Commit | Line | Data |
---|---|---|
a0ed51b3 LW |
1 | /* utf8.c |
2 | * | |
1129b882 | 3 | * Copyright (C) 2000, 2001, 2002, 2003, 2004, 2005, 2006, 2007, 2008 |
b94e2f88 | 4 | * by Larry Wall and others |
a0ed51b3 LW |
5 | * |
6 | * You may distribute under the terms of either the GNU General Public | |
7 | * License or the Artistic License, as specified in the README file. | |
8 | * | |
9 | */ | |
10 | ||
11 | /* | |
4ac71550 TC |
12 | * 'What a fix!' said Sam. 'That's the one place in all the lands we've ever |
13 | * heard of that we don't want to see any closer; and that's the one place | |
14 | * we're trying to get to! And that's just where we can't get, nohow.' | |
15 | * | |
cdad3b53 | 16 | * [p.603 of _The Lord of the Rings_, IV/I: "The Taming of Sméagol"] |
a0ed51b3 LW |
17 | * |
18 | * 'Well do I understand your speech,' he answered in the same language; | |
19 | * 'yet few strangers do so. Why then do you not speak in the Common Tongue, | |
4ac71550 | 20 | * as is the custom in the West, if you wish to be answered?' |
cdad3b53 | 21 | * --Gandalf, addressing Théoden's door wardens |
4ac71550 TC |
22 | * |
23 | * [p.508 of _The Lord of the Rings_, III/vi: "The King of the Golden Hall"] | |
a0ed51b3 LW |
24 | * |
25 | * ...the travellers perceived that the floor was paved with stones of many | |
26 | * hues; branching runes and strange devices intertwined beneath their feet. | |
4ac71550 TC |
27 | * |
28 | * [p.512 of _The Lord of the Rings_, III/vi: "The King of the Golden Hall"] | |
a0ed51b3 LW |
29 | */ |
30 | ||
31 | #include "EXTERN.h" | |
864dbfa3 | 32 | #define PERL_IN_UTF8_C |
a0ed51b3 | 33 | #include "perl.h" |
b992490d | 34 | #include "invlist_inline.h" |
a0ed51b3 | 35 | |
806547a7 | 36 | static const char malformed_text[] = "Malformed UTF-8 character"; |
27da23d5 | 37 | static const char unees[] = |
806547a7 | 38 | "Malformed UTF-8 character (unexpected end of string)"; |
760c7c2f | 39 | static const char cp_above_legal_max[] = |
147e3846 | 40 | "Use of code point 0x%" UVXf " is deprecated; the permissible max is 0x%" UVXf; |
760c7c2f | 41 | |
114d9c4d | 42 | #define MAX_NON_DEPRECATED_CP ((UV) (IV_MAX)) |
901b21bf | 43 | |
48ef279e | 44 | /* |
ccfc67b7 | 45 | =head1 Unicode Support |
7fefc6c1 | 46 | These are various utility functions for manipulating UTF8-encoded |
72d33970 | 47 | strings. For the uninitiated, this is a method of representing arbitrary |
61296642 | 48 | Unicode characters as a variable number of bytes, in such a way that |
56da48f7 DM |
49 | characters in the ASCII range are unmodified, and a zero byte never appears |
50 | within non-zero characters. | |
166f8a29 | 51 | |
eaf7a4d2 CS |
52 | =cut |
53 | */ | |
54 | ||
9cbfb8ab KW |
55 | void |
56 | Perl__force_out_malformed_utf8_message(pTHX_ | |
57 | const U8 *const p, /* First byte in UTF-8 sequence */ | |
58 | const U8 * const e, /* Final byte in sequence (may include | |
59 | multiple chars */ | |
60 | const U32 flags, /* Flags to pass to utf8n_to_uvchr(), | |
61 | usually 0, or some DISALLOW flags */ | |
62 | const bool die_here) /* If TRUE, this function does not return */ | |
63 | { | |
64 | /* This core-only function is to be called when a malformed UTF-8 character | |
65 | * is found, in order to output the detailed information about the | |
66 | * malformation before dieing. The reason it exists is for the occasions | |
67 | * when such a malformation is fatal, but warnings might be turned off, so | |
68 | * that normally they would not be actually output. This ensures that they | |
69 | * do get output. Because a sequence may be malformed in more than one | |
70 | * way, multiple messages may be generated, so we can't make them fatal, as | |
71 | * that would cause the first one to die. | |
72 | * | |
73 | * Instead we pretend -W was passed to perl, then die afterwards. The | |
74 | * flexibility is here to return to the caller so they can finish up and | |
75 | * die themselves */ | |
76 | U32 errors; | |
77 | ||
78 | PERL_ARGS_ASSERT__FORCE_OUT_MALFORMED_UTF8_MESSAGE; | |
79 | ||
80 | ENTER; | |
c15a80f3 | 81 | SAVEI8(PL_dowarn); |
9cbfb8ab KW |
82 | SAVESPTR(PL_curcop); |
83 | ||
84 | PL_dowarn = G_WARN_ALL_ON|G_WARN_ON; | |
85 | if (PL_curcop) { | |
86 | PL_curcop->cop_warnings = pWARN_ALL; | |
87 | } | |
88 | ||
89 | (void) utf8n_to_uvchr_error(p, e - p, NULL, flags & ~UTF8_CHECK_ONLY, &errors); | |
90 | ||
91 | LEAVE; | |
92 | ||
93 | if (! errors) { | |
94 | Perl_croak(aTHX_ "panic: _force_out_malformed_utf8_message should" | |
95 | " be called only when there are errors found"); | |
96 | } | |
97 | ||
98 | if (die_here) { | |
99 | Perl_croak(aTHX_ "Malformed UTF-8 character (fatal)"); | |
100 | } | |
101 | } | |
102 | ||
eaf7a4d2 | 103 | /* |
378516de | 104 | =for apidoc uvoffuni_to_utf8_flags |
eebe1485 | 105 | |
a27992cc | 106 | THIS FUNCTION SHOULD BE USED IN ONLY VERY SPECIALIZED CIRCUMSTANCES. |
de69f3af KW |
107 | Instead, B<Almost all code should use L</uvchr_to_utf8> or |
108 | L</uvchr_to_utf8_flags>>. | |
a27992cc | 109 | |
de69f3af KW |
110 | This function is like them, but the input is a strict Unicode |
111 | (as opposed to native) code point. Only in very rare circumstances should code | |
112 | not be using the native code point. | |
949cf498 | 113 | |
efa9cd84 | 114 | For details, see the description for L</uvchr_to_utf8_flags>. |
949cf498 | 115 | |
eebe1485 SC |
116 | =cut |
117 | */ | |
118 | ||
8ee1cdcb KW |
119 | #define HANDLE_UNICODE_SURROGATE(uv, flags) \ |
120 | STMT_START { \ | |
121 | if (flags & UNICODE_WARN_SURROGATE) { \ | |
122 | Perl_ck_warner_d(aTHX_ packWARN(WARN_SURROGATE), \ | |
147e3846 | 123 | "UTF-16 surrogate U+%04" UVXf, uv); \ |
8ee1cdcb KW |
124 | } \ |
125 | if (flags & UNICODE_DISALLOW_SURROGATE) { \ | |
126 | return NULL; \ | |
127 | } \ | |
128 | } STMT_END; | |
129 | ||
130 | #define HANDLE_UNICODE_NONCHAR(uv, flags) \ | |
131 | STMT_START { \ | |
132 | if (flags & UNICODE_WARN_NONCHAR) { \ | |
133 | Perl_ck_warner_d(aTHX_ packWARN(WARN_NONCHAR), \ | |
147e3846 | 134 | "Unicode non-character U+%04" UVXf " is not " \ |
8ee1cdcb KW |
135 | "recommended for open interchange", uv); \ |
136 | } \ | |
137 | if (flags & UNICODE_DISALLOW_NONCHAR) { \ | |
138 | return NULL; \ | |
139 | } \ | |
140 | } STMT_END; | |
141 | ||
ba6ed43c KW |
142 | /* Use shorter names internally in this file */ |
143 | #define SHIFT UTF_ACCUMULATION_SHIFT | |
144 | #undef MARK | |
145 | #define MARK UTF_CONTINUATION_MARK | |
146 | #define MASK UTF_CONTINUATION_MASK | |
147 | ||
dfe13c55 | 148 | U8 * |
4b31b634 | 149 | Perl_uvoffuni_to_utf8_flags(pTHX_ U8 *d, UV uv, const UV flags) |
a0ed51b3 | 150 | { |
378516de | 151 | PERL_ARGS_ASSERT_UVOFFUNI_TO_UTF8_FLAGS; |
7918f24d | 152 | |
2d1545e5 | 153 | if (OFFUNI_IS_INVARIANT(uv)) { |
4c8cd605 | 154 | *d++ = LATIN1_TO_NATIVE(uv); |
d9432125 KW |
155 | return d; |
156 | } | |
facc1dc2 | 157 | |
3ea68d71 | 158 | if (uv <= MAX_UTF8_TWO_BYTE) { |
facc1dc2 KW |
159 | *d++ = I8_TO_NATIVE_UTF8(( uv >> SHIFT) | UTF_START_MARK(2)); |
160 | *d++ = I8_TO_NATIVE_UTF8(( uv & MASK) | MARK); | |
3ea68d71 KW |
161 | return d; |
162 | } | |
d9432125 | 163 | |
ba6ed43c KW |
164 | /* Not 2-byte; test for and handle 3-byte result. In the test immediately |
165 | * below, the 16 is for start bytes E0-EF (which are all the possible ones | |
166 | * for 3 byte characters). The 2 is for 2 continuation bytes; these each | |
167 | * contribute SHIFT bits. This yields 0x4000 on EBCDIC platforms, 0x1_0000 | |
168 | * on ASCII; so 3 bytes covers the range 0x400-0x3FFF on EBCDIC; | |
169 | * 0x800-0xFFFF on ASCII */ | |
170 | if (uv < (16 * (1U << (2 * SHIFT)))) { | |
171 | *d++ = I8_TO_NATIVE_UTF8(( uv >> ((3 - 1) * SHIFT)) | UTF_START_MARK(3)); | |
172 | *d++ = I8_TO_NATIVE_UTF8(((uv >> ((2 - 1) * SHIFT)) & MASK) | MARK); | |
173 | *d++ = I8_TO_NATIVE_UTF8(( uv /* (1 - 1) */ & MASK) | MARK); | |
174 | ||
175 | #ifndef EBCDIC /* These problematic code points are 4 bytes on EBCDIC, so | |
176 | aren't tested here */ | |
177 | /* The most likely code points in this range are below the surrogates. | |
178 | * Do an extra test to quickly exclude those. */ | |
179 | if (UNLIKELY(uv >= UNICODE_SURROGATE_FIRST)) { | |
180 | if (UNLIKELY( UNICODE_IS_32_CONTIGUOUS_NONCHARS(uv) | |
181 | || UNICODE_IS_END_PLANE_NONCHAR_GIVEN_NOT_SUPER(uv))) | |
182 | { | |
8ee1cdcb KW |
183 | HANDLE_UNICODE_NONCHAR(uv, flags); |
184 | } | |
185 | else if (UNLIKELY(UNICODE_IS_SURROGATE(uv))) { | |
186 | HANDLE_UNICODE_SURROGATE(uv, flags); | |
760c7c2f | 187 | } |
ba6ed43c KW |
188 | } |
189 | #endif | |
190 | return d; | |
191 | } | |
192 | ||
193 | /* Not 3-byte; that means the code point is at least 0x1_0000 on ASCII | |
194 | * platforms, and 0x4000 on EBCDIC. There are problematic cases that can | |
195 | * happen starting with 4-byte characters on ASCII platforms. We unify the | |
196 | * code for these with EBCDIC, even though some of them require 5-bytes on | |
197 | * those, because khw believes the code saving is worth the very slight | |
198 | * performance hit on these high EBCDIC code points. */ | |
199 | ||
200 | if (UNLIKELY(UNICODE_IS_SUPER(uv))) { | |
a5bf80e0 KW |
201 | if ( UNLIKELY(uv > MAX_NON_DEPRECATED_CP) |
202 | && ckWARN_d(WARN_DEPRECATED)) | |
203 | { | |
204 | Perl_warner(aTHX_ packWARN(WARN_DEPRECATED), | |
205 | cp_above_legal_max, uv, MAX_NON_DEPRECATED_CP); | |
206 | } | |
207 | if ( (flags & UNICODE_WARN_SUPER) | |
208 | || ( UNICODE_IS_ABOVE_31_BIT(uv) | |
209 | && (flags & UNICODE_WARN_ABOVE_31_BIT))) | |
210 | { | |
211 | Perl_ck_warner_d(aTHX_ packWARN(WARN_NON_UNICODE), | |
212 | ||
213 | /* Choose the more dire applicable warning */ | |
214 | (UNICODE_IS_ABOVE_31_BIT(uv)) | |
147e3846 KW |
215 | ? "Code point 0x%" UVXf " is not Unicode, and not portable" |
216 | : "Code point 0x%" UVXf " is not Unicode, may not be portable", | |
a5bf80e0 KW |
217 | uv); |
218 | } | |
219 | if (flags & UNICODE_DISALLOW_SUPER | |
220 | || ( UNICODE_IS_ABOVE_31_BIT(uv) | |
221 | && (flags & UNICODE_DISALLOW_ABOVE_31_BIT))) | |
222 | { | |
223 | return NULL; | |
224 | } | |
225 | } | |
ba6ed43c KW |
226 | else if (UNLIKELY(UNICODE_IS_END_PLANE_NONCHAR_GIVEN_NOT_SUPER(uv))) { |
227 | HANDLE_UNICODE_NONCHAR(uv, flags); | |
507b9800 | 228 | } |
d9432125 | 229 | |
ba6ed43c KW |
230 | /* Test for and handle 4-byte result. In the test immediately below, the |
231 | * 8 is for start bytes F0-F7 (which are all the possible ones for 4 byte | |
232 | * characters). The 3 is for 3 continuation bytes; these each contribute | |
233 | * SHIFT bits. This yields 0x4_0000 on EBCDIC platforms, 0x20_0000 on | |
234 | * ASCII, so 4 bytes covers the range 0x4000-0x3_FFFF on EBCDIC; | |
235 | * 0x1_0000-0x1F_FFFF on ASCII */ | |
236 | if (uv < (8 * (1U << (3 * SHIFT)))) { | |
237 | *d++ = I8_TO_NATIVE_UTF8(( uv >> ((4 - 1) * SHIFT)) | UTF_START_MARK(4)); | |
238 | *d++ = I8_TO_NATIVE_UTF8(((uv >> ((3 - 1) * SHIFT)) & MASK) | MARK); | |
239 | *d++ = I8_TO_NATIVE_UTF8(((uv >> ((2 - 1) * SHIFT)) & MASK) | MARK); | |
240 | *d++ = I8_TO_NATIVE_UTF8(( uv /* (1 - 1) */ & MASK) | MARK); | |
241 | ||
242 | #ifdef EBCDIC /* These were handled on ASCII platforms in the code for 3-byte | |
243 | characters. The end-plane non-characters for EBCDIC were | |
244 | handled just above */ | |
245 | if (UNLIKELY(UNICODE_IS_32_CONTIGUOUS_NONCHARS(uv))) { | |
246 | HANDLE_UNICODE_NONCHAR(uv, flags); | |
d528804a | 247 | } |
ba6ed43c KW |
248 | else if (UNLIKELY(UNICODE_IS_SURROGATE(uv))) { |
249 | HANDLE_UNICODE_SURROGATE(uv, flags); | |
250 | } | |
251 | #endif | |
252 | ||
253 | return d; | |
254 | } | |
255 | ||
256 | /* Not 4-byte; that means the code point is at least 0x20_0000 on ASCII | |
257 | * platforms, and 0x4000 on EBCDIC. At this point we switch to a loop | |
258 | * format. The unrolled version above turns out to not save all that much | |
259 | * time, and at these high code points (well above the legal Unicode range | |
260 | * on ASCII platforms, and well above anything in common use in EBCDIC), | |
261 | * khw believes that less code outweighs slight performance gains. */ | |
262 | ||
d9432125 | 263 | { |
5aaebcb3 | 264 | STRLEN len = OFFUNISKIP(uv); |
1d72bdf6 NIS |
265 | U8 *p = d+len-1; |
266 | while (p > d) { | |
4c8cd605 | 267 | *p-- = I8_TO_NATIVE_UTF8((uv & UTF_CONTINUATION_MASK) | UTF_CONTINUATION_MARK); |
1d72bdf6 NIS |
268 | uv >>= UTF_ACCUMULATION_SHIFT; |
269 | } | |
4c8cd605 | 270 | *p = I8_TO_NATIVE_UTF8((uv & UTF_START_MASK(len)) | UTF_START_MARK(len)); |
1d72bdf6 NIS |
271 | return d+len; |
272 | } | |
a0ed51b3 | 273 | } |
a5bf80e0 | 274 | |
646ca15d | 275 | /* |
07693fe6 KW |
276 | =for apidoc uvchr_to_utf8 |
277 | ||
bcb1a2d4 | 278 | Adds the UTF-8 representation of the native code point C<uv> to the end |
f2fc1b45 | 279 | of the string C<d>; C<d> should have at least C<UVCHR_SKIP(uv)+1> (up to |
c749c9fd KW |
280 | C<UTF8_MAXBYTES+1>) free bytes available. The return value is the pointer to |
281 | the byte after the end of the new character. In other words, | |
07693fe6 KW |
282 | |
283 | d = uvchr_to_utf8(d, uv); | |
284 | ||
285 | is the recommended wide native character-aware way of saying | |
286 | ||
287 | *(d++) = uv; | |
288 | ||
760c7c2f KW |
289 | This function accepts any UV as input, but very high code points (above |
290 | C<IV_MAX> on the platform) will raise a deprecation warning. This is | |
291 | typically 0x7FFF_FFFF in a 32-bit word. | |
292 | ||
293 | It is possible to forbid or warn on non-Unicode code points, or those that may | |
294 | be problematic by using L</uvchr_to_utf8_flags>. | |
de69f3af | 295 | |
07693fe6 KW |
296 | =cut |
297 | */ | |
298 | ||
de69f3af KW |
299 | /* This is also a macro */ |
300 | PERL_CALLCONV U8* Perl_uvchr_to_utf8(pTHX_ U8 *d, UV uv); | |
301 | ||
07693fe6 KW |
302 | U8 * |
303 | Perl_uvchr_to_utf8(pTHX_ U8 *d, UV uv) | |
304 | { | |
de69f3af | 305 | return uvchr_to_utf8(d, uv); |
07693fe6 KW |
306 | } |
307 | ||
de69f3af KW |
308 | /* |
309 | =for apidoc uvchr_to_utf8_flags | |
310 | ||
311 | Adds the UTF-8 representation of the native code point C<uv> to the end | |
f2fc1b45 | 312 | of the string C<d>; C<d> should have at least C<UVCHR_SKIP(uv)+1> (up to |
c749c9fd KW |
313 | C<UTF8_MAXBYTES+1>) free bytes available. The return value is the pointer to |
314 | the byte after the end of the new character. In other words, | |
de69f3af KW |
315 | |
316 | d = uvchr_to_utf8_flags(d, uv, flags); | |
317 | ||
318 | or, in most cases, | |
319 | ||
320 | d = uvchr_to_utf8_flags(d, uv, 0); | |
321 | ||
322 | This is the Unicode-aware way of saying | |
323 | ||
324 | *(d++) = uv; | |
325 | ||
760c7c2f KW |
326 | If C<flags> is 0, this function accepts any UV as input, but very high code |
327 | points (above C<IV_MAX> for the platform) will raise a deprecation warning. | |
328 | This is typically 0x7FFF_FFFF in a 32-bit word. | |
329 | ||
330 | Specifying C<flags> can further restrict what is allowed and not warned on, as | |
331 | follows: | |
de69f3af | 332 | |
796b6530 | 333 | If C<uv> is a Unicode surrogate code point and C<UNICODE_WARN_SURROGATE> is set, |
7ee537e6 KW |
334 | the function will raise a warning, provided UTF8 warnings are enabled. If |
335 | instead C<UNICODE_DISALLOW_SURROGATE> is set, the function will fail and return | |
336 | NULL. If both flags are set, the function will both warn and return NULL. | |
de69f3af | 337 | |
760c7c2f KW |
338 | Similarly, the C<UNICODE_WARN_NONCHAR> and C<UNICODE_DISALLOW_NONCHAR> flags |
339 | affect how the function handles a Unicode non-character. | |
93e6dbd6 | 340 | |
760c7c2f KW |
341 | And likewise, the C<UNICODE_WARN_SUPER> and C<UNICODE_DISALLOW_SUPER> flags |
342 | affect the handling of code points that are above the Unicode maximum of | |
343 | 0x10FFFF. Languages other than Perl may not be able to accept files that | |
344 | contain these. | |
93e6dbd6 KW |
345 | |
346 | The flag C<UNICODE_WARN_ILLEGAL_INTERCHANGE> selects all three of | |
347 | the above WARN flags; and C<UNICODE_DISALLOW_ILLEGAL_INTERCHANGE> selects all | |
ecc1615f KW |
348 | three DISALLOW flags. C<UNICODE_DISALLOW_ILLEGAL_INTERCHANGE> restricts the |
349 | allowed inputs to the strict UTF-8 traditionally defined by Unicode. | |
350 | Similarly, C<UNICODE_WARN_ILLEGAL_C9_INTERCHANGE> and | |
351 | C<UNICODE_DISALLOW_ILLEGAL_C9_INTERCHANGE> are shortcuts to select the | |
352 | above-Unicode and surrogate flags, but not the non-character ones, as | |
353 | defined in | |
354 | L<Unicode Corrigendum #9|http://www.unicode.org/versions/corrigendum9.html>. | |
355 | See L<perlunicode/Noncharacter code points>. | |
