872c KW |
4463 | } |
4464 | ||
4465 | /* Successful match. Set output pointers */ | |
4466 | if (pe1) { | |
48ef279e | 4467 | *pe1 = (char*)p1; |
8b35872c KW |
4468 | } |
4469 | if (pe2) { | |
48ef279e | 4470 | *pe2 = (char*)p2; |
8b35872c | 4471 | } |
e6226b18 | 4472 | return 1; |
e6b2e755 | 4473 | } |
701a277b | 4474 | |
f2645549 | 4475 | /* XXX The next two functions should likely be moved to mathoms.c once all |
37e7596b KW |
4476 | * occurrences of them are removed from the core; some cpan-upstream modules |
4477 | * still use them */ | |
4478 | ||
4479 | U8 * | |
4480 | Perl_uvuni_to_utf8(pTHX_ U8 *d, UV uv) | |
4481 | { | |
4482 | PERL_ARGS_ASSERT_UVUNI_TO_UTF8; | |
4483 | ||
4484 | return Perl_uvoffuni_to_utf8_flags(aTHX_ d, uv, 0); | |
4485 | } | |
4486 | ||
e505af10 KW |
4487 | /* |
4488 | =for apidoc utf8n_to_uvuni | |
4489 | ||
4490 | Instead use L</utf8_to_uvchr_buf>, or rarely, L</utf8n_to_uvchr>. | |
4491 | ||
4492 | This function was useful for code that wanted to handle both EBCDIC and | |
4493 | ASCII platforms with Unicode properties, but starting in Perl v5.20, the | |
4494 | distinctions between the platforms have mostly been made invisible to most | |
4495 | code, so this function is quite unlikely to be what you want. If you do need | |
4496 | this precise functionality, use instead | |
4497 | C<L<NATIVE_TO_UNI(utf8_to_uvchr_buf(...))|/utf8_to_uvchr_buf>> | |
4498 | or C<L<NATIVE_TO_UNI(utf8n_to_uvchr(...))|/utf8n_to_uvchr>>. | |
4499 | ||
4500 | =cut | |
4501 | */ | |
4502 | ||
37e7596b KW |
4503 | UV |
4504 | Perl_utf8n_to_uvuni(pTHX_ const U8 *s, STRLEN curlen, STRLEN *retlen, U32 flags) | |
4505 | { | |
4506 | PERL_ARGS_ASSERT_UTF8N_TO_UVUNI; | |
4507 | ||
4508 | return NATIVE_TO_UNI(utf8n_to_uvchr(s, curlen, retlen, flags)); | |
4509 | } | |
4510 | ||
4511 | /* | |
4512 | =for apidoc uvuni_to_utf8_flags | |
4513 | ||
4514 | Instead you almost certainly want to use L</uvchr_to_utf8> or | |
efa9cd84 | 4515 | L</uvchr_to_utf8_flags>. |
37e7596b KW |
4516 | |
4517 | This function is a deprecated synonym for L</uvoffuni_to_utf8_flags>, | |
4518 | which itself, while not deprecated, should be used only in isolated | |
4519 | circumstances. These functions were useful for code that wanted to handle | |
4520 | both EBCDIC and ASCII platforms with Unicode properties, but starting in Perl | |
4521 | v5.20, the distinctions between the platforms have mostly been made invisible | |
4522 | to most code, so this function is quite unlikely to be what you want. | |
4523 | ||
4524 | =cut | |
4525 | */ | |
4526 | ||
4527 | U8 * | |
4528 | Perl_uvuni_to_utf8_flags(pTHX_ U8 *d, UV uv, UV flags) | |
4529 | { | |
4530 | PERL_ARGS_ASSERT_UVUNI_TO_UTF8_FLAGS; | |
4531 | ||
4532 | return uvoffuni_to_utf8_flags(d, uv, flags); | |
4533 | } | |
4534 | ||
4535 | /* | |
14d04a33 | 4536 | * ex: set ts=8 sts=4 sw=4 et: |
37442d52 | 4537 | */ |