93e6dbd6 | 356 | |
ab8e6d41 KW |
357 | Code points above 0x7FFF_FFFF (2**31 - 1) were never specified in any standard, |
358 | so using them is more problematic than other above-Unicode code points. Perl | |
359 | invented an extension to UTF-8 to represent the ones above 2**36-1, so it is | |
360 | likely that non-Perl languages will not be able to read files that contain | |
361 | these that written by the perl interpreter; nor would Perl understand files | |
362 | written by something that uses a different extension. For these reasons, there | |
363 | is a separate set of flags that can warn and/or disallow these extremely high | |
364 | code points, even if other above-Unicode ones are accepted. These are the | |
760c7c2f KW |
365 | C<UNICODE_WARN_ABOVE_31_BIT> and C<UNICODE_DISALLOW_ABOVE_31_BIT> flags. These |
366 | are entirely independent from the deprecation warning for code points above | |
367 | C<IV_MAX>. On 32-bit machines, it will eventually be forbidden to have any | |
368 | code point that needs more than 31 bits to represent. When that happens, | |
369 | effectively the C<UNICODE_DISALLOW_ABOVE_31_BIT> flag will always be set on | |
370 | 32-bit machines. (Of course C<UNICODE_DISALLOW_SUPER> will treat all | |
ab8e6d41 KW |
371 | above-Unicode code points, including these, as malformations; and |
372 | C<UNICODE_WARN_SUPER> warns on these.) | |
373 | ||
374 | On EBCDIC platforms starting in Perl v5.24, the Perl extension for representing | |
375 | extremely high code points kicks in at 0x3FFF_FFFF (2**30 -1), which is lower | |
376 | than on ASCII. Prior to that, code points 2**31 and higher were simply | |
377 | unrepresentable, and a different, incompatible method was used to represent | |
378 | code points between 2**30 and 2**31 - 1. The flags C<UNICODE_WARN_ABOVE_31_BIT> | |
379 | and C<UNICODE_DISALLOW_ABOVE_31_BIT> have the same function as on ASCII | |
380 | platforms, warning and disallowing 2**31 and higher. | |
de69f3af | 381 | |
de69f3af KW |
382 | =cut |
383 | */ | |
384 | ||
385 | /* This is also a macro */ | |
386 | PERL_CALLCONV U8* Perl_uvchr_to_utf8_flags(pTHX_ U8 *d, UV uv, UV flags); | |
387 | ||
07693fe6 KW |
388 | U8 * |
389 | Perl_uvchr_to_utf8_flags(pTHX_ U8 *d, UV uv, UV flags) | |
390 | { | |
de69f3af | 391 | return uvchr_to_utf8_flags(d, uv, flags); |
07693fe6 KW |
392 | } |
393 | ||
83dc0f42 KW |
394 | PERL_STATIC_INLINE bool |
395 | S_is_utf8_cp_above_31_bits(const U8 * const s, const U8 * const e) | |
396 | { | |
397 | /* Returns TRUE if the first code point represented by the Perl-extended- | |
398 | * UTF-8-encoded string starting at 's', and looking no further than 'e - | |
399 | * 1' doesn't fit into 31 bytes. That is, that if it is >= 2**31. | |
400 | * | |
401 | * The function handles the case where the input bytes do not include all | |
402 | * the ones necessary to represent a full character. That is, they may be | |
403 | * the intial bytes of the representation of a code point, but possibly | |
404 | * the final ones necessary for the complete representation may be beyond | |
405 | * 'e - 1'. | |
406 | * | |
407 | * The function assumes that the sequence is well-formed UTF-8 as far as it | |
408 | * goes, and is for a UTF-8 variant code point. If the sequence is | |
409 | * incomplete, the function returns FALSE if there is any well-formed | |
410 | * UTF-8 byte sequence that can complete it in such a way that a code point | |
411 | * < 2**31 is produced; otherwise it returns TRUE. | |
412 | * | |
413 | * Getting this exactly right is slightly tricky, and has to be done in | |
414 | * several places in this file, so is centralized here. It is based on the | |
415 | * following table: | |
416 | * | |
417 | * U+7FFFFFFF (2 ** 31 - 1) | |
418 | * ASCII: \xFD\xBF\xBF\xBF\xBF\xBF | |
419 | * IBM-1047: \xFE\x41\x41\x41\x41\x41\x41\x42\x73\x73\x73\x73\x73\x73 | |
420 | * IBM-037: \xFE\x41\x41\x41\x41\x41\x41\x42\x72\x72\x72\x72\x72\x72 | |
421 | * POSIX-BC: \xFE\x41\x41\x41\x41\x41\x41\x42\x75\x75\x75\x75\x75\x75 | |
422 | * I8: \xFF\xA0\xA0\xA0\xA0\xA0\xA0\xA1\xBF\xBF\xBF\xBF\xBF\xBF | |
423 | * U+80000000 (2 ** 31): | |
424 | * ASCII: \xFE\x82\x80\x80\x80\x80\x80 | |
425 | * [0] [1] [2] [3] [4] [5] [6] [7] [8] [9] 10 11 12 13 | |
426 | * IBM-1047: \xFE\x41\x41\x41\x41\x41\x41\x43\x41\x41\x41\x41\x41\x41 | |
427 | * IBM-037: \xFE\x41\x41\x41\x41\x41\x41\x43\x41\x41\x41\x41\x41\x41 | |
428 | * POSIX-BC: \xFE\x41\x41\x41\x41\x41\x41\x43\x41\x41\x41\x41\x41\x41 | |
429 | * I8: \xFF\xA0\xA0\xA0\xA0\xA0\xA0\xA2\xA0\xA0\xA0\xA0\xA0\xA0 | |
430 | */ | |
431 | ||
432 | #ifdef EBCDIC | |
433 | ||
37086697 KW |
434 | /* [0] is start byte [1] [2] [3] [4] [5] [6] [7] */ |
435 | const U8 prefix[] = "\x41\x41\x41\x41\x41\x41\x42"; | |
83dc0f42 KW |
436 | const STRLEN prefix_len = sizeof(prefix) - 1; |
437 | const STRLEN len = e - s; | |
f880f78a | 438 | const STRLEN cmp_len = MIN(prefix_len, len - 1); |
83dc0f42 KW |
439 | |
440 | #else | |
441 | ||
442 | PERL_UNUSED_ARG(e); | |
443 | ||
444 | #endif | |
445 | ||
446 | PERL_ARGS_ASSERT_IS_UTF8_CP_ABOVE_31_BITS; | |
447 | ||
448 | assert(! UTF8_IS_INVARIANT(*s)); | |
449 | ||
450 | #ifndef EBCDIC | |
451 | ||
452 | /* Technically, a start byte of FE can be for a code point that fits into | |
453 | * 31 bytes, but not for well-formed UTF-8: doing that requires an overlong | |
454 | * malformation. */ | |
455 | return (*s >= 0xFE); | |
456 | ||
457 | #else | |
458 | ||
459 | /* On the EBCDIC code pages we handle, only 0xFE can mean a 32-bit or | |
460 | * larger code point (0xFF is an invariant). For 0xFE, we need at least 2 | |
461 | * bytes, and maybe up through 8 bytes, to be sure if the value is above 31 | |
462 | * bits. */ | |
463 | if (*s != 0xFE || len == 1) { | |
464 | return FALSE; | |
465 | } | |
466 | ||
467 | /* Note that in UTF-EBCDIC, the two lowest possible continuation bytes are | |
468 | * \x41 and \x42. */ | |
469 | return cBOOL(memGT(s + 1, prefix, cmp_len)); | |
470 | ||
471 | #endif | |
472 | ||
473 | } | |
474 | ||
12a4bed3 KW |
475 | PERL_STATIC_INLINE bool |
476 | S_does_utf8_overflow(const U8 * const s, const U8 * e) | |
477 | { | |
478 | const U8 *x; | |
479 | const U8 * y = (const U8 *) HIGHEST_REPRESENTABLE_UTF8; | |
480 | ||
c551bb01 KW |
481 | #if ! defined(UV_IS_QUAD) && ! defined(EBCDIC) |
482 | ||
483 | const STRLEN len = e - s; | |
484 | ||
485 | #endif | |
486 | ||
12a4bed3 KW |
487 | /* Returns a boolean as to if this UTF-8 string would overflow a UV on this |
488 | * platform, that is if it represents a code point larger than the highest | |
489 | * representable code point. (For ASCII platforms, we could use memcmp() | |
490 | * because we don't have to convert each byte to I8, but it's very rare | |
491 | * input indeed that would approach overflow, so the loop below will likely | |
492 | * only get executed once. | |
493 | * | |
494 | * 'e' must not be beyond a full character. If it is less than a full | |
495 | * character, the function returns FALSE if there is any input beyond 'e' | |
496 | * that could result in a non-overflowing code point */ | |
497 | ||
498 | PERL_ARGS_ASSERT_DOES_UTF8_OVERFLOW; | |
b0b342d4 KW |
499 | assert(s <= e && s + UTF8SKIP(s) >= e); |
500 | ||
501 | #if ! defined(UV_IS_QUAD) && ! defined(EBCDIC) | |
502 | ||
503 | /* On 32 bit ASCII machines, many overlongs that start with FF don't | |
504 | * overflow */ | |
505 | ||
c551bb01 | 506 | if (isFF_OVERLONG(s, len)) { |
b0b342d4 KW |
507 | const U8 max_32_bit_overlong[] = "\xFF\x80\x80\x80\x80\x80\x80\x84"; |
508 | return memGE(s, max_32_bit_overlong, | |
c551bb01 | 509 | MIN(len, sizeof(max_32_bit_overlong) - 1)); |
b0b342d4 KW |
510 | } |
511 | ||
512 | #endif | |
12a4bed3 KW |
513 | |
514 | for (x = s; x < e; x++, y++) { | |
515 | ||
516 | /* If this byte is larger than the corresponding highest UTF-8 byte, it | |
517 | * overflows */ | |
518 | if (UNLIKELY(NATIVE_UTF8_TO_I8(*x) > *y)) { | |
519 | return TRUE; | |
520 | } | |
521 | ||
522 | /* If not the same as this byte, it must be smaller, doesn't overflow */ | |
523 | if (LIKELY(NATIVE_UTF8_TO_I8(*x) != *y)) { | |
524 | return FALSE; | |
525 | } | |
526 | } | |
527 | ||
528 | /* Got to the end and all bytes are the same. If the input is a whole | |
529 | * character, it doesn't overflow. And if it is a partial character, | |
530 | * there's not enough information to tell, so assume doesn't overflow */ | |
531 | return FALSE; | |
532 | } | |
533 | ||
534 | PERL_STATIC_INLINE bool | |
535 | S_is_utf8_overlong_given_start_byte_ok(const U8 * const s, const STRLEN len) | |
536 | { | |
537 | /* Overlongs can occur whenever the number of continuation bytes | |
538 | * changes. That means whenever the number of leading 1 bits in a start | |
539 | * byte increases from the next lower start byte. That happens for start | |
540 | * bytes C0, E0, F0, F8, FC, FE, and FF. On modern perls, the following | |
541 | * illegal start bytes have already been excluded, so don't need to be | |
542 | * tested here; | |
543 | * ASCII platforms: C0, C1 | |
544 | * EBCDIC platforms C0, C1, C2, C3, C4, E0 | |
545 | * | |
546 | * At least a second byte is required to determine if other sequences will | |
547 | * be an overlong. */ | |
548 | ||
549 | const U8 s0 = NATIVE_UTF8_TO_I8(s[0]); | |
550 | const U8 s1 = NATIVE_UTF8_TO_I8(s[1]); | |
551 | ||
552 | PERL_ARGS_ASSERT_IS_UTF8_OVERLONG_GIVEN_START_BYTE_OK; | |
553 | assert(len > 1 && UTF8_IS_START(*s)); | |
554 | ||
555 | /* Each platform has overlongs after the start bytes given above (expressed | |
556 | * in I8 for EBCDIC). What constitutes an overlong varies by platform, but | |
557 | * the logic is the same, except the E0 overlong has already been excluded | |
558 | * on EBCDIC platforms. The values below were found by manually | |
559 | * inspecting the UTF-8 patterns. See the tables in utf8.h and | |
560 | * utfebcdic.h. */ | |
561 | ||
562 | # ifdef EBCDIC | |
563 | # define F0_ABOVE_OVERLONG 0xB0 | |
564 | # define F8_ABOVE_OVERLONG 0xA8 | |
565 | # define FC_ABOVE_OVERLONG 0xA4 | |
566 | # define FE_ABOVE_OVERLONG 0xA2 | |
567 | # define FF_OVERLONG_PREFIX "\xfe\x41\x41\x41\x41\x41\x41\x41" | |
568 | /* I8(0xfe) is FF */ | |
569 | # else | |
570 | ||
571 | if (s0 == 0xE0 && UNLIKELY(s1 < 0xA0)) { | |
572 | return TRUE; | |
573 | } | |
574 | ||
575 | # define F0_ABOVE_OVERLONG 0x90 | |
576 | # define F8_ABOVE_OVERLONG 0x88 | |
577 | # define FC_ABOVE_OVERLONG 0x84 | |
578 | # define FE_ABOVE_OVERLONG 0x82 | |
579 | # define FF_OVERLONG_PREFIX "\xff\x80\x80\x80\x80\x80\x80" | |
580 | # endif | |
581 | ||
582 | ||
583 | if ( (s0 == 0xF0 && UNLIKELY(s1 < F0_ABOVE_OVERLONG)) | |
584 | || (s0 == 0xF8 && UNLIKELY(s1 < F8_ABOVE_OVERLONG)) | |
585 | || (s0 == 0xFC && UNLIKELY(s1 < FC_ABOVE_OVERLONG)) | |
586 | || (s0 == 0xFE && UNLIKELY(s1 < FE_ABOVE_OVERLONG))) | |
587 | { | |
588 | return TRUE; | |
589 | } | |
590 | ||
b0b342d4 KW |
591 | /* Check for the FF overlong */ |
592 | return isFF_OVERLONG(s, len); | |
593 | } | |
594 | ||
595 | PERL_STATIC_INLINE bool | |
596 | S_isFF_OVERLONG(const U8 * const s, const STRLEN len) | |
597 | { | |
598 | PERL_ARGS_ASSERT_ISFF_OVERLONG; | |
12a4bed3 KW |
599 | |
600 | /* Check for the FF overlong. This happens only if all these bytes match; | |
601 | * what comes after them doesn't matter. See tables in utf8.h, | |
b0b342d4 | 602 | * utfebcdic.h. */ |
12a4bed3 | 603 | |
b0b342d4 KW |
604 | return len >= sizeof(FF_OVERLONG_PREFIX) - 1 |
605 | && UNLIKELY(memEQ(s, FF_OVERLONG_PREFIX, | |
606 | sizeof(FF_OVERLONG_PREFIX) - 1)); | |
12a4bed3 KW |
607 | } |
608 | ||
609 | #undef F0_ABOVE_OVERLONG | |
610 | #undef F8_ABOVE_OVERLONG | |
611 | #undef FC_ABOVE_OVERLONG | |
612 | #undef FE_ABOVE_OVERLONG | |
613 | #undef FF_OVERLONG_PREFIX | |
614 | ||
35f8c9bd | 615 | STRLEN |
edc2c47a | 616 | Perl__is_utf8_char_helper(const U8 * const s, const U8 * e, const U32 flags) |
35f8c9bd | 617 | { |
2b479609 | 618 | STRLEN len; |
12a4bed3 | 619 | const U8 *x; |
35f8c9bd | 620 | |
2b479609 KW |
621 | /* A helper function that should not be called directly. |
622 | * | |
623 | * This function returns non-zero if the string beginning at 's' and | |
624 | * looking no further than 'e - 1' is well-formed Perl-extended-UTF-8 for a | |
625 | * code point; otherwise it returns 0. The examination stops after the | |
626 | * first code point in 's' is validated, not looking at the rest of the | |
627 | * input. If 'e' is such that there are not enough bytes to represent a | |
628 | * complete code point, this function will return non-zero anyway, if the | |
629 | * bytes it does have are well-formed UTF-8 as far as they go, and aren't | |
630 | * excluded by 'flags'. | |
631 | * | |
632 | * A non-zero return gives the number of bytes required to represent the | |
633 | * code point. Be aware that if the input is for a partial character, the | |
634 | * return will be larger than 'e - s'. | |
635 | * | |
636 | * This function assumes that the code point represented is UTF-8 variant. | |
637 | * The caller should have excluded this possibility before calling this | |
638 | * function. | |
639 | * | |
640 | * 'flags' can be 0, or any combination of the UTF8_DISALLOW_foo flags | |
641 | * accepted by L</utf8n_to_uvchr>. If non-zero, this function will return | |
642 | * 0 if the code point represented is well-formed Perl-extended-UTF-8, but | |
643 | * disallowed by the flags. If the input is only for a partial character, | |
644 | * the function will return non-zero if there is any sequence of | |
645 | * well-formed UTF-8 that, when appended to the input sequence, could | |
646 | * result in an allowed code point; otherwise it returns 0. Non characters | |
647 | * cannot be determined based on partial character input. But many of the | |
648 | * other excluded types can be determined with just the first one or two | |
649 | * bytes. | |
650 | * | |
651 | */ | |
652 | ||
653 | PERL_ARGS_ASSERT__IS_UTF8_CHAR_HELPER; | |
654 | ||
655 | assert(0 == (flags & ~(UTF8_DISALLOW_ILLEGAL_INTERCHANGE | |
656 | |UTF8_DISALLOW_ABOVE_31_BIT))); | |
657 | assert(! UTF8_IS_INVARIANT(*s)); | |
35f8c9bd | 658 | |
2b479609 | 659 | /* A variant char must begin with a start byte */ |
35f8c9bd KW |
660 | if (UNLIKELY(! UTF8_IS_START(*s))) { |
661 | return 0; | |
662 | } | |
663 | ||
edc2c47a KW |
664 | /* Examine a maximum of a single whole code point */ |
665 | if (e - s > UTF8SKIP(s)) { | |
666 | e = s + UTF8SKIP(s); | |
667 | } | |
668 | ||
2b479609 KW |
669 | len = e - s; |
670 | ||
671 | if (flags && isUTF8_POSSIBLY_PROBLEMATIC(*s)) { | |
672 | const U8 s0 = NATIVE_UTF8_TO_I8(s[0]); | |
35f8c9bd | 673 | |
2b479609 KW |
674 | /* The code below is derived from this table. Keep in mind that legal |
675 | * continuation bytes range between \x80..\xBF for UTF-8, and | |
676 | * \xA0..\xBF for I8. Anything above those aren't continuation bytes. | |
677 | * Hence, we don't have to test the upper edge because if any of those | |
678 | * are encountered, the sequence is malformed, and will fail elsewhere | |
679 | * in this function. | |
680 | * UTF-8 UTF-EBCDIC I8 | |
681 | * U+D800: \xED\xA0\x80 \xF1\xB6\xA0\xA0 First surrogate | |
682 | * U+DFFF: \xED\xBF\xBF \xF1\xB7\xBF\xBF Final surrogate | |
683 | * U+110000: \xF4\x90\x80\x80 \xF9\xA2\xA0\xA0\xA0 First above Unicode | |
684 | * | |
685 | */ | |
686 | ||
687 | #ifdef EBCDIC /* On EBCDIC, these are actually I8 bytes */ | |
688 | # define FIRST_START_BYTE_THAT_IS_DEFINITELY_SUPER 0xFA | |
19794540 | 689 | # define IS_UTF8_2_BYTE_SUPER(s0, s1) ((s0) == 0xF9 && (s1) >= 0xA2) |
2b479609 | 690 | |
19794540 KW |
691 | # define IS_UTF8_2_BYTE_SURROGATE(s0, s1) ((s0) == 0xF1 \ |
692 | /* B6 and B7 */ \ | |
693 | && ((s1) & 0xFE ) == 0xB6) | |
2b479609 KW |
694 | #else |
695 | # define FIRST_START_BYTE_THAT_IS_DEFINITELY_SUPER 0xF5 | |
19794540 KW |
696 | # define IS_UTF8_2_BYTE_SUPER(s0, s1) ((s0) == 0xF4 && (s1) >= 0x90) |
697 | # define IS_UTF8_2_BYTE_SURROGATE(s0, s1) ((s0) == 0xED && (s1) >= 0xA0) | |
2b479609 KW |
698 | #endif |
699 | ||
700 | if ( (flags & UTF8_DISALLOW_SUPER) | |
ddb65933 KW |
701 | && UNLIKELY(s0 >= FIRST_START_BYTE_THAT_IS_DEFINITELY_SUPER)) |
702 | { | |
2b479609 KW |
703 | return 0; /* Above Unicode */ |
704 | } | |
705 | ||
706 | if ( (flags & UTF8_DISALLOW_ABOVE_31_BIT) | |
707 | && UNLIKELY(is_utf8_cp_above_31_bits(s, e))) | |
708 | { | |
709 | return 0; /* Above 31 bits */ | |
710 | } | |
711 | ||
712 | if (len > 1) { | |
713 | const U8 s1 = NATIVE_UTF8_TO_I8(s[1]); | |
714 | ||
715 | if ( (flags & UTF8_DISALLOW_SUPER) | |
19794540 | 716 | && UNLIKELY(IS_UTF8_2_BYTE_SUPER(s0, s1))) |
2b479609 KW |
717 | { |
718 | return 0; /* Above Unicode */ | |
719 | } | |
720 | ||
721 | if ( (flags & UTF8_DISALLOW_SURROGATE) | |
19794540 | 722 | && UNLIKELY(IS_UTF8_2_BYTE_SURROGATE(s0, s1))) |
2b479609 KW |
723 | { |
724 | return 0; /* Surrogate */ | |
725 | } | |
726 | ||
727 | if ( (flags & UTF8_DISALLOW_NONCHAR) | |
728 | && UNLIKELY(UTF8_IS_NONCHAR(s, e))) | |
729 | { | |
730 | return 0; /* Noncharacter code point */ | |
731 | } | |
732 | } | |
733 | } | |
734 | ||
735 | /* Make sure that all that follows are continuation bytes */ | |
35f8c9bd KW |
736 | for (x = s + 1; x < e; x++) { |
737 | if (UNLIKELY(! UTF8_IS_CONTINUATION(*x))) { | |
738 | return 0; | |
739 | } | |
740 | } | |
741 | ||
af13dd8a | 742 | /* Here is syntactically valid. Next, make sure this isn't the start of an |
12a4bed3 KW |
743 | * overlong. */ |
744 | if (len > 1 && is_utf8_overlong_given_start_byte_ok(s, len)) { | |
745 | return 0; | |
af13dd8a KW |
746 | } |
747 | ||
12a4bed3 KW |
748 | /* And finally, that the code point represented fits in a word on this |
749 | * platform */ | |
750 | if (does_utf8_overflow(s, e)) { | |
751 | return 0; | |
35f8c9bd KW |
752 | } |
753 | ||
2b479609 | 754 | return UTF8SKIP(s); |
35f8c9bd KW |
755 | } |
756 | ||
7cf8d05d KW |
757 | STATIC char * |
758 | S__byte_dump_string(pTHX_ const U8 * s, const STRLEN len) | |
759 | { | |
760 | /* Returns a mortalized C string that is a displayable copy of the 'len' | |
761 | * bytes starting at 's', each in a \xXY format. */ | |
762 | ||
763 | const STRLEN output_len = 4 * len + 1; /* 4 bytes per each input, plus a | |
764 | trailing NUL */ | |
765 | const U8 * const e = s + len; | |
766 | char * output; | |
767 | char * d; | |
768 | ||
769 | PERL_ARGS_ASSERT__BYTE_DUMP_STRING; | |
770 | ||
771 | Newx(output, output_len, char); | |
772 | SAVEFREEPV(output); | |
773 | ||
774 | d = output; | |
775 | for (; s < e; s++) { | |
776 | const unsigned high_nibble = (*s & 0xF0) >> 4; | |
777 | const unsigned low_nibble = (*s & 0x0F); | |
778 | ||
779 | *d++ = '\\'; | |
780 | *d++ = 'x'; | |
781 | ||
782 | if (high_nibble < 10) { | |
783 | *d++ = high_nibble + '0'; | |
784 | } | |
785 | else { | |
786 | *d++ = high_nibble - 10 + 'a'; | |
787 | } | |
788 | ||
789 | if (low_nibble < 10) { | |
790 | *d++ = low_nibble + '0'; | |
791 | } | |
792 | else { | |
793 | *d++ = low_nibble - 10 + 'a'; | |
794 | } | |
795 | } | |
796 | ||
797 | *d = '\0'; | |
798 | return output; | |
799 | } | |
800 | ||
806547a7 | 801 | PERL_STATIC_INLINE char * |
7cf8d05d KW |
802 | S_unexpected_non_continuation_text(pTHX_ const U8 * const s, |
803 | ||
804 | /* How many bytes to print */ | |
3cc6a05e | 805 | STRLEN print_len, |
7cf8d05d KW |
806 | |
807 | /* Which one is the non-continuation */ | |
808 | const STRLEN non_cont_byte_pos, | |
809 | ||
810 | /* How many bytes should there be? */ | |
811 | const STRLEN expect_len) | |
806547a7 KW |
812 | { |
813 | /* Return the malformation warning text for an unexpected continuation | |
814 | * byte. */ | |
815 | ||
7cf8d05d | 816 | const char * const where = (non_cont_byte_pos == 1) |
806547a7 | 817 | ? "immediately" |
7cf8d05d KW |
818 | : Perl_form(aTHX_ "%d bytes", |
819 | (int) non_cont_byte_pos); | |
806547a7 KW |
820 | |
821 | PERL_ARGS_ASSERT_UNEXPECTED_NON_CONTINUATION_TEXT; | |
822 | ||
7cf8d05d KW |
823 | /* We don't need to pass this parameter, but since it has already been |
824 | * calculated, it's likely faster to pass it; verify under DEBUGGING */ | |
825 | assert(expect_len == UTF8SKIP(s)); | |
826 | ||
827 | return Perl_form(aTHX_ "%s: %s (unexpected non-continuation byte 0x%02x," | |
828 | " %s after start byte 0x%02x; need %d bytes, got %d)", | |
829 | malformed_text, | |
830 | _byte_dump_string(s, print_len), | |
831 | *(s + non_cont_byte_pos), | |
832 | where, | |
833 | *s, | |
834 | (int) expect_len, | |
835 | (int) non_cont_byte_pos); | |
806547a7 KW |
836 | } |
837 | ||
35f8c9bd KW |
838 | /* |
839 | ||
de69f3af | 840 | =for apidoc utf8n_to_uvchr |
378516de KW |
841 | |
842 | THIS FUNCTION SHOULD BE USED IN ONLY VERY SPECIALIZED CIRCUMSTANCES. | |
de69f3af | 843 | Most code should use L</utf8_to_uvchr_buf>() rather than call this directly. |
67e989fb | 844 | |
9041c2e3 | 845 | Bottom level UTF-8 decode routine. |
de69f3af | 846 | Returns the native code point value of the first character in the string C<s>, |
746afd53 KW |
847 | which is assumed to be in UTF-8 (or UTF-EBCDIC) encoding, and no longer than |
848 | C<curlen> bytes; C<*retlen> (if C<retlen> isn't NULL) will be set to | |
849 | the length, in bytes, of that character. | |
949cf498 KW |
850 | |
851 | The value of C<flags> determines the behavior when C<s> does not point to a | |
2b5e7bc2 KW |
852 | well-formed UTF-8 character. If C<flags> is 0, encountering a malformation |
853 | causes zero to be returned and C<*retlen> is set so that (S<C<s> + C<*retlen>>) | |
854 | is the next possible position in C<s> that could begin a non-malformed | |
855 | character. Also, if UTF-8 warnings haven't been lexically disabled, a warning | |
856 | is raised. Some UTF-8 input sequences may contain multiple malformations. | |
857 | This function tries to find every possible one in each call, so multiple | |
858 | warnings can be raised for each sequence. | |
949cf498 KW |
859 | |
860 | Various ALLOW flags can be set in C<flags> to allow (and not warn on) | |
861 | individual types of malformations, such as the sequence being overlong (that | |
862 | is, when there is a shorter sequence that can express the same code point; | |
863 | overlong sequences are expressly forbidden in the UTF-8 standard due to | |
864 | potential security issues). Another malformation example is the first byte of | |
865 | a character not being a legal first byte. See F<utf8.h> for the list of such | |
94953955 KW |
866 | flags. Even if allowed, this function generally returns the Unicode |
867 | REPLACEMENT CHARACTER when it encounters a malformation. There are flags in | |
868 | F<utf8.h> to override this behavior for the overlong malformations, but don't | |
869 | do that except for very specialized purposes. | |
949cf498 | 870 | |
796b6530 | 871 | The C<UTF8_CHECK_ONLY> flag overrides the behavior when a non-allowed (by other |
949cf498 KW |
872 | flags) malformation is found. If this flag is set, the routine assumes that |
873 | the caller will raise a warning, and this function will silently just set | |
d088425d KW |
874 | C<retlen> to C<-1> (cast to C<STRLEN>) and return zero. |
875 | ||
75200dff | 876 | Note that this API requires disambiguation between successful decoding a C<NUL> |
796b6530 | 877 | character, and an error return (unless the C<UTF8_CHECK_ONLY> flag is set), as |
111fa700 KW |
878 | in both cases, 0 is returned, and, depending on the malformation, C<retlen> may |
879 | be set to 1. To disambiguate, upon a zero return, see if the first byte of | |
880 | C<s> is 0 as well. If so, the input was a C<NUL>; if not, the input had an | |
f9380377 | 881 | error. Or you can use C<L</utf8n_to_uvchr_error>>. |
949cf498 KW |
882 | |
883 | Certain code points are considered problematic. These are Unicode surrogates, | |
746afd53 | 884 | Unicode non-characters, and code points above the Unicode maximum of 0x10FFFF. |
949cf498 | 885 | By default these are considered regular code points, but certain situations |
ecc1615f KW |
886 | warrant special handling for them, which can be specified using the C<flags> |
887 | parameter. If C<flags> contains C<UTF8_DISALLOW_ILLEGAL_INTERCHANGE>, all | |
888 | three classes are treated as malformations and handled as such. The flags | |
889 | C<UTF8_DISALLOW_SURROGATE>, C<UTF8_DISALLOW_NONCHAR>, and | |
890 | C<UTF8_DISALLOW_SUPER> (meaning above the legal Unicode maximum) can be set to | |
891 | disallow these categories individually. C<UTF8_DISALLOW_ILLEGAL_INTERCHANGE> | |
892 | restricts the allowed inputs to the strict UTF-8 traditionally defined by | |
893 | Unicode. Use C<UTF8_DISALLOW_ILLEGAL_C9_INTERCHANGE> to use the strictness | |
894 | definition given by | |
895 | L<Unicode Corrigendum #9|http://www.unicode.org/versions/corrigendum9.html>. | |
896 | The difference between traditional strictness and C9 strictness is that the | |
897 | latter does not forbid non-character code points. (They are still discouraged, | |
898 | however.) For more discussion see L<perlunicode/Noncharacter code points>. | |
899 | ||
900 | The flags C<UTF8_WARN_ILLEGAL_INTERCHANGE>, | |
901 | C<UTF8_WARN_ILLEGAL_C9_INTERCHANGE>, C<UTF8_WARN_SURROGATE>, | |
796b6530 KW |
902 | C<UTF8_WARN_NONCHAR>, and C<UTF8_WARN_SUPER> will cause warning messages to be |
903 | raised for their respective categories, but otherwise the code points are | |
904 | considered valid (not malformations). To get a category to both be treated as | |
905 | a malformation and raise a warning, specify both the WARN and DISALLOW flags. | |
949cf498 | 906 | (But note that warnings are not raised if lexically disabled nor if |
796b6530 | 907 | C<UTF8_CHECK_ONLY> is also specified.) |
949cf498 | 908 | |
760c7c2f KW |
909 | It is now deprecated to have very high code points (above C<IV_MAX> on the |
910 | platforms) and this function will raise a deprecation warning for these (unless | |
d5944cab | 911 | such warnings are turned off). This value is typically 0x7FFF_FFFF (2**31 -1) |
760c7c2f | 912 | in a 32-bit word. |
ab8e6d41 KW |
913 | |
914 | Code points above 0x7FFF_FFFF (2**31 - 1) were never specified in any standard, | |
915 | so using them is more problematic than other above-Unicode code points. Perl | |
916 | invented an extension to UTF-8 to represent the ones above 2**36-1, so it is | |
917 | likely that non-Perl languages will not be able to read files that contain | |
aff2be59 | 918 | these; nor would Perl understand files |
ab8e6d41 KW |
919 | written by something that uses a different extension. For these reasons, there |
920 | is a separate set of flags that can warn and/or disallow these extremely high | |
921 | code points, even if other above-Unicode ones are accepted. These are the | |
760c7c2f KW |
922 | C<UTF8_WARN_ABOVE_31_BIT> and C<UTF8_DISALLOW_ABOVE_31_BIT> flags. These |
923 | are entirely independent from the deprecation warning for code points above | |
924 | C<IV_MAX>. On 32-bit machines, it will eventually be forbidden to have any | |
925 | code point that needs more than 31 bits to represent. When that happens, | |
926 | effectively the C<UTF8_DISALLOW_ABOVE_31_BIT> flag will always be set on | |
927 | 32-bit machines. (Of course C<UTF8_DISALLOW_SUPER> will treat all | |
ab8e6d41 KW |
928 | above-Unicode code points, including these, as malformations; and |
929 | C<UTF8_WARN_SUPER> warns on these.) | |
930 | ||
931 | On EBCDIC platforms starting in Perl v5.24, the Perl extension for representing | |
932 | extremely high code points kicks in at 0x3FFF_FFFF (2**30 -1), which is lower | |
933 | than on ASCII. Prior to that, code points 2**31 and higher were simply | |
934 | unrepresentable, and a different, incompatible method was used to represent | |
935 | code points between 2**30 and 2**31 - 1. The flags C<UTF8_WARN_ABOVE_31_BIT> | |
936 | and C<UTF8_DISALLOW_ABOVE_31_BIT> have the same function as on ASCII | |
937 | platforms, warning and disallowing 2**31 and higher. | |
949cf498 KW |
938 | |
939 | All other code points corresponding to Unicode characters, including private | |
940 | use and those yet to be assigned, are never considered malformed and never | |
941 | warn. | |
67e989fb | 942 | |
37607a96 | 943 | =cut |
f9380377 KW |
944 | |
945 | Also implemented as a macro in utf8.h | |
946 | */ | |
947 | ||
948 | UV | |
949 | Perl_utf8n_to_uvchr(pTHX_ const U8 *s, | |
950 | STRLEN curlen, | |
951 | STRLEN *retlen, | |
952 | const U32 flags) | |
953 | { | |
954 | PERL_ARGS_ASSERT_UTF8N_TO_UVCHR; | |
955 | ||
956 | return utf8n_to_uvchr_error(s, curlen, retlen, flags, NULL); | |
957 | } | |
958 | ||
959 | /* | |
960 | ||
961 | =for apidoc utf8n_to_uvchr_error | |
962 | ||
963 | THIS FUNCTION SHOULD BE USED IN ONLY VERY SPECIALIZED CIRCUMSTANCES. | |
964 | Most code should use L</utf8_to_uvchr_buf>() rather than call this directly. | |
965 | ||
966 | This function is for code that needs to know what the precise malformation(s) | |
967 | are when an error is found. | |
968 | ||
969 | It is like C<L</utf8n_to_uvchr>> but it takes an extra parameter placed after | |
970 | all the others, C<errors>. If this parameter is 0, this function behaves | |
971 | identically to C<L</utf8n_to_uvchr>>. Otherwise, C<errors> should be a pointer | |
972 | to a C<U32> variable, which this function sets to indicate any errors found. | |
973 | Upon return, if C<*errors> is 0, there were no errors found. Otherwise, | |
974 | C<*errors> is the bit-wise C<OR> of the bits described in the list below. Some | |
975 | of these bits will be set if a malformation is found, even if the input | |
7a65503b | 976 | C<flags> parameter indicates that the given malformation is allowed; those |
f9380377 KW |
977 | exceptions are noted: |
978 | ||
979 | =over 4 | |
980 | ||
981 | =item C<UTF8_GOT_ABOVE_31_BIT> | |
982 | ||
983 | The code point represented by the input UTF-8 sequence occupies more than 31 | |
984 | bits. | |
985 | This bit is set only if the input C<flags> parameter contains either the | |
986 | C<UTF8_DISALLOW_ABOVE_31_BIT> or the C<UTF8_WARN_ABOVE_31_BIT> flags. | |
987 | ||
988 | =item C<UTF8_GOT_CONTINUATION> | |
989 | ||
990 | The input sequence was malformed in that the first byte was a a UTF-8 | |
991 | continuation byte. | |
992 | ||
993 | =item C<UTF8_GOT_EMPTY> | |
994 | ||
995 | The input C<curlen> parameter was 0. | |
996 | ||
997 | =item C<UTF8_GOT_LONG> | |
998 | ||
999 | The input sequence was malformed in that there is some other sequence that | |
1000 | evaluates to the same code point, but that sequence is shorter than this one. | |
1001 | ||
1002 | =item C<UTF8_GOT_NONCHAR> | |
1003 | ||
1004 | The code point represented by the input UTF-8 sequence is for a Unicode | |
1005 | non-character code point. | |
1006 | This bit is set only if the input C<flags> parameter contains either the | |
1007 | C<UTF8_DISALLOW_NONCHAR> or the C<UTF8_WARN_NONCHAR> flags. | |
1008 | ||
1009 | =item C<UTF8_GOT_NON_CONTINUATION> | |
1010 | ||
1011 | The input sequence was malformed in that a non-continuation type byte was found | |
1012 | in a position where only a continuation type one should be. | |
1013 | ||
1014 | =item C<UTF8_GOT_OVERFLOW> | |
1015 | ||
1016 | The input sequence was malformed in that it is for a code point that is not | |
1017 | representable in the number of bits available in a UV on the current platform. | |
1018 | ||
1019 | =item C<UTF8_GOT_SHORT> | |
1020 | ||
1021 | The input sequence was malformed in that C<curlen> is smaller than required for | |
1022 | a complete sequence. In other words, the input is for a partial character | |
1023 | sequence. | |
1024 | ||
1025 | =item C<UTF8_GOT_SUPER> | |
1026 | ||
1027 | The input sequence was malformed in that it is for a non-Unicode code point; | |
1028 | that is, one above the legal Unicode maximum. | |
1029 | This bit is set only if the input C<flags> parameter contains either the | |
1030 | C<UTF8_DISALLOW_SUPER> or the C<UTF8_WARN_SUPER> flags. | |
1031 | ||
1032 | =item C<UTF8_GOT_SURROGATE> | |
1033 | ||
1034 | The input sequence was malformed in that it is for a -Unicode UTF-16 surrogate | |
1035 | code point. | |
1036 | This bit is set only if the input C<flags> parameter contains either the | |
1037 | C<UTF8_DISALLOW_SURROGATE> or the C<UTF8_WARN_SURROGATE> flags. | |
1038 | ||
1039 | =back | |
1040 | ||
133551d8 KW |
1041 | To do your own error handling, call this function with the C<UTF8_CHECK_ONLY> |
1042 | flag to suppress any warnings, and then examine the C<*errors> return. | |
1043 | ||
f9380377 | 1044 | =cut |
37607a96 | 1045 | */ |
67e989fb | 1046 | |
a0ed51b3 | 1047 | UV |
f9380377 KW |
1048 | Perl_utf8n_to_uvchr_error(pTHX_ const U8 *s, |
1049 | STRLEN curlen, | |
1050 | STRLEN *retlen, | |
1051 | const U32 flags, | |
1052 | U32 * errors) | |
a0ed51b3 | 1053 | { |
d4c19fe8 | 1054 | const U8 * const s0 = s; |
2b5e7bc2 KW |
1055 | U8 * send = NULL; /* (initialized to silence compilers' wrong |
1056 | warning) */ | |
1057 | U32 possible_problems = 0; /* A bit is set here for each potential problem | |
1058 | found as we go along */ | |
eb83ed87 | 1059 | UV uv = *s; |
2b5e7bc2 KW |
1060 | STRLEN expectlen = 0; /* How long should this sequence be? |
1061 | (initialized to silence compilers' wrong | |
1062 | warning) */ | |
e308b348 | 1063 | STRLEN avail_len = 0; /* When input is too short, gives what that is */ |
f9380377 KW |
1064 | U32 discard_errors = 0; /* Used to save branches when 'errors' is NULL; |
1065 | this gets set and discarded */ | |
a0dbb045 | 1066 | |
2b5e7bc2 KW |
1067 | /* The below are used only if there is both an overlong malformation and a |
1068 | * too short one. Otherwise the first two are set to 's0' and 'send', and | |
1069 | * the third not used at all */ | |
1070 | U8 * adjusted_s0 = (U8 *) s0; | |
5ec712b1 KW |
1071 | U8 * adjusted_send = NULL; /* (Initialized to silence compilers' wrong |
1072 | warning) */ | |
2b5e7bc2 | 1073 | UV uv_so_far = 0; /* (Initialized to silence compilers' wrong warning) */ |
7918f24d | 1074 | |
f9380377 KW |
1075 | PERL_ARGS_ASSERT_UTF8N_TO_UVCHR_ERROR; |
1076 | ||
1077 | if (errors) { | |
1078 | *errors = 0; | |
1079 | } | |
1080 | else { | |
1081 | errors = &discard_errors; | |
1082 | } | |
a0dbb045 | 1083 | |
eb83ed87 KW |
1084 | /* The order of malformation tests here is important. We should consume as |
1085 | * few bytes as possible in order to not skip any valid character. This is | |
1086 | * required by the Unicode Standard (section 3.9 of Unicode 6.0); see also | |
1087 | * http://unicode.org/reports/tr36 for more discussion as to why. For | |
1088 | * example, once we've done a UTF8SKIP, we can tell the expected number of | |
1089 | * bytes, and could fail right off the bat if the input parameters indicate | |
1090 | * that there are too few available. But it could be that just that first | |
1091 | * byte is garbled, and the intended character occupies fewer bytes. If we | |
1092 | * blindly assumed that the first byte is correct, and skipped based on | |
1093 | * that number, we could skip over a valid input character. So instead, we | |
1094 | * always examine the sequence byte-by-byte. | |
1095 | * | |
1096 | * We also should not consume too few bytes, otherwise someone could inject | |
1097 | * things. For example, an input could be deliberately designed to | |
1098 | * overflow, and if this code bailed out immediately upon discovering that, | |
e2660c54 | 1099 | * returning to the caller C<*retlen> pointing to the very next byte (one |
eb83ed87 KW |
1100 | * which is actually part of of the overflowing sequence), that could look |
1101 | * legitimate to the caller, which could discard the initial partial | |
2b5e7bc2 KW |
1102 | * sequence and process the rest, inappropriately. |
1103 | * | |
1104 | * Some possible input sequences are malformed in more than one way. This | |
1105 | * function goes to lengths to try to find all of them. This is necessary | |
1106 | * for correctness, as the inputs may allow one malformation but not | |
1107 | * another, and if we abandon searching for others after finding the | |
1108 | * allowed one, we could allow in something that shouldn't have been. | |
1109 | */ | |
eb83ed87 | 1110 | |
b5b9af04 | 1111 | if (UNLIKELY(curlen == 0)) { |
2b5e7bc2 KW |
1112 | possible_problems |= UTF8_GOT_EMPTY; |
1113 | curlen = 0; | |
5a48568d | 1114 | uv = UNICODE_REPLACEMENT; |
2b5e7bc2 | 1115 | goto ready_to_handle_errors; |
0c443dc2 JH |
1116 | } |
1117 | ||
eb83ed87 KW |
1118 | expectlen = UTF8SKIP(s); |
1119 | ||
1120 | /* A well-formed UTF-8 character, as the vast majority of calls to this | |
1121 | * function will be for, has this expected length. For efficiency, set | |
1122 | * things up here to return it. It will be overriden only in those rare | |
1123 | * cases where a malformation is found */ | |
1124 | if (retlen) { | |
1125 | *retlen = expectlen; | |
1126 | } | |
1127 | ||
1128 | /* An invariant is trivially well-formed */ | |
1d72bdf6 | 1129 | if (UTF8_IS_INVARIANT(uv)) { |
de69f3af | 1130 | return uv; |
a0ed51b3 | 1131 | } |
67e989fb | 1132 | |
eb83ed87 | 1133 | /* A continuation character can't start a valid sequence */ |
b5b9af04 | 1134 | if (UNLIKELY(UTF8_IS_CONTINUATION(uv))) { |
2b5e7bc2 KW |
1135 | possible_problems |= UTF8_GOT_CONTINUATION; |
1136 | curlen = 1; | |
1137 | uv = UNICODE_REPLACEMENT; | |
1138 | goto ready_to_handle_errors; | |
ba210ebe | 1139 | } |
9041c2e3 | 1140 | |
dcd27b3c | 1141 | /* Here is not a continuation byte, nor an invariant. The only thing left |
ddb65933 KW |
1142 | * is a start byte (possibly for an overlong). (We can't use UTF8_IS_START |
1143 | * because it excludes start bytes like \xC0 that always lead to | |
1144 | * overlongs.) */ | |
dcd27b3c | 1145 | |
534752c1 KW |
1146 | /* Convert to I8 on EBCDIC (no-op on ASCII), then remove the leading bits |
1147 | * that indicate the number of bytes in the character's whole UTF-8 | |
1148 | * sequence, leaving just the bits that are part of the value. */ | |
1149 | uv = NATIVE_UTF8_TO_I8(uv) & UTF_START_MASK(expectlen); | |
ba210ebe | 1150 | |
e308b348 KW |
1151 | /* Setup the loop end point, making sure to not look past the end of the |
1152 | * input string, and flag it as too short if the size isn't big enough. */ | |
1153 | send = (U8*) s0; | |
1154 | if (UNLIKELY(curlen < expectlen)) { | |
1155 | possible_problems |= UTF8_GOT_SHORT; | |
1156 | avail_len = curlen; | |
1157 | send += curlen; | |
1158 | } | |
1159 | else { | |
1160 | send += expectlen; | |
1161 | } | |
1162 | adjusted_send = send; | |
1163 | ||
eb83ed87 | 1164 | /* Now, loop through the remaining bytes in the character's sequence, |
e308b348 | 1165 | * accumulating each into the working value as we go. */ |
eb83ed87 | 1166 | for (s = s0 + 1; s < send; s++) { |
b5b9af04 | 1167 | if (LIKELY(UTF8_IS_CONTINUATION(*s))) { |
8850bf83 | 1168 | uv = UTF8_ACCUMULATE(uv, *s); |
2b5e7bc2 KW |
1169 | continue; |
1170 | } | |
1171 | ||
1172 | /* Here, found a non-continuation before processing all expected bytes. | |
1173 | * This byte indicates the beginning of a new character, so quit, even | |
1174 | * if allowing this malformation. */ | |
2b5e7bc2 | 1175 | possible_problems |= UTF8_GOT_NON_CONTINUATION; |
e308b348 | 1176 | break; |
eb83ed87 KW |
1177 | } /* End of loop through the character's bytes */ |
1178 | ||
1179 | /* Save how many bytes were actually in the character */ | |
1180 | curlen = s - s0; | |
1181 | ||
e308b348 KW |
1182 | /* A convenience macro that matches either of the too-short conditions. */ |
1183 | # define UTF8_GOT_TOO_SHORT (UTF8_GOT_SHORT|UTF8_GOT_NON_CONTINUATION) | |
2f8f112e | 1184 | |
e308b348 | 1185 | if (UNLIKELY(possible_problems & UTF8_GOT_TOO_SHORT)) { |
2b5e7bc2 KW |
1186 | uv_so_far = uv; |
1187 | uv = UNICODE_REPLACEMENT; | |
eb83ed87 KW |
1188 | } |
1189 | ||
2b5e7bc2 KW |
1190 | /* Note that there are two types of too-short malformation. One is when |
1191 | * there is actual wrong data before the normal termination of the | |
1192 | * sequence. The other is that the sequence wasn't complete before the end | |
1193 | * of the data we are allowed to look at, based on the input 'curlen'. | |
1194 | * This means that we were passed data for a partial character, but it is | |
1195 | * valid as far as we saw. The other is definitely invalid. This | |
1196 | * distinction could be important to a caller, so the two types are kept | |
1197 | * separate. */ | |
1198 | ||
1199 | /* Check for overflow */ | |
1200 | if (UNLIKELY(does_utf8_overflow(s0, send))) { | |
1201 | possible_problems |= UTF8_GOT_OVERFLOW; | |
1202 | uv = UNICODE_REPLACEMENT; | |
eb83ed87 | 1203 | } |
eb83ed87 | 1204 | |
2b5e7bc2 KW |
1205 | /* Check for overlong. If no problems so far, 'uv' is the correct code |
1206 | * point value. Simply see if it is expressible in fewer bytes. Otherwise | |
1207 | * we must look at the UTF-8 byte sequence itself to see if it is for an | |
1208 | * overlong */ | |
1209 | if ( ( LIKELY(! possible_problems) | |
1210 | && UNLIKELY(expectlen > (STRLEN) OFFUNISKIP(uv))) | |
1211 | || ( UNLIKELY( possible_problems) | |
1212 | && ( UNLIKELY(! UTF8_IS_START(*s0)) | |
1213 | || ( curlen > 1 | |
1214 | && UNLIKELY(is_utf8_overlong_given_start_byte_ok(s0, | |
1215 | send - s0)))))) | |
2f8f112e | 1216 | { |
2b5e7bc2 KW |
1217 | possible_problems |= UTF8_GOT_LONG; |
1218 | ||
2b5e7bc2 KW |
1219 | if (UNLIKELY(possible_problems & UTF8_GOT_TOO_SHORT)) { |
1220 | UV min_uv = uv_so_far; | |
1221 | STRLEN i; | |
1222 | ||
1223 | /* Here, the input is both overlong and is missing some trailing | |
1224 | * bytes. There is no single code point it could be for, but there | |
1225 | * may be enough information present to determine if what we have | |
1226 | * so far is for an unallowed code point, such as for a surrogate. | |
1227 | * The code below has the intelligence to determine this, but just | |
1228 | * for non-overlong UTF-8 sequences. What we do here is calculate | |
1229 | * the smallest code point the input could represent if there were | |
1230 | * no too short malformation. Then we compute and save the UTF-8 | |
1231 | * for that, which is what the code below looks at instead of the | |
1232 | * raw input. It turns out that the smallest such code point is | |
1233 | * all we need. */ | |
1234 | for (i = curlen; i < expectlen; i++) { | |
1235 | min_uv = UTF8_ACCUMULATE(min_uv, | |
1236 | I8_TO_NATIVE_UTF8(UTF_CONTINUATION_MARK)); | |
1237 | } | |
1238 | ||
1239 | Newx(adjusted_s0, OFFUNISKIP(min_uv) + 1, U8); | |
1240 | SAVEFREEPV((U8 *) adjusted_s0); /* Needed because we may not get | |
1241 | to free it ourselves if | |
1242 | warnings are made fatal */ | |
1243 | adjusted_send = uvoffuni_to_utf8_flags(adjusted_s0, min_uv, 0); | |
1244 | } | |
eb83ed87 KW |
1245 | } |
1246 | ||
2b5e7bc2 KW |
1247 | /* Now check that the input isn't for a problematic code point not allowed |
1248 | * by the input parameters. */ | |
1249 | /* isn't problematic if < this */ | |
1250 | if ( ( ( LIKELY(! possible_problems) && uv >= UNICODE_SURROGATE_FIRST) | |
1251 | || ( UNLIKELY(possible_problems) | |
d60baaa7 KW |
1252 | |
1253 | /* if overflow, we know without looking further | |
1254 | * precisely which of the problematic types it is, | |
1255 | * and we deal with those in the overflow handling | |
1256 | * code */ | |
1257 | && LIKELY(! (possible_problems & UTF8_GOT_OVERFLOW)) | |
2b5e7bc2 | 1258 | && isUTF8_POSSIBLY_PROBLEMATIC(*adjusted_s0))) |
760c7c2f KW |
1259 | && ((flags & ( UTF8_DISALLOW_NONCHAR |
1260 | |UTF8_DISALLOW_SURROGATE | |
1261 | |UTF8_DISALLOW_SUPER | |
1262 | |UTF8_DISALLOW_ABOVE_31_BIT | |
1263 | |UTF8_WARN_NONCHAR | |
1264 | |UTF8_WARN_SURROGATE | |
1265 | |UTF8_WARN_SUPER | |
1266 | |UTF8_WARN_ABOVE_31_BIT)) | |
2b5e7bc2 KW |
1267 | /* In case of a malformation, 'uv' is not valid, and has |
1268 | * been changed to something in the Unicode range. | |
1269 | * Currently we don't output a deprecation message if there | |
1270 | * is already a malformation, so we don't have to special | |
1271 | * case the test immediately below */ | |
760c7c2f KW |
1272 | || ( UNLIKELY(uv > MAX_NON_DEPRECATED_CP) |
1273 | && ckWARN_d(WARN_DEPRECATED)))) | |
eb83ed87 | 1274 | { |
2b5e7bc2 KW |
1275 | /* If there were no malformations, or the only malformation is an |
1276 | * overlong, 'uv' is valid */ | |
1277 | if (LIKELY(! (possible_problems & ~UTF8_GOT_LONG))) { | |
1278 | if (UNLIKELY(UNICODE_IS_SURROGATE(uv))) { | |
1279 | possible_problems |= UTF8_GOT_SURROGATE; | |
1280 | } | |
1281 | else if (UNLIKELY(uv > PERL_UNICODE_MAX)) { | |
1282 | possible_problems |= UTF8_GOT_SUPER; | |
1283 | } | |
1284 | else if (UNLIKELY(UNICODE_IS_NONCHAR(uv))) { | |
1285 | possible_problems |= UTF8_GOT_NONCHAR; | |
1286 | } | |
1287 | } | |
1288 | else { /* Otherwise, need to look at the source UTF-8, possibly | |
1289 | adjusted to be non-overlong */ | |
1290 | ||
1291 | if (UNLIKELY(NATIVE_UTF8_TO_I8(*adjusted_s0) | |
1292 | >= FIRST_START_BYTE_THAT_IS_DEFINITELY_SUPER)) | |
ea5ced44 | 1293 | { |
2b5e7bc2 KW |
1294 | possible_problems |= UTF8_GOT_SUPER; |
1295 | } | |
1296 | else if (curlen > 1) { | |
1297 | if (UNLIKELY(IS_UTF8_2_BYTE_SUPER( | |
1298 | NATIVE_UTF8_TO_I8(*adjusted_s0), | |
1299 | NATIVE_UTF8_TO_I8(*(adjusted_s0 + 1))))) | |
ea5ced44 | 1300 | { |
2b5e7bc2 | 1301 | possible_problems |= UTF8_GOT_SUPER; |
ea5ced44 | 1302 | } |
2b5e7bc2 KW |
1303 | else if (UNLIKELY(IS_UTF8_2_BYTE_SURROGATE( |
1304 | NATIVE_UTF8_TO_I8(*adjusted_s0), | |
1305 | NATIVE_UTF8_TO_I8(*(adjusted_s0 + 1))))) | |
1306 | { | |
1307 | possible_problems |= UTF8_GOT_SURROGATE; | |
ea5ced44 KW |
1308 | } |
1309 | } | |
c0236afe | 1310 | |
2b5e7bc2 KW |
1311 | /* We need a complete well-formed UTF-8 character to discern |
1312 | * non-characters, so can't look for them here */ | |
1313 | } | |
1314 | } | |
949cf498 | 1315 | |
2b5e7bc2 KW |
1316 | ready_to_handle_errors: |
1317 | ||
1318 | /* At this point: | |
1319 | * curlen contains the number of bytes in the sequence that | |
1320 | * this call should advance the input by. | |
e308b348 KW |
1321 | * avail_len gives the available number of bytes passed in, but |
1322 | * only if this is less than the expected number of | |
1323 | * bytes, based on the code point's start byte. | |
2b5e7bc2 KW |
1324 | * possible_problems' is 0 if there weren't any problems; otherwise a bit |
1325 | * is set in it for each potential problem found. | |
1326 | * uv contains the code point the input sequence | |
1327 | * represents; or if there is a problem that prevents | |
1328 | * a well-defined value from being computed, it is | |
1329 | * some subsitute value, typically the REPLACEMENT | |
1330 | * CHARACTER. | |
1331 | * s0 points to the first byte of the character | |
1332 | * send points to just after where that (potentially | |
1333 | * partial) character ends | |
1334 | * adjusted_s0 normally is the same as s0, but in case of an | |
1335 | * overlong for which the UTF-8 matters below, it is | |
1336 | * the first byte of the shortest form representation | |
1337 | * of the input. | |
1338 | * adjusted_send normally is the same as 'send', but if adjusted_s0 | |
1339 | * is set to something other than s0, this points one | |
1340 | * beyond its end | |
1341 | */ | |
eb83ed87 | 1342 | |
2b5e7bc2 KW |
1343 | if (UNLIKELY(possible_problems)) { |
1344 | bool disallowed = FALSE; | |
1345 | const U32 orig_problems = possible_problems; | |
1346 | ||
1347 | while (possible_problems) { /* Handle each possible problem */ | |
1348 | UV pack_warn = 0; | |
1349 | char * message = NULL; | |
1350 | ||
1351 | /* Each 'if' clause handles one problem. They are ordered so that | |
1352 | * the first ones' messages will be displayed before the later | |
1353 | * ones; this is kinda in decreasing severity order */ | |
1354 | if (possible_problems & UTF8_GOT_OVERFLOW) { | |
1355 | ||
1356 | /* Overflow means also got a super and above 31 bits, but we | |
1357 | * handle all three cases here */ | |
1358 | possible_problems | |
1359 | &= ~(UTF8_GOT_OVERFLOW|UTF8_GOT_SUPER|UTF8_GOT_ABOVE_31_BIT); | |
f9380377 KW |
1360 | *errors |= UTF8_GOT_OVERFLOW; |
1361 | ||
1362 | /* But the API says we flag all errors found */ | |
1363 | if (flags & (UTF8_WARN_SUPER|UTF8_DISALLOW_SUPER)) { | |
1364 | *errors |= UTF8_GOT_SUPER; | |
1365 | } | |
ddb65933 KW |
1366 | if (flags |
1367 | & (UTF8_WARN_ABOVE_31_BIT|UTF8_DISALLOW_ABOVE_31_BIT)) | |
1368 | { | |
f9380377 KW |
1369 | *errors |= UTF8_GOT_ABOVE_31_BIT; |
1370 | } | |
2b5e7bc2 | 1371 | |
d60baaa7 KW |
1372 | /* Disallow if any of the three categories say to */ |
1373 | if ( ! (flags & UTF8_ALLOW_OVERFLOW) | |
1374 | || (flags & ( UTF8_DISALLOW_SUPER | |
1375 | |UTF8_DISALLOW_ABOVE_31_BIT))) | |
1376 | { | |
1377 | disallowed = TRUE; | |
1378 | } | |
1379 | ||
1380 | ||
1381 | /* Likewise, warn if any say to, plus if deprecation warnings | |
1382 | * are on, because this code point is above IV_MAX */ | |
1383 | if ( ckWARN_d(WARN_DEPRECATED) | |
1384 | || ! (flags & UTF8_ALLOW_OVERFLOW) | |
1385 | || (flags & (UTF8_WARN_SUPER|UTF8_WARN_ABOVE_31_BIT))) | |
1386 | { | |
2b5e7bc2 | 1387 | |
ddb65933 KW |
1388 | /* The warnings code explicitly says it doesn't handle the |
1389 | * case of packWARN2 and two categories which have | |
1390 | * parent-child relationship. Even if it works now to | |
1391 | * raise the warning if either is enabled, it wouldn't | |
1392 | * necessarily do so in the future. We output (only) the | |
1393 | * most dire warning*/ | |
1394 | if (! (flags & UTF8_CHECK_ONLY)) { | |
1395 | if (ckWARN_d(WARN_UTF8)) { | |
1396 | pack_warn = packWARN(WARN_UTF8); | |
1397 | } | |
1398 | else if (ckWARN_d(WARN_NON_UNICODE)) { | |
1399 | pack_warn = packWARN(WARN_NON_UNICODE); | |
1400 | } | |
1401 | if (pack_warn) { | |
1402 | message = Perl_form(aTHX_ "%s: %s (overflows)", | |
1403 | malformed_text, | |
1404 | _byte_dump_string(s0, send - s0)); | |
1405 | } | |
2b5e7bc2 KW |
1406 | } |
1407 | } | |
1408 | } | |
1409 | else if (possible_problems & UTF8_GOT_EMPTY) { | |
1410 | possible_problems &= ~UTF8_GOT_EMPTY; | |
f9380377 | 1411 | *errors |= UTF8_GOT_EMPTY; |
2b5e7bc2 KW |
1412 | |
1413 | if (! (flags & UTF8_ALLOW_EMPTY)) { | |
d1f8d421 KW |
1414 | |
1415 | /* This so-called malformation is now treated as a bug in | |
1416 | * the caller. If you have nothing to decode, skip calling | |
1417 | * this function */ | |
1418 | assert(0); | |
1419 | ||
2b5e7bc2 KW |
1420 | disallowed = TRUE; |
1421 | if (ckWARN_d(WARN_UTF8) && ! (flags & UTF8_CHECK_ONLY)) { | |
1422 | pack_warn = packWARN(WARN_UTF8); | |
1423 | message = Perl_form(aTHX_ "%s (empty string)", | |
1424 | malformed_text); | |
1425 | } | |
1426 | } | |
1427 | } | |
1428 | else if (possible_problems & UTF8_GOT_CONTINUATION) { | |
1429 | possible_problems &= ~UTF8_GOT_CONTINUATION; | |
f9380377 | 1430 | *errors |= UTF8_GOT_CONTINUATION; |
2b5e7bc2 KW |
1431 | |
1432 | if (! (flags & UTF8_ALLOW_CONTINUATION)) { | |
1433 | disallowed = TRUE; | |
1434 | if (ckWARN_d(WARN_UTF8) && ! (flags & UTF8_CHECK_ONLY)) { | |
1435 | pack_warn = packWARN(WARN_UTF8); | |
1436 | message = Perl_form(aTHX_ | |
1437 | "%s: %s (unexpected continuation byte 0x%02x," | |
1438 | " with no preceding start byte)", | |
1439 | malformed_text, | |
1440 | _byte_dump_string(s0, 1), *s0); | |
1441 | } | |
1442 | } | |
1443 | } | |
2b5e7bc2 KW |
1444 | else if (possible_problems & UTF8_GOT_SHORT) { |
1445 | possible_problems &= ~UTF8_GOT_SHORT; | |
f9380377 | 1446 | *errors |= UTF8_GOT_SHORT; |
2b5e7bc2 KW |
1447 | |
1448 | if (! (flags & UTF8_ALLOW_SHORT)) { | |
1449 | disallowed = TRUE; | |
1450 | if (ckWARN_d(WARN_UTF8) && ! (flags & UTF8_CHECK_ONLY)) { | |
1451 | pack_warn = packWARN(WARN_UTF8); | |
1452 | message = Perl_form(aTHX_ | |
9a6c9c81 | 1453 | "%s: %s (too short; %d byte%s available, need %d)", |
2b5e7bc2 KW |
1454 | malformed_text, |
1455 | _byte_dump_string(s0, send - s0), | |
e308b348 KW |
1456 | (int)avail_len, |
1457 | avail_len == 1 ? "" : "s", | |
2b5e7bc2 KW |
1458 | (int)expectlen); |
1459 | } | |
1460 | } | |
ba210ebe | 1461 | |
2b5e7bc2 | 1462 | } |
e308b348 KW |
1463 | else if (possible_problems & UTF8_GOT_NON_CONTINUATION) { |
1464 | possible_problems &= ~UTF8_GOT_NON_CONTINUATION; | |
1465 | *errors |= UTF8_GOT_NON_CONTINUATION; | |
1466 | ||
1467 | if (! (flags & UTF8_ALLOW_NON_CONTINUATION)) { | |
1468 | disallowed = TRUE; | |
1469 | if (ckWARN_d(WARN_UTF8) && ! (flags & UTF8_CHECK_ONLY)) { | |
99a765e9 KW |
1470 | |
1471 | /* If we don't know for sure that the input length is | |
1472 | * valid, avoid as much as possible reading past the | |
1473 | * end of the buffer */ | |
1474 | int printlen = (flags & _UTF8_NO_CONFIDENCE_IN_CURLEN) | |
1475 | ? s - s0 | |
1476 | : send - s0; | |
e308b348 KW |
1477 | pack_warn = packWARN(WARN_UTF8); |
1478 | message = Perl_form(aTHX_ "%s", | |
1479 | unexpected_non_continuation_text(s0, | |
99a765e9 | 1480 | printlen, |
e308b348 KW |
1481 | s - s0, |
1482 | (int) expectlen)); | |
1483 | } | |
1484 | } | |
1485 | } | |
2b5e7bc2 KW |
1486 | else if (possible_problems & UTF8_GOT_LONG) { |
1487 | possible_problems &= ~UTF8_GOT_LONG; | |
f9380377 | 1488 | *errors |= UTF8_GOT_LONG; |
2b5e7bc2 | 1489 | |
94953955 KW |
1490 | if (flags & UTF8_ALLOW_LONG) { |
1491 | ||
1492 | /* We don't allow the actual overlong value, unless the | |
1493 | * special extra bit is also set */ | |
1494 | if (! (flags & ( UTF8_ALLOW_LONG_AND_ITS_VALUE | |
1495 | & ~UTF8_ALLOW_LONG))) | |
1496 | { | |
1497 | uv = UNICODE_REPLACEMENT; | |
1498 | } | |
1499 | } | |
1500 | else { | |
2b5e7bc2 KW |
1501 | disallowed = TRUE; |
1502 | ||
1503 | if (ckWARN_d(WARN_UTF8) && ! (flags & UTF8_CHECK_ONLY)) { | |
1504 | pack_warn = packWARN(WARN_UTF8); | |
1505 | ||
1506 | /* These error types cause 'uv' to be something that | |
1507 | * isn't what was intended, so can't use it in the | |
1508 | * message. The other error types either can't | |
1509 | * generate an overlong, or else the 'uv' is valid */ | |
1510 | if (orig_problems & | |
1511 | (UTF8_GOT_TOO_SHORT|UTF8_GOT_OVERFLOW)) | |
1512 | { | |
1513 | message = Perl_form(aTHX_ | |
1514 | "%s: %s (any UTF-8 sequence that starts" | |
1515 | " with \"%s\" is overlong which can and" | |
1516 | " should be represented with a" | |
1517 | " different, shorter sequence)", | |
1518 | malformed_text, | |
1519 | _byte_dump_string(s0, send - s0), | |
1520 | _byte_dump_string(s0, curlen)); | |
1521 | } | |
1522 | else { | |
1523 | U8 tmpbuf[UTF8_MAXBYTES+1]; | |
1524 | const U8 * const e = uvoffuni_to_utf8_flags(tmpbuf, | |
1525 | uv, 0); | |
1526 | message = Perl_form(aTHX_ | |
1527 | "%s: %s (overlong; instead use %s to represent" | |
147e3846 | 1528 | " U+%0*" UVXf ")", |
2b5e7bc2 KW |
1529 | malformed_text, |
1530 | _byte_dump_string(s0, send - s0), | |
1531 | _byte_dump_string(tmpbuf, e - tmpbuf), | |
1532 | ((uv < 256) ? 2 : 4), /* Field width of 2 for | |
1533 | small code points */ | |
1534 | uv); | |
1535 | } | |
1536 | } | |
1537 | } | |
1538 | } | |
1539 | else if (possible_problems & UTF8_GOT_SURROGATE) { | |
1540 | possible_problems &= ~UTF8_GOT_SURROGATE; | |
1541 | ||
f9380377 KW |
1542 | if (flags & UTF8_WARN_SURROGATE) { |
1543 | *errors |= UTF8_GOT_SURROGATE; | |
1544 | ||
1545 | if ( ! (flags & UTF8_CHECK_ONLY) | |
1546 | && ckWARN_d(WARN_SURROGATE)) | |
1547 | { | |
2b5e7bc2 KW |
1548 | pack_warn = packWARN(WARN_SURROGATE); |
1549 | ||
1550 | /* These are the only errors that can occur with a | |
1551 | * surrogate when the 'uv' isn't valid */ | |
1552 | if (orig_problems & UTF8_GOT_TOO_SHORT) { | |
1553 | message = Perl_form(aTHX_ | |
1554 | "UTF-16 surrogate (any UTF-8 sequence that" | |
1555 | " starts with \"%s\" is for a surrogate)", | |
1556 | _byte_dump_string(s0, curlen)); | |
1557 | } | |
1558 | else { | |
1559 | message = Perl_form(aTHX_ | |
147e3846 | 1560 | "UTF-16 surrogate U+%04" UVXf, uv); |
2b5e7bc2 | 1561 | } |
f9380377 | 1562 | } |
2b5e7bc2 | 1563 | } |
ba210ebe | 1564 | |
2b5e7bc2 KW |
1565 | if (flags & UTF8_DISALLOW_SURROGATE) { |
1566 | disallowed = TRUE; | |
f9380377 | 1567 | *errors |= UTF8_GOT_SURROGATE; |
2b5e7bc2 KW |
1568 | } |
1569 | } | |
1570 | else if (possible_problems & UTF8_GOT_SUPER) { | |
1571 | possible_problems &= ~UTF8_GOT_SUPER; | |
949cf498 | 1572 | |
f9380377 KW |
1573 | if (flags & UTF8_WARN_SUPER) { |
1574 | *errors |= UTF8_GOT_SUPER; | |
1575 | ||
1576 | if ( ! (flags & UTF8_CHECK_ONLY) | |
1577 | && ckWARN_d(WARN_NON_UNICODE)) | |
1578 | { | |
2b5e7bc2 KW |
1579 | pack_warn = packWARN(WARN_NON_UNICODE); |
1580 | ||
1581 | if (orig_problems & UTF8_GOT_TOO_SHORT) { | |
1582 | message = Perl_form(aTHX_ | |
1583 | "Any UTF-8 sequence that starts with" | |
1584 | " \"%s\" is for a non-Unicode code point," | |
1585 | " may not be portable", | |
1586 | _byte_dump_string(s0, curlen)); | |
1587 | } | |
1588 | else { | |
1589 | message = Perl_form(aTHX_ | |
147e3846 | 1590 | "Code point 0x%04" UVXf " is not" |
2b5e7bc2 KW |
1591 | " Unicode, may not be portable", |
1592 | uv); | |
1593 | } | |
f9380377 | 1594 | } |
2b5e7bc2 | 1595 | } |
ba210ebe | 1596 | |
2b5e7bc2 KW |
1597 | /* The maximum code point ever specified by a standard was |
1598 | * 2**31 - 1. Anything larger than that is a Perl extension | |
1599 | * that very well may not be understood by other applications | |
1600 | * (including earlier perl versions on EBCDIC platforms). We | |
1601 | * test for these after the regular SUPER ones, and before | |
1602 | * possibly bailing out, so that the slightly more dire warning | |
1603 | * will override the regular one. */ | |
1604 | if ( (flags & (UTF8_WARN_ABOVE_31_BIT | |
1605 | |UTF8_WARN_SUPER | |
1606 | |UTF8_DISALLOW_ABOVE_31_BIT)) | |
1607 | && ( ( UNLIKELY(orig_problems & UTF8_GOT_TOO_SHORT) | |
1608 | && UNLIKELY(is_utf8_cp_above_31_bits( | |
1609 | adjusted_s0, | |
1610 | adjusted_send))) | |
1611 | || ( LIKELY(! (orig_problems & UTF8_GOT_TOO_SHORT)) | |
1612 | && UNLIKELY(UNICODE_IS_ABOVE_31_BIT(uv))))) | |
1613 | { | |
1614 | if ( ! (flags & UTF8_CHECK_ONLY) | |
1615 | && (flags & (UTF8_WARN_ABOVE_31_BIT|UTF8_WARN_SUPER)) | |
1616 | && ckWARN_d(WARN_UTF8)) | |
1617 | { | |
1618 | pack_warn = packWARN(WARN_UTF8); | |
1619 | ||
1620 | if (orig_problems & UTF8_GOT_TOO_SHORT) { | |
1621 | message = Perl_form(aTHX_ | |
1622 | "Any UTF-8 sequence that starts with" | |
1623 | " \"%s\" is for a non-Unicode code" | |
1624 | " point, and is not portable", | |
1625 | _byte_dump_string(s0, curlen)); | |
1626 | } | |
1627 | else { | |
1628 | message = Perl_form(aTHX_ | |
147e3846 | 1629 | "Code point 0x%" UVXf " is not Unicode," |
2b5e7bc2 KW |
1630 | " and not portable", |
1631 | uv); | |
1632 | } | |
1633 | } | |
1634 | ||
ddb65933 KW |
1635 | if (flags & ( UTF8_WARN_ABOVE_31_BIT |
1636 | |UTF8_DISALLOW_ABOVE_31_BIT)) | |
1637 | { | |
f9380377 KW |
1638 | *errors |= UTF8_GOT_ABOVE_31_BIT; |
1639 | ||
1640 | if (flags & UTF8_DISALLOW_ABOVE_31_BIT) { | |
1641 | disallowed = TRUE; | |
1642 | } | |
2b5e7bc2 KW |
1643 | } |
1644 | } | |
eb83ed87 | 1645 | |
2b5e7bc2 | 1646 | if (flags & UTF8_DISALLOW_SUPER) { |
f9380377 | 1647 | *errors |= UTF8_GOT_SUPER; |
2b5e7bc2 KW |
1648 | disallowed = TRUE; |
1649 | } | |
eb83ed87 | 1650 | |
2b5e7bc2 KW |
1651 | /* The deprecated warning overrides any non-deprecated one. If |
1652 | * there are other problems, a deprecation message is not | |
1653 | * really helpful, so don't bother to raise it in that case. | |
1654 | * This also keeps the code from having to handle the case | |
1655 | * where 'uv' is not valid. */ | |
1656 | if ( ! (orig_problems | |
1657 | & (UTF8_GOT_TOO_SHORT|UTF8_GOT_OVERFLOW)) | |
1658 | && UNLIKELY(uv > MAX_NON_DEPRECATED_CP) | |
1659 | && ckWARN_d(WARN_DEPRECATED)) | |
1660 | { | |
1661 | message = Perl_form(aTHX_ cp_above_legal_max, | |
1662 | uv, MAX_NON_DEPRECATED_CP); | |
1663 | pack_warn = packWARN(WARN_DEPRECATED); | |
1664 | } | |
1665 | } | |
1666 | else if (possible_problems & UTF8_GOT_NONCHAR) { | |
1667 | possible_problems &= ~UTF8_GOT_NONCHAR; | |
ba210ebe | 1668 | |
f9380377 KW |
1669 | if (flags & UTF8_WARN_NONCHAR) { |
1670 | *errors |= UTF8_GOT_NONCHAR; | |
1671 | ||
1672 | if ( ! (flags & UTF8_CHECK_ONLY) | |
1673 | && ckWARN_d(WARN_NONCHAR)) | |
1674 | { | |
2b5e7bc2 KW |
1675 | /* The code above should have guaranteed that we don't |
1676 | * get here with errors other than overlong */ | |
1677 | assert (! (orig_problems | |
1678 | & ~(UTF8_GOT_LONG|UTF8_GOT_NONCHAR))); | |
1679 | ||
1680 | pack_warn = packWARN(WARN_NONCHAR); | |
1681 | message = Perl_form(aTHX_ "Unicode non-character" | |
147e3846 | 1682 | " U+%04" UVXf " is not recommended" |
2b5e7bc2 | 1683 | " for open interchange", uv); |
f9380377 | 1684 | } |
2b5e7bc2 | 1685 | } |
5b311467 | 1686 | |
2b5e7bc2 KW |
1687 | if (flags & UTF8_DISALLOW_NONCHAR) { |
1688 | disallowed = TRUE; | |
f9380377 | 1689 | *errors |= UTF8_GOT_NONCHAR; |
2b5e7bc2 KW |
1690 | } |
1691 | } /* End of looking through the possible flags */ | |
1692 | ||
1693 | /* Display the message (if any) for the problem being handled in | |
1694 | * this iteration of the loop */ | |
1695 | if (message) { | |
1696 | if (PL_op) | |
1697 | Perl_warner(aTHX_ pack_warn, "%s in %s", message, | |
1698 | OP_DESC(PL_op)); | |
1699 | else | |
1700 | Perl_warner(aTHX_ pack_warn, "%s", message); | |
1701 | } | |
ddb65933 | 1702 | } /* End of 'while (possible_problems)' */ |
a0dbb045 | 1703 | |
2b5e7bc2 KW |
1704 | /* Since there was a possible problem, the returned length may need to |
1705 | * be changed from the one stored at the beginning of this function. | |
1706 | * Instead of trying to figure out if that's needed, just do it. */ | |
1707 | if (retlen) { | |
1708 | *retlen = curlen; | |
1709 | } | |
a0dbb045 | 1710 | |
2b5e7bc2 KW |
1711 | if (disallowed) { |
1712 | if (flags & UTF8_CHECK_ONLY && retlen) { | |
1713 | *retlen = ((STRLEN) -1); | |
1714 | } | |
1715 | return 0; | |
1716 | } | |
eb83ed87 | 1717 | } |
ba210ebe | 1718 | |
2b5e7bc2 | 1719 | return UNI_TO_NATIVE(uv); |
a0ed51b3 LW |
1720 | } |
1721 | ||
8e84507e | 1722 | /* |
ec5f19d0 KW |
1723 | =for apidoc utf8_to_uvchr_buf |
1724 | ||
1725 | Returns the native code point of the first character in the string C<s> which | |
1726 | is assumed to be in UTF-8 encoding; C<send> points to 1 beyond the end of C<s>. | |
524080c4 | 1727 | C<*retlen> will be set to the length, in bytes, of that character. |
ec5f19d0 | 1728 | |
524080c4 KW |
1729 | If C<s> does not point to a well-formed UTF-8 character and UTF8 warnings are |
1730 | enabled, zero is returned and C<*retlen> is set (if C<retlen> isn't | |
796b6530 | 1731 | C<NULL>) to -1. If those warnings are off, the computed value, if well-defined |
173db420 | 1732 | (or the Unicode REPLACEMENT CHARACTER if not), is silently returned, and |
796b6530 | 1733 | C<*retlen> is set (if C<retlen> isn't C<NULL>) so that (S<C<s> + C<*retlen>>) is |
173db420 | 1734 | the next possible position in C<s> that could begin a non-malformed character. |
de69f3af | 1735 | See L</utf8n_to_uvchr> for details on when the REPLACEMENT CHARACTER is |
173db420 | 1736 | returned. |
ec5f19d0 | 1737 | |
760c7c2f KW |
1738 | Code points above the platform's C<IV_MAX> will raise a deprecation warning, |
1739 | unless those are turned off. | |
1740 | ||
ec5f19d0 | 1741 | =cut |
52be2536 KW |
1742 | |
1743 | Also implemented as a macro in utf8.h | |
1744 | ||
ec5f19d0 KW |
1745 | */ |
1746 | ||
1747 | ||
1748 | UV | |
1749 | Perl_utf8_to_uvchr_buf(pTHX_ const U8 *s, const U8 *send, STRLEN *retlen) | |
1750 | { | |
ec5f19d0 KW |
1751 | assert(s < send); |
1752 | ||
1753 | return utf8n_to_uvchr(s, send - s, retlen, | |
ddb65933 | 1754 | ckWARN_d(WARN_UTF8) ? 0 : UTF8_ALLOW_ANY); |
ec5f19d0 KW |
1755 | } |
1756 | ||
52be2536 KW |
1757 | /* This is marked as deprecated |
1758 | * | |
ec5f19d0 KW |
1759 | =for apidoc utf8_to_uvuni_buf |
1760 | ||
de69f3af KW |
1761 | Only in very rare circumstances should code need to be dealing in Unicode |
1762 | (as opposed to native) code points. In those few cases, use | |
1763 | C<L<NATIVE_TO_UNI(utf8_to_uvchr_buf(...))|/utf8_to_uvchr_buf>> instead. | |
4f83cdcd KW |
1764 | |
1765 | Returns the Unicode (not-native) code point of the first character in the | |
1766 | string C<s> which | |
ec5f19d0 KW |
1767 | is assumed to be in UTF-8 encoding; C<send> points to 1 beyond the end of C<s>. |
1768 | C<retlen> will be set to the length, in bytes, of that character. | |
1769 | ||
524080c4 KW |
1770 | If C<s> does not point to a well-formed UTF-8 character and UTF8 warnings are |
1771 | enabled, zero is returned and C<*retlen> is set (if C<retlen> isn't | |
1772 | NULL) to -1. If those warnings are off, the computed value if well-defined (or | |
1773 | the Unicode REPLACEMENT CHARACTER, if not) is silently returned, and C<*retlen> | |
1774 | is set (if C<retlen> isn't NULL) so that (S<C<s> + C<*retlen>>) is the | |
1775 | next possible position in C<s> that could begin a non-malformed character. | |
de69f3af | 1776 | See L</utf8n_to_uvchr> for details on when the REPLACEMENT CHARACTER is returned. |
ec5f19d0 | 1777 | |
760c7c2f KW |
1778 | Code points above the platform's C<IV_MAX> will raise a deprecation warning, |
1779 | unless those are turned off. | |
1780 | ||
ec5f19d0 KW |
1781 | =cut |
1782 | */ | |
1783 | ||
1784 | UV | |
1785 | Perl_utf8_to_uvuni_buf(pTHX_ const U8 *s, const U8 *send, STRLEN *retlen) | |
1786 | { | |
1787 | PERL_ARGS_ASSERT_UTF8_TO_UVUNI_BUF; | |
1788 | ||
1789 | assert(send > s); | |
1790 | ||
5962d97e KW |
1791 | /* Call the low level routine, asking for checks */ |
1792 | return NATIVE_TO_UNI(utf8_to_uvchr_buf(s, send, retlen)); | |
ec5f19d0 KW |
1793 | } |
1794 | ||
b76347f2 | 1795 | /* |
87cea99e | 1796 | =for apidoc utf8_length |
b76347f2 JH |
1797 | |
1798 | Return the length of the UTF-8 char encoded string C<s> in characters. | |
02eb7b47 JH |
1799 | Stops at C<e> (inclusive). If C<e E<lt> s> or if the scan would end |
1800 | up past C<e>, croaks. | |
b76347f2 JH |
1801 | |
1802 | =cut | |
1803 | */ | |
1804 | ||
1805 | STRLEN | |
35a4481c | 1806 | Perl_utf8_length(pTHX_ const U8 *s, const U8 *e) |
b76347f2 JH |
1807 | { |
1808 | STRLEN len = 0; | |
1809 | ||
7918f24d NC |
1810 | PERL_ARGS_ASSERT_UTF8_LENGTH; |
1811 | ||
8850bf83 JH |
1812 | /* Note: cannot use UTF8_IS_...() too eagerly here since e.g. |
1813 | * the bitops (especially ~) can create illegal UTF-8. | |
1814 | * In other words: in Perl UTF-8 is not just for Unicode. */ | |
1815 | ||
a3b680e6 AL |
1816 | if (e < s) |
1817 | goto warn_and_return; | |
b76347f2 | 1818 | while (s < e) { |
4cbf4130 | 1819 | s += UTF8SKIP(s); |
8e91ec7f AV |
1820 | len++; |
1821 | } | |
1822 | ||
1823 | if (e != s) { | |
1824 | len--; | |
1825 | warn_and_return: | |
9b387841 NC |
1826 | if (PL_op) |
1827 | Perl_ck_warner_d(aTHX_ packWARN(WARN_UTF8), | |
1828 | "%s in %s", unees, OP_DESC(PL_op)); | |
1829 | else | |
61a12c31 | 1830 | Perl_ck_warner_d(aTHX_ packWARN(WARN_UTF8), "%s", unees); |
b76347f2 JH |
1831 | } |
1832 | ||
1833 | return len; | |
1834 | } | |
1835 | ||
b06226ff | 1836 | /* |
fed3ba5d NC |
1837 | =for apidoc bytes_cmp_utf8 |
1838 | ||
a1433954 | 1839 | Compares the sequence of characters (stored as octets) in C<b>, C<blen> with the |
72d33970 FC |
1840 | sequence of characters (stored as UTF-8) |
1841 | in C<u>, C<ulen>. Returns 0 if they are | |
fed3ba5d NC |
1842 | equal, -1 or -2 if the first string is less than the second string, +1 or +2 |
1843 | if the first string is greater than the second string. | |
1844 | ||
1845 | -1 or +1 is returned if the shorter string was identical to the start of the | |
72d33970 FC |
1846 | longer string. -2 or +2 is returned if |
1847 | there was a difference between characters | |
fed3ba5d NC |
1848 | within the strings. |
1849 | ||
1850 | =cut | |
1851 | */ | |
1852 | ||
1853 | int | |
1854 | Perl_bytes_cmp_utf8(pTHX_ const U8 *b, STRLEN blen, const U8 *u, STRLEN ulen) | |
1855 | { | |
1856 | const U8 *const bend = b + blen; | |
1857 | const U8 *const uend = u + ulen; | |
1858 | ||
1859 | PERL_ARGS_ASSERT_BYTES_CMP_UTF8; | |
fed3ba5d NC |
1860 | |
1861 | while (b < bend && u < uend) { | |
1862 | U8 c = *u++; | |
1863 | if (!UTF8_IS_INVARIANT(c)) { | |
1864 | if (UTF8_IS_DOWNGRADEABLE_START(c)) { | |
1865 | if (u < uend) { | |
1866 | U8 c1 = *u++; | |
1867 | if (UTF8_IS_CONTINUATION(c1)) { | |
a62b247b | 1868 | c = EIGHT_BIT_UTF8_TO_NATIVE(c, c1); |
fed3ba5d | 1869 | } else { |
2b5e7bc2 | 1870 | /* diag_listed_as: Malformed UTF-8 character%s */ |
fed3ba5d | 1871 | Perl_ck_warner_d(aTHX_ packWARN(WARN_UTF8), |
806547a7 | 1872 | "%s %s%s", |
7cf8d05d | 1873 | unexpected_non_continuation_text(u - 1, 2, 1, 2), |
806547a7 KW |
1874 | PL_op ? " in " : "", |
1875 | PL_op ? OP_DESC(PL_op) : ""); | |
fed3ba5d NC |
1876 | return -2; |
1877 | } | |
1878 | } else { | |
1879 | if (PL_op) | |
1880 | Perl_ck_warner_d(aTHX_ packWARN(WARN_UTF8), | |
1881 | "%s in %s", unees, OP_DESC(PL_op)); | |
1882 | else | |
61a12c31 | 1883 | Perl_ck_warner_d(aTHX_ packWARN(WARN_UTF8), "%s", unees); |
fed3ba5d NC |
1884 | return -2; /* Really want to return undef :-) */ |
1885 | } | |
1886 | } else { | |
1887 | return -2; | |
1888 | } | |
1889 | } | |
1890 | if (*b != c) { | |
1891 | return *b < c ? -2 : +2; | |
1892 | } | |
1893 | ++b; | |
1894 | } | |
1895 | ||
1896 | if (b == bend && u == uend) | |
1897 | return 0; | |
1898 | ||
1899 | return b < bend ? +1 : -1; | |
1900 | } | |
1901 | ||
1902 | /* | |
87cea99e | 1903 | =for apidoc utf8_to_bytes |
6940069f | 1904 | |
2bbc8d55 | 1905 | Converts a string C<s> of length C<len> from UTF-8 into native byte encoding. |
a1433954 KW |
1906 | Unlike L</bytes_to_utf8>, this over-writes the original string, and |
1907 | updates C<len> to contain the new length. | |
67e989fb | 1908 | Returns zero on failure, setting C<len> to -1. |
6940069f | 1909 | |
a1433954 | 1910 | If you need a copy of the string, see L</bytes_from_utf8>. |
95be277c | 1911 | |
6940069f GS |
1912 | =cut |
1913 | */ | |
1914 | ||
1915 | U8 * | |
37607a96 | 1916 | Perl_utf8_to_bytes(pTHX_ U8 *s, STRLEN *len) |
6940069f | 1917 | { |
d4c19fe8 AL |
1918 | U8 * const save = s; |
1919 | U8 * const send = s + *len; | |
6940069f | 1920 | U8 *d; |
246fae53 | 1921 | |
7918f24d | 1922 | PERL_ARGS_ASSERT_UTF8_TO_BYTES; |
81611534 | 1923 | PERL_UNUSED_CONTEXT; |
7918f24d | 1924 | |
1e54db1a | 1925 | /* ensure valid UTF-8 and chars < 256 before updating string */ |
d4c19fe8 | 1926 | while (s < send) { |
d59937ca KW |
1927 | if (! UTF8_IS_INVARIANT(*s)) { |
1928 | if (! UTF8_IS_NEXT_CHAR_DOWNGRADEABLE(s, send)) { | |
1929 | *len = ((STRLEN) -1); | |
1930 | return 0; | |
1931 | } | |
1932 | s++; | |
dcad2880 | 1933 | } |
d59937ca | 1934 | s++; |
246fae53 | 1935 | } |
dcad2880 JH |
1936 | |
1937 | d = s = save; | |
6940069f | 1938 | while (s < send) { |
80e0b38f KW |
1939 | U8 c = *s++; |
1940 | if (! UTF8_IS_INVARIANT(c)) { | |
1941 | /* Then it is two-byte encoded */ | |
a62b247b | 1942 | c = EIGHT_BIT_UTF8_TO_NATIVE(c, *s); |
80e0b38f KW |
1943 | s++; |
1944 | } | |
1945 | *d++ = c; | |
6940069f GS |
1946 | } |
1947 | *d = '\0'; | |
246fae53 | 1948 | *len = d - save; |
6940069f GS |
1949 | return save; |
1950 | } | |
1951 | ||
1952 | /* | |
87cea99e | 1953 | =for apidoc bytes_from_utf8 |
f9a63242 | 1954 | |
2bbc8d55 | 1955 | Converts a string C<s> of length C<len> from UTF-8 into native byte encoding. |
a1433954 | 1956 | Unlike L</utf8_to_bytes> but like L</bytes_to_utf8>, returns a pointer to |
ef9edfd0 JH |
1957 | the newly-created string, and updates C<len> to contain the new |
1958 | length. Returns the original string if no conversion occurs, C<len> | |
72d33970 | 1959 | is unchanged. Do nothing if C<is_utf8> points to 0. Sets C<is_utf8> to |
2bbc8d55 | 1960 | 0 if C<s> is converted or consisted entirely of characters that are invariant |
4a4088c4 | 1961 | in UTF-8 (i.e., US-ASCII on non-EBCDIC machines). |
f9a63242 | 1962 | |
37607a96 PK |
1963 | =cut |
1964 | */ | |
f9a63242 JH |
1965 | |
1966 | U8 * | |
e1ec3a88 | 1967 | Perl_bytes_from_utf8(pTHX_ const U8 *s, STRLEN *len, bool *is_utf8) |
f9a63242 | 1968 | { |
f9a63242 | 1969 | U8 *d; |
e1ec3a88 AL |
1970 | const U8 *start = s; |
1971 | const U8 *send; | |
f9a63242 JH |
1972 | I32 count = 0; |
1973 | ||
7918f24d | 1974 | PERL_ARGS_ASSERT_BYTES_FROM_UTF8; |
96a5add6 | 1975 | PERL_UNUSED_CONTEXT; |
f9a63242 | 1976 | if (!*is_utf8) |
73d840c0 | 1977 | return (U8 *)start; |
f9a63242 | 1978 | |
1e54db1a | 1979 | /* ensure valid UTF-8 and chars < 256 before converting string */ |
f9a63242 | 1980 | for (send = s + *len; s < send;) { |
d59937ca KW |
1981 | if (! UTF8_IS_INVARIANT(*s)) { |
1982 | if (! UTF8_IS_NEXT_CHAR_DOWNGRADEABLE(s, send)) { | |
73d840c0 | 1983 | return (U8 *)start; |
d59937ca KW |
1984 | } |
1985 | count++; | |
1986 | s++; | |
db42d148 | 1987 | } |
d59937ca | 1988 | s++; |
f9a63242 JH |
1989 | } |
1990 | ||
35da51f7 | 1991 | *is_utf8 = FALSE; |
f9a63242 | 1992 | |
212542aa | 1993 | Newx(d, (*len) - count + 1, U8); |
ef9edfd0 | 1994 | s = start; start = d; |
f9a63242 JH |
1995 | while (s < send) { |
1996 | U8 c = *s++; | |
1a91c45d | 1997 | if (! UTF8_IS_INVARIANT(c)) { |
c4d5f83a | 1998 | /* Then it is two-byte encoded */ |
a62b247b | 1999 | c = EIGHT_BIT_UTF8_TO_NATIVE(c, *s); |
1a91c45d | 2000 | s++; |
c4d5f83a NIS |
2001 | } |
2002 | *d++ = c; | |
f9a63242 JH |
2003 | } |
2004 | *d = '\0'; | |
2005 | *len = d - start; | |
73d840c0 | 2006 | return (U8 *)start; |
f9a63242 JH |
2007 | } |
2008 | ||
2009 | /* | |
87cea99e | 2010 | =for apidoc bytes_to_utf8 |
6940069f | 2011 | |
ff97e5cf KW |
2012 | Converts a string C<s> of length C<len> bytes from the native encoding into |
2013 | UTF-8. | |
6662521e | 2014 | Returns a pointer to the newly-created string, and sets C<len> to |
ff97e5cf | 2015 | reflect the new length in bytes. |
6940069f | 2016 | |
75200dff | 2017 | A C<NUL> character will be written after the end of the string. |
2bbc8d55 SP |
2018 | |
2019 | If you want to convert to UTF-8 from encodings other than | |
2020 | the native (Latin1 or EBCDIC), | |
a1433954 | 2021 | see L</sv_recode_to_utf8>(). |
c9ada85f | 2022 | |
497711e7 | 2023 | =cut |
6940069f GS |
2024 | */ |
2025 | ||
c682ebef FC |
2026 | /* This logic is duplicated in sv_catpvn_flags, so any bug fixes will |
2027 | likewise need duplication. */ | |
2028 | ||
6940069f | 2029 | U8* |
35a4481c | 2030 | Perl_bytes_to_utf8(pTHX_ const U8 *s, STRLEN *len) |
6940069f | 2031 | { |
35a4481c | 2032 | const U8 * const send = s + (*len); |
6940069f GS |
2033 | U8 *d; |
2034 | U8 *dst; | |
7918f24d NC |
2035 | |
2036 | PERL_ARGS_ASSERT_BYTES_TO_UTF8; | |
96a5add6 | 2037 | PERL_UNUSED_CONTEXT; |
6940069f | 2038 | |
212542aa | 2039 | Newx(d, (*len) * 2 + 1, U8); |
6940069f GS |
2040 | dst = d; |
2041 | ||
2042 | while (s < send) { | |
55d09dc8 KW |
2043 | append_utf8_from_native_byte(*s, &d); |
2044 | s++; | |
6940069f GS |
2045 | } |
2046 | *d = '\0'; | |
6662521e | 2047 | *len = d-dst; |
6940069f GS |
2048 | return dst; |
2049 | } | |
2050 | ||
a0ed51b3 | 2051 | /* |
dea0fc0b | 2052 | * Convert native (big-endian) or reversed (little-endian) UTF-16 to UTF-8. |
a0ed51b3 LW |
2053 | * |
2054 | * Destination must be pre-extended to 3/2 source. Do not use in-place. | |
2055 | * We optimize for native, for obvious reasons. */ | |
2056 | ||
2057 | U8* | |
dea0fc0b | 2058 | Perl_utf16_to_utf8(pTHX_ U8* p, U8* d, I32 bytelen, I32 *newlen) |
a0ed51b3 | 2059 | { |
dea0fc0b JH |
2060 | U8* pend; |
2061 | U8* dstart = d; | |
2062 | ||
7918f24d NC |
2063 | PERL_ARGS_ASSERT_UTF16_TO_UTF8; |
2064 | ||
dea0fc0b | 2065 | if (bytelen & 1) |
147e3846 | 2066 | Perl_croak(aTHX_ "panic: utf16_to_utf8: odd bytelen %" UVuf, (UV)bytelen); |
dea0fc0b JH |
2067 | |
2068 | pend = p + bytelen; | |
2069 | ||
a0ed51b3 | 2070 | while (p < pend) { |
dea0fc0b JH |
2071 | UV uv = (p[0] << 8) + p[1]; /* UTF-16BE */ |
2072 | p += 2; | |
2d1545e5 | 2073 | if (OFFUNI_IS_INVARIANT(uv)) { |
56d37426 | 2074 | *d++ = LATIN1_TO_NATIVE((U8) uv); |
a0ed51b3 LW |
2075 | continue; |
2076 | } | |
56d37426 KW |
2077 | if (uv <= MAX_UTF8_TWO_BYTE) { |
2078 | *d++ = UTF8_TWO_BYTE_HI(UNI_TO_NATIVE(uv)); | |
2079 | *d++ = UTF8_TWO_BYTE_LO(UNI_TO_NATIVE(uv)); | |
a0ed51b3 LW |
2080 | continue; |
2081 | } | |
46956fad KW |
2082 | #define FIRST_HIGH_SURROGATE UNICODE_SURROGATE_FIRST |
2083 | #define LAST_HIGH_SURROGATE 0xDBFF | |
2084 | #define FIRST_LOW_SURROGATE 0xDC00 | |
2085 | #define LAST_LOW_SURROGATE UNICODE_SURROGATE_LAST | |
e23c50db KW |
2086 | |
2087 | /* This assumes that most uses will be in the first Unicode plane, not | |
2088 | * needing surrogates */ | |
2089 | if (UNLIKELY(uv >= UNICODE_SURROGATE_FIRST | |
2090 | && uv <= UNICODE_SURROGATE_LAST)) | |
2091 | { | |
2092 | if (UNLIKELY(p >= pend) || UNLIKELY(uv > LAST_HIGH_SURROGATE)) { | |
2093 | Perl_croak(aTHX_ "Malformed UTF-16 surrogate"); | |
2094 | } | |
2095 | else { | |
01ea242b | 2096 | UV low = (p[0] << 8) + p[1]; |
e23c50db KW |
2097 | if ( UNLIKELY(low < FIRST_LOW_SURROGATE) |
2098 | || UNLIKELY(low > LAST_LOW_SURROGATE)) | |
2099 | { | |
01ea242b | 2100 | Perl_croak(aTHX_ "Malformed UTF-16 surrogate"); |
e23c50db KW |
2101 | } |
2102 | p += 2; | |
46956fad KW |
2103 | uv = ((uv - FIRST_HIGH_SURROGATE) << 10) |
2104 | + (low - FIRST_LOW_SURROGATE) + 0x10000; | |
01ea242b | 2105 | } |
a0ed51b3 | 2106 | } |
56d37426 KW |
2107 | #ifdef EBCDIC |
2108 | d = uvoffuni_to_utf8_flags(d, uv, 0); | |
2109 | #else | |
a0ed51b3 | 2110 | if (uv < 0x10000) { |
eb160463 GS |
2111 | *d++ = (U8)(( uv >> 12) | 0xe0); |
2112 | *d++ = (U8)(((uv >> 6) & 0x3f) | 0x80); | |
2113 | *d++ = (U8)(( uv & 0x3f) | 0x80); | |
a0ed51b3 LW |
2114 | continue; |
2115 | } | |
2116 | else { | |
eb160463 GS |
2117 | *d++ = (U8)(( uv >> 18) | 0xf0); |
2118 | *d++ = (U8)(((uv >> 12) & 0x3f) | 0x80); | |
2119 | *d++ = (U8)(((uv >> 6) & 0x3f) | 0x80); | |
2120 | *d++ = (U8)(( uv & 0x3f) | 0x80); | |
a0ed51b3 LW |
2121 | continue; |
2122 | } | |
56d37426 | 2123 | #endif |
a0ed51b3 | 2124 | } |
dea0fc0b | 2125 | *newlen = d - dstart; |
a0ed51b3 LW |
2126 | return d; |
2127 | } | |
2128 | ||
2129 | /* Note: this one is slightly destructive of the source. */ | |
2130 | ||
2131 | U8* | |
dea0fc0b | 2132 | Perl_utf16_to_utf8_reversed(pTHX_ U8* p, U8* d, I32 bytelen, I32 *newlen) |
a0ed51b3 LW |
2133 | { |
2134 | U8* s = (U8*)p; | |
d4c19fe8 | 2135 | U8* const send = s + bytelen; |
7918f24d NC |
2136 | |
2137 | PERL_ARGS_ASSERT_UTF16_TO_UTF8_REVERSED; | |
2138 | ||
e0ea5e2d | 2139 | if (bytelen & 1) |
147e3846 | 2140 | Perl_croak(aTHX_ "panic: utf16_to_utf8_reversed: odd bytelen %" UVuf, |
e0ea5e2d NC |
2141 | (UV)bytelen); |
2142 | ||
a0ed51b3 | 2143 | while (s < send) { |
d4c19fe8 | 2144 | const U8 tmp = s[0]; |
a0ed51b3 LW |
2145 | s[0] = s[1]; |
2146 | s[1] = tmp; | |
2147 | s += 2; | |
2148 | } | |
dea0fc0b | 2149 | return utf16_to_utf8(p, d, bytelen, newlen); |
a0ed51b3 LW |
2150 | } |
2151 | ||
922e8cb4 KW |
2152 | bool |
2153 | Perl__is_uni_FOO(pTHX_ const U8 classnum, const UV c) | |
2154 | { | |
2155 | U8 tmpbuf[UTF8_MAXBYTES+1]; | |
2156 | uvchr_to_utf8(tmpbuf, c); | |
da8c1a98 | 2157 | return _is_utf8_FOO_with_len(classnum, tmpbuf, tmpbuf + sizeof(tmpbuf)); |
922e8cb4 KW |
2158 | } |
2159 | ||
f9ae8fb6 JD |
2160 | /* Internal function so we can deprecate the external one, and call |
2161 | this one from other deprecated functions in this file */ | |
2162 | ||
f2645549 KW |
2163 | bool |
2164 | Perl__is_utf8_idstart(pTHX_ const U8 *p) | |
61b19385 | 2165 | { |
f2645549 | 2166 | PERL_ARGS_ASSERT__IS_UTF8_IDSTART; |
61b19385 KW |
2167 | |
2168 | if (*p == '_') | |
2169 | return TRUE; | |
f25ce844 | 2170 | return is_utf8_common(p, &PL_utf8_idstart, "IdStart", NULL); |
61b19385 KW |
2171 | } |
2172 | ||
5092f92a | 2173 | bool |
eba68aa0 KW |
2174 | Perl__is_uni_perl_idcont(pTHX_ UV c) |
2175 | { | |
2176 | U8 tmpbuf[UTF8_MAXBYTES+1]; | |
2177 | uvchr_to_utf8(tmpbuf, c); | |
da8c1a98 | 2178 | return _is_utf8_perl_idcont_with_len(tmpbuf, tmpbuf + sizeof(tmpbuf)); |
eba68aa0 KW |
2179 | } |
2180 | ||
2181 | bool | |
f91dcd13 KW |
2182 | Perl__is_uni_perl_idstart(pTHX_ UV c) |
2183 | { | |
2184 | U8 tmpbuf[UTF8_MAXBYTES+1]; | |
2185 | uvchr_to_utf8(tmpbuf, c); | |
da8c1a98 | 2186 | return _is_utf8_perl_idstart_with_len(tmpbuf, tmpbuf + sizeof(tmpbuf)); |
f91dcd13 KW |
2187 | } |
2188 | ||
3a4c58c9 KW |
2189 | UV |
2190 | Perl__to_upper_title_latin1(pTHX_ const U8 c, U8* p, STRLEN *lenp, const char S_or_s) | |
2191 | { | |
2192 | /* We have the latin1-range values compiled into the core, so just use | |
4a4088c4 | 2193 | * those, converting the result to UTF-8. The only difference between upper |
3a4c58c9 KW |
2194 | * and title case in this range is that LATIN_SMALL_LETTER_SHARP_S is |
2195 | * either "SS" or "Ss". Which one to use is passed into the routine in | |
2196 | * 'S_or_s' to avoid a test */ | |
2197 | ||
2198 | UV converted = toUPPER_LATIN1_MOD(c); | |
2199 | ||
2200 | PERL_ARGS_ASSERT__TO_UPPER_TITLE_LATIN1; | |
2201 | ||
2202 | assert(S_or_s == 'S' || S_or_s == 's'); | |
2203 | ||
6f2d5cbc | 2204 | if (UVCHR_IS_INVARIANT(converted)) { /* No difference between the two for |
f4cd282c | 2205 | characters in this range */ |
3a4c58c9 KW |
2206 | *p = (U8) converted; |
2207 | *lenp = 1; | |
2208 | return converted; | |
2209 | } | |
2210 | ||
2211 | /* toUPPER_LATIN1_MOD gives the correct results except for three outliers, | |
2212 | * which it maps to one of them, so as to only have to have one check for | |
2213 | * it in the main case */ | |
2214 | if (UNLIKELY(converted == LATIN_SMALL_LETTER_Y_WITH_DIAERESIS)) { | |
2215 | switch (c) { | |
2216 | case LATIN_SMALL_LETTER_Y_WITH_DIAERESIS: | |
2217 | converted = LATIN_CAPITAL_LETTER_Y_WITH_DIAERESIS; | |
2218 | break; | |
2219 | case MICRO_SIGN: | |
2220 | converted = GREEK_CAPITAL_LETTER_MU; | |
2221 | break; | |
79e064b9 KW |
2222 | #if UNICODE_MAJOR_VERSION > 2 \ |
2223 | || (UNICODE_MAJOR_VERSION == 2 && UNICODE_DOT_VERSION >= 1 \ | |
2224 | && UNICODE_DOT_DOT_VERSION >= 8) | |
3a4c58c9 KW |
2225 | case LATIN_SMALL_LETTER_SHARP_S: |
2226 | *(p)++ = 'S'; | |
2227 | *p = S_or_s; | |
2228 | *lenp = 2; | |
2229 | return 'S'; | |
79e064b9 | 2230 | #endif |
3a4c58c9 KW |
2231 | default: |
2232 | Perl_croak(aTHX_ "panic: to_upper_title_latin1 did not expect '%c' to map to '%c'", c, LATIN_SMALL_LETTER_Y_WITH_DIAERESIS); | |
e5964223 | 2233 | NOT_REACHED; /* NOTREACHED */ |
3a4c58c9 KW |
2234 | } |
2235 | } | |
2236 | ||
2237 | *(p)++ = UTF8_TWO_BYTE_HI(converted); | |
2238 | *p = UTF8_TWO_BYTE_LO(converted); | |
2239 | *lenp = 2; | |
2240 | ||
2241 | return converted; | |
2242 | } | |
2243 | ||
50bda2c3 KW |
2244 | /* Call the function to convert a UTF-8 encoded character to the specified case. |
2245 | * Note that there may be more than one character in the result. | |
2246 | * INP is a pointer to the first byte of the input character | |
2247 | * OUTP will be set to the first byte of the string of changed characters. It | |
2248 | * needs to have space for UTF8_MAXBYTES_CASE+1 bytes | |
2249 | * LENP will be set to the length in bytes of the string of changed characters | |
2250 | * | |
2251 | * The functions return the ordinal of the first character in the string of OUTP */ | |
b9992569 KW |
2252 | #define CALL_UPPER_CASE(uv, s, d, lenp) _to_utf8_case(uv, s, d, lenp, &PL_utf8_toupper, "ToUc", "") |
2253 | #define CALL_TITLE_CASE(uv, s, d, lenp) _to_utf8_case(uv, s, d, lenp, &PL_utf8_totitle, "ToTc", "") | |
2254 | #define CALL_LOWER_CASE(uv, s, d, lenp) _to_utf8_case(uv, s, d, lenp, &PL_utf8_tolower, "ToLc", "") | |
50bda2c3 | 2255 | |
b9992569 KW |
2256 | /* This additionally has the input parameter 'specials', which if non-zero will |
2257 | * cause this to use the specials hash for folding (meaning get full case | |
50bda2c3 | 2258 | * folding); otherwise, when zero, this implies a simple case fold */ |
b9992569 | 2259 | #define CALL_FOLD_CASE(uv, s, d, lenp, specials) _to_utf8_case(uv, s, d, lenp, &PL_utf8_tofold, "ToCf", (specials) ? "" : NULL) |
c3fd2246 | 2260 | |
84afefe6 JH |
2261 | UV |
2262 | Perl_to_uni_upper(pTHX_ UV c, U8* p, STRLEN *lenp) | |
a0ed51b3 | 2263 | { |
a1433954 KW |
2264 | /* Convert the Unicode character whose ordinal is <c> to its uppercase |
2265 | * version and store that in UTF-8 in <p> and its length in bytes in <lenp>. | |
2266 | * Note that the <p> needs to be at least UTF8_MAXBYTES_CASE+1 bytes since | |
c3fd2246 KW |
2267 | * the changed version may be longer than the original character. |
2268 | * | |
2269 | * The ordinal of the first character of the changed version is returned | |
2270 | * (but note, as explained above, that there may be more.) */ | |
2271 | ||
7918f24d NC |
2272 | PERL_ARGS_ASSERT_TO_UNI_UPPER; |
2273 | ||
3a4c58c9 KW |
2274 | if (c < 256) { |
2275 | return _to_upper_title_latin1((U8) c, p, lenp, 'S'); | |
2276 | } | |
2277 | ||
0ebc6274 | 2278 | uvchr_to_utf8(p, c); |
b9992569 | 2279 | return CALL_UPPER_CASE(c, p, p, lenp); |
a0ed51b3 LW |
2280 | } |
2281 | ||
84afefe6 JH |
2282 | UV |
2283 | Perl_to_uni_title(pTHX_ UV c, U8* p, STRLEN *lenp) | |
a0ed51b3 | 2284 | { |
7918f24d NC |
2285 | PERL_ARGS_ASSERT_TO_UNI_TITLE; |
2286 | ||
3a4c58c9 KW |
2287 | if (c < 256) { |
2288 | return _to_upper_title_latin1((U8) c, p, lenp, 's'); | |
2289 | } | |
2290 | ||
0ebc6274 | 2291 | uvchr_to_utf8(p, c); |
b9992569 | 2292 | return CALL_TITLE_CASE(c, p, p, lenp); |
a0ed51b3 LW |
2293 | } |
2294 | ||
afc16117 | 2295 | STATIC U8 |
eaf412bf | 2296 | S_to_lower_latin1(const U8 c, U8* p, STRLEN *lenp, const char dummy) |
afc16117 KW |
2297 | { |
2298 | /* We have the latin1-range values compiled into the core, so just use | |
4a4088c4 | 2299 | * those, converting the result to UTF-8. Since the result is always just |
a1433954 | 2300 | * one character, we allow <p> to be NULL */ |
afc16117 KW |
2301 | |
2302 | U8 converted = toLOWER_LATIN1(c); | |
2303 | ||
eaf412bf KW |
2304 | PERL_UNUSED_ARG(dummy); |
2305 | ||
afc16117 | 2306 | if (p != NULL) { |
6f2d5cbc | 2307 | if (NATIVE_BYTE_IS_INVARIANT(converted)) { |
afc16117 KW |
2308 | *p = converted; |
2309 | *lenp = 1; | |
2310 | } | |
2311 | else { | |
430c9760 KW |
2312 | /* Result is known to always be < 256, so can use the EIGHT_BIT |
2313 | * macros */ | |
2314 | *p = UTF8_EIGHT_BIT_HI(converted); | |
2315 | *(p+1) = UTF8_EIGHT_BIT_LO(converted); | |
afc16117 KW |
2316 | *lenp = 2; |
2317 | } | |
2318 | } | |
2319 | return converted; | |
2320 | } | |
2321 | ||
84afefe6 JH |
2322 | UV |
2323 | Perl_to_uni_lower(pTHX_ UV c, U8* p, STRLEN *lenp) | |
a0ed51b3 | 2324 | { |
7918f24d NC |
2325 | PERL_ARGS_ASSERT_TO_UNI_LOWER; |
2326 | ||
afc16117 | 2327 | if (c < 256) { |
eaf412bf | 2328 | return to_lower_latin1((U8) c, p, lenp, 0 /* 0 is a dummy arg */ ); |
bca00c02 KW |
2329 | } |
2330 | ||
afc16117 | 2331 | uvchr_to_utf8(p, c); |
b9992569 | 2332 | return CALL_LOWER_CASE(c, p, p, lenp); |
a0ed51b3 LW |
2333 | } |
2334 | ||
84afefe6 | 2335 | UV |
51910141 | 2336 | Perl__to_fold_latin1(pTHX_ const U8 c, U8* p, STRLEN *lenp, const unsigned int flags) |
a1dde8de | 2337 | { |
51910141 | 2338 | /* Corresponds to to_lower_latin1(); <flags> bits meanings: |
1ca267a5 | 2339 | * FOLD_FLAGS_NOMIX_ASCII iff non-ASCII to ASCII folds are prohibited |
51910141 | 2340 | * FOLD_FLAGS_FULL iff full folding is to be used; |
1ca267a5 KW |
2341 | * |
2342 | * Not to be used for locale folds | |
51910141 | 2343 | */ |
f673fad4 | 2344 | |
a1dde8de KW |
2345 | UV converted; |
2346 | ||
2347 | PERL_ARGS_ASSERT__TO_FOLD_LATIN1; | |
81611534 | 2348 | PERL_UNUSED_CONTEXT; |
a1dde8de | 2349 | |
1ca267a5 KW |
2350 | assert (! (flags & FOLD_FLAGS_LOCALE)); |
2351 | ||
659a7c2d | 2352 | if (UNLIKELY(c == MICRO_SIGN)) { |
a1dde8de KW |
2353 | converted = GREEK_SMALL_LETTER_MU; |
2354 | } | |
9b63e895 KW |
2355 | #if UNICODE_MAJOR_VERSION > 3 /* no multifolds in early Unicode */ \ |
2356 | || (UNICODE_MAJOR_VERSION == 3 && ( UNICODE_DOT_VERSION > 0) \ | |
2357 | || UNICODE_DOT_DOT_VERSION > 0) | |
659a7c2d KW |
2358 | else if ( (flags & FOLD_FLAGS_FULL) |
2359 | && UNLIKELY(c == LATIN_SMALL_LETTER_SHARP_S)) | |
2360 | { | |
1ca267a5 KW |
2361 | /* If can't cross 127/128 boundary, can't return "ss"; instead return |
2362 | * two U+017F characters, as fc("\df") should eq fc("\x{17f}\x{17f}") | |
2363 | * under those circumstances. */ | |
2364 | if (flags & FOLD_FLAGS_NOMIX_ASCII) { | |
2365 | *lenp = 2 * sizeof(LATIN_SMALL_LETTER_LONG_S_UTF8) - 2; | |
2366 | Copy(LATIN_SMALL_LETTER_LONG_S_UTF8 LATIN_SMALL_LETTER_LONG_S_UTF8, | |
2367 | p, *lenp, U8); | |
2368 | return LATIN_SMALL_LETTER_LONG_S; | |
2369 | } | |
2370 | else { | |
4f489194 KW |
2371 | *(p)++ = 's'; |
2372 | *p = 's'; | |
2373 | *lenp = 2; | |
2374 | return 's'; | |
1ca267a5 | 2375 | } |
a1dde8de | 2376 | } |
9b63e895 | 2377 | #endif |
a1dde8de KW |
2378 | else { /* In this range the fold of all other characters is their lower |
2379 | case */ | |
2380 | converted = toLOWER_LATIN1(c); | |
2381 | } | |
2382 | ||
6f2d5cbc | 2383 | if (UVCHR_IS_INVARIANT(converted)) { |
a1dde8de KW |
2384 | *p = (U8) converted; |
2385 | *lenp = 1; | |
2386 | } | |
2387 | else { | |
2388 | *(p)++ = UTF8_TWO_BYTE_HI(converted); | |
2389 | *p = UTF8_TWO_BYTE_LO(converted); | |
2390 | *lenp = 2; | |
2391 | } | |
2392 | ||
2393 | return converted; | |
2394 | } | |
2395 | ||
2396 | UV | |
31f05a37 | 2397 | Perl__to_uni_fold_flags(pTHX_ UV c, U8* p, STRLEN *lenp, U8 flags) |
84afefe6 | 2398 | { |
4b593389 | 2399 | |
a0270393 KW |
2400 | /* Not currently externally documented, and subject to change |
2401 | * <flags> bits meanings: | |
2402 | * FOLD_FLAGS_FULL iff full folding is to be used; | |
31f05a37 KW |
2403 | * FOLD_FLAGS_LOCALE is set iff the rules from the current underlying |
2404 | * locale are to be used. | |
a0270393 KW |
2405 | * FOLD_FLAGS_NOMIX_ASCII iff non-ASCII to ASCII folds are prohibited |
2406 | */ | |
4b593389 | 2407 | |
36bb2ab6 | 2408 | PERL_ARGS_ASSERT__TO_UNI_FOLD_FLAGS; |
7918f24d | 2409 | |
780fcc9f KW |
2410 | if (flags & FOLD_FLAGS_LOCALE) { |
2411 | /* Treat a UTF-8 locale as not being in locale at all */ | |
2412 | if (IN_UTF8_CTYPE_LOCALE) { | |
2413 | flags &= ~FOLD_FLAGS_LOCALE; | |
2414 | } | |
2415 | else { | |
2416 | _CHECK_AND_WARN_PROBLEMATIC_LOCALE; | |
e7b7ac46 | 2417 | goto needs_full_generality; |
780fcc9f | 2418 | } |
31f05a37 KW |
2419 | } |
2420 | ||
a1dde8de | 2421 | if (c < 256) { |
e7b7ac46 | 2422 | return _to_fold_latin1((U8) c, p, lenp, |
31f05a37 | 2423 | flags & (FOLD_FLAGS_FULL | FOLD_FLAGS_NOMIX_ASCII)); |
a1dde8de KW |
2424 | } |
2425 | ||
2f306ab9 | 2426 | /* Here, above 255. If no special needs, just use the macro */ |
a0270393 KW |
2427 | if ( ! (flags & (FOLD_FLAGS_LOCALE|FOLD_FLAGS_NOMIX_ASCII))) { |
2428 | uvchr_to_utf8(p, c); | |
b9992569 | 2429 | return CALL_FOLD_CASE(c, p, p, lenp, flags & FOLD_FLAGS_FULL); |
a0270393 | 2430 | } |
567b353c | 2431 | else { /* Otherwise, _toFOLD_utf8_flags has the intelligence to deal with |
a0270393 KW |
2432 | the special flags. */ |
2433 | U8 utf8_c[UTF8_MAXBYTES + 1]; | |
e7b7ac46 KW |
2434 | |
2435 | needs_full_generality: | |
a0270393 | 2436 | uvchr_to_utf8(utf8_c, c); |
a1a5ec35 | 2437 | return _toFOLD_utf8_flags(utf8_c, utf8_c + sizeof(utf8_c), p, lenp, flags); |
a0270393 | 2438 | } |
84afefe6 JH |
2439 | } |
2440 | ||
26483009 | 2441 | PERL_STATIC_INLINE bool |
5141f98e | 2442 | S_is_utf8_common(pTHX_ const U8 *const p, SV **swash, |
f25ce844 | 2443 | const char *const swashname, SV* const invlist) |
bde6a22d | 2444 | { |
ea317ccb KW |
2445 | /* returns a boolean giving whether or not the UTF8-encoded character that |
2446 | * starts at <p> is in the swash indicated by <swashname>. <swash> | |
2447 | * contains a pointer to where the swash indicated by <swashname> | |
2448 | * is to be stored; which this routine will do, so that future calls will | |
f25ce844 KW |
2449 | * look at <*swash> and only generate a swash if it is not null. <invlist> |
2450 | * is NULL or an inversion list that defines the swash. If not null, it | |
2451 | * saves time during initialization of the swash. | |
ea317ccb KW |
2452 | * |
2453 | * Note that it is assumed that the buffer length of <p> is enough to | |
2454 | * contain all the bytes that comprise the character. Thus, <*p> should | |
2455 | * have been checked before this call for mal-formedness enough to assure | |
2456 | * that. */ | |
2457 | ||
7918f24d NC |
2458 | PERL_ARGS_ASSERT_IS_UTF8_COMMON; |
2459 | ||
492a624f | 2460 | /* The API should have included a length for the UTF-8 character in <p>, |
28123549 | 2461 | * but it doesn't. We therefore assume that p has been validated at least |
492a624f KW |
2462 | * as far as there being enough bytes available in it to accommodate the |
2463 | * character without reading beyond the end, and pass that number on to the | |
2464 | * validating routine */ | |
6302f837 | 2465 | if (! isUTF8_CHAR(p, p + UTF8SKIP(p))) { |
86ae6e94 | 2466 | _force_out_malformed_utf8_message(p, p + UTF8SKIP(p), |
99a765e9 | 2467 | _UTF8_NO_CONFIDENCE_IN_CURLEN, |
86ae6e94 KW |
2468 | 1 /* Die */ ); |
2469 | NOT_REACHED; /* NOTREACHED */ | |
28123549 | 2470 | } |
86ae6e94 | 2471 | |
87367d5f KW |
2472 | if (!*swash) { |
2473 | U8 flags = _CORE_SWASH_INIT_ACCEPT_INVLIST; | |
f25ce844 KW |
2474 | *swash = _core_swash_init("utf8", |
2475 | ||
2476 | /* Only use the name if there is no inversion | |
2477 | * list; otherwise will go out to disk */ | |
2478 | (invlist) ? "" : swashname, | |
2479 | ||
2480 | &PL_sv_undef, 1, 0, invlist, &flags); | |
87367d5f | 2481 | } |
28123549 | 2482 | |
bde6a22d NC |
2483 | return swash_fetch(*swash, p, TRUE) != 0; |
2484 | } | |
2485 | ||
da8c1a98 KW |
2486 | PERL_STATIC_INLINE bool |
2487 | S_is_utf8_common_with_len(pTHX_ const U8 *const p, const U8 * const e, SV **swash, | |
2488 | const char *const swashname, SV* const invlist) | |
2489 | { | |
2490 | /* returns a boolean giving whether or not the UTF8-encoded character that | |
2491 | * starts at <p>, and extending no further than <e - 1> is in the swash | |
2492 | * indicated by <swashname>. <swash> contains a pointer to where the swash | |
2493 | * indicated by <swashname> is to be stored; which this routine will do, so | |
2494 | * that future calls will look at <*swash> and only generate a swash if it | |
2495 | * is not null. <invlist> is NULL or an inversion list that defines the | |
2496 | * swash. If not null, it saves time during initialization of the swash. | |
2497 | */ | |
2498 | ||
2499 | PERL_ARGS_ASSERT_IS_UTF8_COMMON_WITH_LEN; | |
2500 | ||
2501 | if (! isUTF8_CHAR(p, e)) { | |
2502 | _force_out_malformed_utf8_message(p, e, 0, 1); | |
2503 | NOT_REACHED; /* NOTREACHED */ | |
2504 | } | |
2505 | ||
2506 | if (!*swash) { | |
2507 | U8 flags = _CORE_SWASH_INIT_ACCEPT_INVLIST; | |
2508 | *swash = _core_swash_init("utf8", | |
2509 | ||
2510 | /* Only use the name if there is no inversion | |
2511 | * list; otherwise will go out to disk */ | |
2512 | (invlist) ? "" : swashname, | |
2513 | ||
2514 | &PL_sv_undef, 1, 0, invlist, &flags); | |
2515 | } | |
2516 | ||
2517 | return swash_fetch(*swash, p, TRUE) != 0; | |
2518 | } | |
2519 | ||
34aeb2e9 KW |
2520 | STATIC void |
2521 | S_warn_on_first_deprecated_use(pTHX_ const char * const name, | |
2522 | const char * const alternative, | |
2523 | const bool use_locale, | |
2524 | const char * const file, | |
2525 | const unsigned line) | |
2526 | { | |
2527 | const char * key; | |
2528 | ||
2529 | PERL_ARGS_ASSERT_WARN_ON_FIRST_DEPRECATED_USE; | |
2530 | ||
2531 | if (ckWARN_d(WARN_DEPRECATED)) { | |
2532 | ||
2533 | key = Perl_form(aTHX_ "%s;%d;%s;%d", name, use_locale, file, line); | |
2534 | if (! hv_fetch(PL_seen_deprecated_macro, key, strlen(key), 0)) { | |
2535 | if (! PL_seen_deprecated_macro) { | |
2536 | PL_seen_deprecated_macro = newHV(); | |
2537 | } | |
2538 | if (! hv_store(PL_seen_deprecated_macro, key, | |
2539 | strlen(key), &PL_sv_undef, 0)) | |
2540 | { | |
2541 | Perl_croak(aTHX_ "panic: hv_store() unexpectedly failed"); | |
2542 | } | |
2543 | ||
c44e9413 | 2544 | if (instr(file, "mathoms.c")) { |
607313a1 KW |
2545 | Perl_warner(aTHX_ WARN_DEPRECATED, |
2546 | "In %s, line %d, starting in Perl v5.30, %s()" | |
2547 | " will be removed. Avoid this message by" | |
2548 | " converting to use %s().\n", | |
2549 | file, line, name, alternative); | |
2550 | } | |
2551 | else { | |
34aeb2e9 KW |
2552 | Perl_warner(aTHX_ WARN_DEPRECATED, |
2553 | "In %s, line %d, starting in Perl v5.30, %s() will" | |
2554 | " require an additional parameter. Avoid this" | |
2555 | " message by converting to use %s().\n", | |
2556 | file, line, name, alternative); | |
607313a1 | 2557 | } |
34aeb2e9 KW |
2558 | } |
2559 | } | |
2560 | } | |
2561 | ||
bde6a22d | 2562 | bool |
34aeb2e9 KW |
2563 | Perl__is_utf8_FOO(pTHX_ U8 classnum, |
2564 | const U8 *p, | |
2565 | const char * const name, | |
2566 | const char * const alternative, | |
2567 | const bool use_utf8, | |
2568 | const bool use_locale, | |
2569 | const char * const file, | |
2570 | const unsigned line) | |
922e8cb4 | 2571 | { |
922e8cb4 KW |
2572 | PERL_ARGS_ASSERT__IS_UTF8_FOO; |
2573 | ||
34aeb2e9 KW |
2574 | warn_on_first_deprecated_use(name, alternative, use_locale, file, line); |
2575 | ||
2576 | if (use_utf8 && UTF8_IS_ABOVE_LATIN1(*p)) { | |
34aeb2e9 KW |
2577 | |
2578 | switch (classnum) { | |
2579 | case _CC_WORDCHAR: | |
2580 | case _CC_DIGIT: | |
2581 | case _CC_ALPHA: | |
2582 | case _CC_LOWER: | |
2583 | case _CC_UPPER: | |
2584 | case _CC_PUNCT: | |
2585 | case _CC_PRINT: | |
2586 | case _CC_ALPHANUMERIC: | |
2587 | case _CC_GRAPH: | |
2588 | case _CC_CASED: | |
2589 | ||
2590 | return is_utf8_common(p, | |
2591 | &PL_utf8_swash_ptrs[classnum], | |
2592 | swash_property_names[classnum], | |
2593 | PL_XPosix_ptrs[classnum]); | |
2594 | ||
2595 | case _CC_SPACE: | |
2596 | return is_XPERLSPACE_high(p); | |
2597 | case _CC_BLANK: | |
2598 | return is_HORIZWS_high(p); | |
2599 | case _CC_XDIGIT: | |
2600 | return is_XDIGIT_high(p); | |
2601 | case _CC_CNTRL: | |
2602 | return 0; | |
2603 | case _CC_ASCII: | |
2604 | return 0; | |
2605 | case _CC_VERTSPACE: | |
2606 | return is_VERTWS_high(p); | |
2607 | case _CC_IDFIRST: | |
2608 | if (! PL_utf8_perl_idstart) { | |
22f0498f KW |
2609 | PL_utf8_perl_idstart |
2610 | = _new_invlist_C_array(_Perl_IDStart_invlist); | |
34aeb2e9 | 2611 | } |
22f0498f KW |
2612 | return is_utf8_common(p, &PL_utf8_perl_idstart, |
2613 | "_Perl_IDStart", NULL); | |
34aeb2e9 KW |
2614 | case _CC_IDCONT: |
2615 | if (! PL_utf8_perl_idcont) { | |
22f0498f KW |
2616 | PL_utf8_perl_idcont |
2617 | = _new_invlist_C_array(_Perl_IDCont_invlist); | |
34aeb2e9 | 2618 | } |
22f0498f KW |
2619 | return is_utf8_common(p, &PL_utf8_perl_idcont, |
2620 | "_Perl_IDCont", NULL); | |
34aeb2e9 KW |
2621 | } |
2622 | } | |
2623 | ||
2624 | /* idcont is the same as wordchar below 256 */ | |
2625 | if (classnum == _CC_IDCONT) { | |
2626 | classnum = _CC_WORDCHAR; | |
2627 | } | |
2628 | else if (classnum == _CC_IDFIRST) { | |
2629 | if (*p == '_') { | |
2630 | return TRUE; | |
2631 | } | |
2632 | classnum = _CC_ALPHA; | |
2633 | } | |
2634 | ||
2635 | if (! use_locale) { | |
2636 | if (! use_utf8 || UTF8_IS_INVARIANT(*p)) { | |
2637 | return _generic_isCC(*p, classnum); | |
2638 | } | |
922e8cb4 | 2639 | |
34aeb2e9 KW |
2640 | return _generic_isCC(EIGHT_BIT_UTF8_TO_NATIVE(*p, *(p + 1 )), classnum); |
2641 | } | |
2642 | else { | |
2643 | if (! use_utf8 || UTF8_IS_INVARIANT(*p)) { | |
2644 | return isFOO_lc(classnum, *p); | |
2645 | } | |
2646 | ||
2647 | return isFOO_lc(classnum, EIGHT_BIT_UTF8_TO_NATIVE(*p, *(p + 1 ))); | |
2648 | } | |
2649 | ||
2650 | NOT_REACHED; /* NOTREACHED */ | |
922e8cb4 KW |
2651 | } |
2652 | ||
2653 | bool | |
da8c1a98 KW |
2654 | Perl__is_utf8_FOO_with_len(pTHX_ const U8 classnum, const U8 *p, |
2655 | const U8 * const e) | |
2656 | { | |
2657 | PERL_ARGS_ASSERT__IS_UTF8_FOO_WITH_LEN; | |
2658 | ||
2659 | assert(classnum < _FIRST_NON_SWASH_CC); | |
2660 | ||
2661 | return is_utf8_common_with_len(p, | |
2662 | e, | |
2663 | &PL_utf8_swash_ptrs[classnum], | |
2664 | swash_property_names[classnum], | |
2665 | PL_XPosix_ptrs[classnum]); | |
2666 | } | |
2667 | ||
2668 | bool | |
da8c1a98 KW |
2669 | Perl__is_utf8_perl_idstart_with_len(pTHX_ const U8 *p, const U8 * const e) |
2670 | { | |
2671 | SV* invlist = NULL; | |
2672 | ||
2673 | PERL_ARGS_ASSERT__IS_UTF8_PERL_IDSTART_WITH_LEN; | |
2674 | ||
2675 | if (! PL_utf8_perl_idstart) { | |
2676 | invlist = _new_invlist_C_array(_Perl_IDStart_invlist); | |
2677 | } | |
2678 | return is_utf8_common_with_len(p, e, &PL_utf8_perl_idstart, | |
2679 | "_Perl_IDStart", invlist); | |
2680 | } | |
2681 | ||
2682 | bool | |
f2645549 | 2683 | Perl__is_utf8_xidstart(pTHX_ const U8 *p) |
c11ff943 | 2684 | { |
f2645549 | 2685 | PERL_ARGS_ASSERT__IS_UTF8_XIDSTART; |
c11ff943 KW |
2686 | |
2687 | if (*p == '_') | |
2688 | return TRUE; | |
f25ce844 | 2689 | return is_utf8_common(p, &PL_utf8_xidstart, "XIdStart", NULL); |
c11ff943 KW |
2690 | } |
2691 | ||
2692 | bool | |
da8c1a98 KW |
2693 | Perl__is_utf8_perl_idcont_with_len(pTHX_ const U8 *p, const U8 * const e) |
2694 | { | |
2695 | SV* invlist = NULL; | |
2696 | ||
2697 | PERL_ARGS_ASSERT__IS_UTF8_PERL_IDCONT_WITH_LEN; | |
2698 | ||
2699 | if (! PL_utf8_perl_idcont) { | |
2700 | invlist = _new_invlist_C_array(_Perl_IDCont_invlist); | |
2701 | } | |
2702 | return is_utf8_common_with_len(p, e, &PL_utf8_perl_idcont, | |
2703 | "_Perl_IDCont", invlist); | |
2704 | } | |
2705 | ||
2706 | bool | |
f2645549 | 2707 | Perl__is_utf8_idcont(pTHX_ const U8 *p) |
82686b01 | 2708 | { |
f2645549 | 2709 | PERL_ARGS_ASSERT__IS_UTF8_IDCONT; |
7918f24d | 2710 | |
f25ce844 | 2711 | return is_utf8_common(p, &PL_utf8_idcont, "IdContinue", NULL); |
a0ed51b3 LW |
2712 | } |
2713 | ||
2714 | bool | |
f2645549 | 2715 | Perl__is_utf8_xidcont(pTHX_ const U8 *p) |
c11ff943 | 2716 | { |
f2645549 | 2717 | PERL_ARGS_ASSERT__IS_UTF8_XIDCONT; |
c11ff943 | 2718 | |
f25ce844 | 2719 | return is_utf8_common(p, &PL_utf8_idcont, "XIdContinue", NULL); |
c11ff943 KW |
2720 | } |
2721 | ||
2722 | bool | |
7dbf68d2 KW |
2723 | Perl__is_utf8_mark(pTHX_ const U8 *p) |
2724 | { | |
7dbf68d2 KW |
2725 | PERL_ARGS_ASSERT__IS_UTF8_MARK; |
2726 | ||
f25ce844 | 2727 | return is_utf8_common(p, &PL_utf8_mark, "IsM", NULL); |
7dbf68d2 KW |
2728 | } |
2729 | ||
6b5c0936 | 2730 | /* |
87cea99e | 2731 | =for apidoc to_utf8_case |
6b5c0936 | 2732 | |
a239b1e2 KW |
2733 | Instead use the appropriate one of L</toUPPER_utf8_safe>, |
2734 | L</toTITLE_utf8_safe>, | |
2735 | L</toLOWER_utf8_safe>, | |
2736 | or L</toFOLD_utf8_safe>. | |
9da1e7cb | 2737 | |
45de9c41 KW |
2738 | This function will be removed in Perl v5.28. |
2739 | ||
6fae5207 | 2740 | C<p> contains the pointer to the UTF-8 string encoding |
a1433954 KW |
2741 | the character that is being converted. This routine assumes that the character |
2742 | at C<p> is well-formed. | |
6b5c0936 | 2743 | |
6fae5207 KW |
2744 | C<ustrp> is a pointer to the character buffer to put the |
2745 | conversion result to. C<lenp> is a pointer to the length | |
6b5c0936 JH |
2746 | of the result. |
2747 | ||
6fae5207 | 2748 | C<swashp> is a pointer to the swash to use. |
6b5c0936 | 2749 | |
a1433954 | 2750 | Both the special and normal mappings are stored in F<lib/unicore/To/Foo.pl>, |
796b6530 | 2751 | and loaded by C<SWASHNEW>, using F<lib/utf8_heavy.pl>. C<special> (usually, |
0134edef | 2752 | but not always, a multicharacter mapping), is tried first. |
6b5c0936 | 2753 | |
4a8240a3 KW |
2754 | C<special> is a string, normally C<NULL> or C<"">. C<NULL> means to not use |
2755 | any special mappings; C<""> means to use the special mappings. Values other | |
2756 | than these two are treated as the name of the hash containing the special | |
2757 | mappings, like C<"utf8::ToSpecLower">. | |
6b5c0936 | 2758 | |
796b6530 KW |
2759 | C<normal> is a string like C<"ToLower"> which means the swash |
2760 | C<%utf8::ToLower>. | |
0134edef | 2761 | |
760c7c2f KW |
2762 | Code points above the platform's C<IV_MAX> will raise a deprecation warning, |
2763 | unless those are turned off. | |
2764 | ||
0134edef | 2765 | =cut */ |
6b5c0936 | 2766 | |
2104c8d9 | 2767 | UV |
9a957fbc AL |
2768 | Perl_to_utf8_case(pTHX_ const U8 *p, U8* ustrp, STRLEN *lenp, |
2769 | SV **swashp, const char *normal, const char *special) | |
a0ed51b3 | 2770 | { |
45de9c41 KW |
2771 | STRLEN len_cp; |
2772 | UV cp; | |
2773 | const U8 * e = p + UTF8SKIP(p); | |
2774 | ||
b9992569 KW |
2775 | PERL_ARGS_ASSERT_TO_UTF8_CASE; |
2776 | ||
45de9c41 KW |
2777 | cp = utf8n_to_uvchr(p, e - p, &len_cp, UTF8_CHECK_ONLY); |
2778 | if (len_cp == (STRLEN) -1) { | |
2779 | _force_out_malformed_utf8_message(p, e, | |
2780 | _UTF8_NO_CONFIDENCE_IN_CURLEN, 1 /* Die */ ); | |
2781 | } | |
2782 | ||
2783 | return _to_utf8_case(cp, p, ustrp, lenp, swashp, normal, special); | |
b9992569 KW |
2784 | } |
2785 | ||
2786 | /* change namve uv1 to 'from' */ | |
6a4a25f4 | 2787 | STATIC UV |
b9992569 KW |
2788 | S__to_utf8_case(pTHX_ const UV uv1, const U8 *p, U8* ustrp, STRLEN *lenp, |
2789 | SV **swashp, const char *normal, const char *special) | |
2790 | { | |
0134edef | 2791 | STRLEN len = 0; |
7918f24d | 2792 | |
b9992569 | 2793 | PERL_ARGS_ASSERT__TO_UTF8_CASE; |
7918f24d | 2794 | |
36eaa811 KW |
2795 | /* For code points that don't change case, we already know that the output |
2796 | * of this function is the unchanged input, so we can skip doing look-ups | |
2797 | * for them. Unfortunately the case-changing code points are scattered | |
2798 | * around. But there are some long consecutive ranges where there are no | |
2799 | * case changing code points. By adding tests, we can eliminate the lookup | |
2800 | * for all the ones in such ranges. This is currently done here only for | |
2801 | * just a few cases where the scripts are in common use in modern commerce | |
2802 | * (and scripts adjacent to those which can be included without additional | |
2803 | * tests). */ | |
2804 | ||
2805 | if (uv1 >= 0x0590) { | |
2806 | /* This keeps from needing further processing the code points most | |
2807 | * likely to be used in the following non-cased scripts: Hebrew, | |
2808 | * Arabic, Syriac, Thaana, NKo, Samaritan, Mandaic, Devanagari, | |
2809 | * Bengali, Gurmukhi, Gujarati, Oriya, Tamil, Telugu, Kannada, | |
2810 | * Malayalam, Sinhala, Thai, Lao, Tibetan, Myanmar */ | |
2811 | if (uv1 < 0x10A0) { | |
2812 | goto cases_to_self; | |
2813 | } | |
2814 | ||
2815 | /* The following largish code point ranges also don't have case | |
2816 | * changes, but khw didn't think they warranted extra tests to speed | |
2817 | * them up (which would slightly slow down everything else above them): | |
2818 | * 1100..139F Hangul Jamo, Ethiopic | |
2819 | * 1400..1CFF Unified Canadian Aboriginal Syllabics, Ogham, Runic, | |
2820 | * Tagalog, Hanunoo, Buhid, Tagbanwa, Khmer, Mongolian, | |
2821 | * Limbu, Tai Le, New Tai Lue, Buginese, Tai Tham, | |
2822 | * Combining Diacritical Marks Extended, Balinese, | |
2823 | * Sundanese, Batak, Lepcha, Ol Chiki | |
2824 | * 2000..206F General Punctuation | |
2825 | */ | |
2826 | ||
2827 | if (uv1 >= 0x2D30) { | |
2828 | ||
2829 | /* This keeps the from needing further processing the code points | |
2830 | * most likely to be used in the following non-cased major scripts: | |
2831 | * CJK, Katakana, Hiragana, plus some less-likely scripts. | |
2832 | * | |
2833 | * (0x2D30 above might have to be changed to 2F00 in the unlikely | |
2834 | * event that Unicode eventually allocates the unused block as of | |
2835 | * v8.0 2FE0..2FEF to code points that are cased. khw has verified | |
2836 | * that the test suite will start having failures to alert you | |
2837 | * should that happen) */ | |
2838 | if (uv1 < 0xA640) { | |
2839 | goto cases_to_self; | |
2840 | } | |
2841 | ||
2842 | if (uv1 >= 0xAC00) { | |
2843 | if (UNLIKELY(UNICODE_IS_SURROGATE(uv1))) { | |
5af9bc97 KW |
2844 | if (ckWARN_d(WARN_SURROGATE)) { |
2845 | const char* desc = (PL_op) ? OP_DESC(PL_op) : normal; | |
2846 | Perl_warner(aTHX_ packWARN(WARN_SURROGATE), | |
147e3846 | 2847 | "Operation \"%s\" returns its argument for UTF-16 surrogate U+%04" UVXf, desc, uv1); |
5af9bc97 KW |
2848 | } |
2849 | goto cases_to_self; | |
2850 | } | |
36eaa811 KW |
2851 | |
2852 | /* AC00..FAFF Catches Hangul syllables and private use, plus | |
2853 | * some others */ | |
2854 | if (uv1 < 0xFB00) { | |
2855 | goto cases_to_self; | |
2856 | ||
2857 | } | |
2858 | ||
5af9bc97 KW |
2859 | if (UNLIKELY(UNICODE_IS_SUPER(uv1))) { |
2860 | if ( UNLIKELY(uv1 > MAX_NON_DEPRECATED_CP) | |
2861 | && ckWARN_d(WARN_DEPRECATED)) | |
2862 | { | |
2863 | Perl_warner(aTHX_ packWARN(WARN_DEPRECATED), | |
2864 | cp_above_legal_max, uv1, MAX_NON_DEPRECATED_CP); | |
2865 | } | |
2866 | if (ckWARN_d(WARN_NON_UNICODE)) { | |
2867 | const char* desc = (PL_op) ? OP_DESC(PL_op) : normal; | |
2868 | Perl_warner(aTHX_ packWARN(WARN_NON_UNICODE), | |
147e3846 | 2869 | "Operation \"%s\" returns its argument for non-Unicode code point 0x%04" UVXf, desc, uv1); |
5af9bc97 KW |
2870 | } |
2871 | goto cases_to_self; | |
2872 | } | |
3bfc1e70 KW |
2873 | #ifdef HIGHEST_CASE_CHANGING_CP_FOR_USE_ONLY_BY_UTF8_DOT_C |
2874 | if (UNLIKELY(uv1 | |
2875 | > HIGHEST_CASE_CHANGING_CP_FOR_USE_ONLY_BY_UTF8_DOT_C)) | |
2876 | { | |
2877 | ||
2878 | /* As of this writing, this means we avoid swash creation | |
2879 | * for anything beyond low Plane 1 */ | |
2880 | goto cases_to_self; | |
2881 | } | |
2882 | #endif | |
36eaa811 KW |
2883 | } |
2884 | } | |
9ae3ac1a | 2885 | |
36eaa811 KW |
2886 | /* Note that non-characters are perfectly legal, so no warning should |
2887 | * be given. There are so few of them, that it isn't worth the extra | |
2888 | * tests to avoid swash creation */ | |
9ae3ac1a KW |
2889 | } |
2890 | ||
0134edef | 2891 | if (!*swashp) /* load on-demand */ |
5ab9d2ef | 2892 | *swashp = _core_swash_init("utf8", normal, &PL_sv_undef, 4, 0, NULL, NULL); |
0134edef | 2893 | |
a6f87d8c | 2894 | if (special) { |
0134edef | 2895 | /* It might be "special" (sometimes, but not always, |
2a37f04d | 2896 | * a multicharacter mapping) */ |
4a8240a3 | 2897 | HV *hv = NULL; |
b08cf34e JH |
2898 | SV **svp; |
2899 | ||
4a8240a3 KW |
2900 | /* If passed in the specials name, use that; otherwise use any |
2901 | * given in the swash */ | |
2902 | if (*special != '\0') { | |
2903 | hv = get_hv(special, 0); | |
2904 | } | |
2905 | else { | |
2906 | svp = hv_fetchs(MUTABLE_HV(SvRV(*swashp)), "SPECIALS", 0); | |
2907 | if (svp) { | |
2908 | hv = MUTABLE_HV(SvRV(*svp)); | |
2909 | } | |
2910 | } | |
2911 | ||
176fe009 | 2912 | if (hv |
5f560d8a | 2913 | && (svp = hv_fetch(hv, (const char*)p, UVCHR_SKIP(uv1), FALSE)) |
176fe009 KW |
2914 | && (*svp)) |
2915 | { | |
cfd0369c | 2916 | const char *s; |
47654450 | 2917 | |
cfd0369c | 2918 | s = SvPV_const(*svp, len); |
47654450 | 2919 | if (len == 1) |
f4cd282c | 2920 | /* EIGHTBIT */ |
c80e42f3 | 2921 | len = uvchr_to_utf8(ustrp, *(U8*)s) - ustrp; |
2a37f04d | 2922 | else { |
d2dcd0fb | 2923 | Copy(s, ustrp, len, U8); |
29e98929 | 2924 | } |
983ffd37 | 2925 | } |
0134edef JH |
2926 | } |
2927 | ||
2928 | if (!len && *swashp) { | |
4a4088c4 | 2929 | const UV uv2 = swash_fetch(*swashp, p, TRUE /* => is UTF-8 */); |
d4c19fe8 | 2930 | |
0134edef JH |
2931 | if (uv2) { |
2932 | /* It was "normal" (a single character mapping). */ | |
f4cd282c | 2933 | len = uvchr_to_utf8(ustrp, uv2) - ustrp; |
2a37f04d JH |
2934 | } |
2935 | } | |
1feea2c7 | 2936 | |
cbe07460 KW |
2937 | if (len) { |
2938 | if (lenp) { | |
2939 | *lenp = len; | |
2940 | } | |
2941 | return valid_utf8_to_uvchr(ustrp, 0); | |
2942 | } | |
2943 | ||
2944 | /* Here, there was no mapping defined, which means that the code point maps | |
2945 | * to itself. Return the inputs */ | |
e24dfe9c | 2946 | cases_to_self: |
bfdf22ec | 2947 | len = UTF8SKIP(p); |
ca9fab46 KW |
2948 | if (p != ustrp) { /* Don't copy onto itself */ |
2949 | Copy(p, ustrp, len, U8); | |
2950 | } | |
0134edef | 2951 | |
2a37f04d JH |
2952 | if (lenp) |
2953 | *lenp = len; | |
2954 | ||
f4cd282c | 2955 | return uv1; |
cbe07460 | 2956 | |
a0ed51b3 LW |
2957 | } |
2958 | ||
051a06d4 | 2959 | STATIC UV |
357aadde | 2960 | S_check_locale_boundary_crossing(pTHX_ const U8* const p, const UV result, U8* const ustrp, STRLEN *lenp) |
051a06d4 | 2961 | { |
4a4088c4 | 2962 | /* This is called when changing the case of a UTF-8-encoded character above |
31f05a37 KW |
2963 | * the Latin1 range, and the operation is in a non-UTF-8 locale. If the |
2964 | * result contains a character that crosses the 255/256 boundary, disallow | |
2965 | * the change, and return the original code point. See L<perlfunc/lc> for | |
2966 | * why; | |
051a06d4 | 2967 | * |
a1433954 KW |
2968 | * p points to the original string whose case was changed; assumed |
2969 | * by this routine to be well-formed | |
051a06d4 KW |
2970 | * result the code point of the first character in the changed-case string |
2971 | * ustrp points to the changed-case string (<result> represents its first char) | |
2972 | * lenp points to the length of <ustrp> */ | |
2973 | ||
2974 | UV original; /* To store the first code point of <p> */ | |
2975 | ||
2976 | PERL_ARGS_ASSERT_CHECK_LOCALE_BOUNDARY_CROSSING; | |
2977 | ||
a4f12ed7 | 2978 | assert(UTF8_IS_ABOVE_LATIN1(*p)); |
051a06d4 KW |
2979 | |
2980 | /* We know immediately if the first character in the string crosses the | |
2981 | * boundary, so can skip */ | |
2982 | if (result > 255) { | |
2983 | ||
2984 | /* Look at every character in the result; if any cross the | |
2985 | * boundary, the whole thing is disallowed */ | |
2986 | U8* s = ustrp + UTF8SKIP(ustrp); | |
2987 | U8* e = ustrp + *lenp; | |
2988 | while (s < e) { | |
a4f12ed7 | 2989 | if (! UTF8_IS_ABOVE_LATIN1(*s)) { |
051a06d4 KW |
2990 | goto bad_crossing; |
2991 | } | |
2992 | s += UTF8SKIP(s); | |
2993 | } | |
2994 | ||
613abc6d KW |
2995 | /* Here, no characters crossed, result is ok as-is, but we warn. */ |
2996 | _CHECK_AND_OUTPUT_WIDE_LOCALE_UTF8_MSG(p, p + UTF8SKIP(p)); | |
051a06d4 KW |
2997 | return result; |
2998 | } | |
2999 | ||
7b52d656 | 3000 | bad_crossing: |
051a06d4 KW |
3001 | |
3002 | /* Failed, have to return the original */ | |
4b88fb76 | 3003 | original = valid_utf8_to_uvchr(p, lenp); |
ab0b796c KW |
3004 | |
3005 | /* diag_listed_as: Can't do %s("%s") on non-UTF-8 locale; resolved to "%s". */ | |
3006 | Perl_ck_warner(aTHX_ packWARN(WARN_LOCALE), | |
147e3846 KW |
3007 | "Can't do %s(\"\\x{%" UVXf "}\") on non-UTF-8 locale; " |
3008 | "resolved to \"\\x{%" UVXf "}\".", | |
357aadde | 3009 | OP_DESC(PL_op), |
ab0b796c KW |
3010 | original, |
3011 | original); | |
051a06d4 KW |
3012 | Copy(p, ustrp, *lenp, char); |
3013 | return original; | |
3014 | } | |
3015 | ||
607313a1 KW |
3016 | STATIC U32 |
3017 | S_check_and_deprecate(pTHX_ const U8 *p, | |
3018 | const U8 **e, | |
3019 | const unsigned int type, /* See below */ | |
3020 | const bool use_locale, /* Is this a 'LC_' | |
3021 | macro call? */ | |
3022 | const char * const file, | |
3023 | const unsigned line) | |
3024 | { | |
3025 | /* This is a temporary function to deprecate the unsafe calls to the case | |
3026 | * changing macros and functions. It keeps all the special stuff in just | |
3027 | * one place. | |
3028 | * | |
3029 | * It updates *e with the pointer to the end of the input string. If using | |
3030 | * the old-style macros, *e is NULL on input, and so this function assumes | |
3031 | * the input string is long enough to hold the entire UTF-8 sequence, and | |
3032 | * sets *e accordingly, but it then returns a flag to pass the | |
3033 | * utf8n_to_uvchr(), to tell it that this size is a guess, and to avoid | |
3034 | * using the full length if possible. | |
3035 | * | |
3036 | * It also does the assert that *e > p when *e is not NULL. This should be | |
3037 | * migrated to the callers when this function gets deleted. | |
3038 | * | |
3039 | * The 'type' parameter is used for the caller to specify which case | |
3040 | * changing function this is called from: */ | |
3041 | ||
3042 | # define DEPRECATE_TO_UPPER 0 | |
3043 | # define DEPRECATE_TO_TITLE 1 | |
3044 | # define DEPRECATE_TO_LOWER 2 | |
3045 | # define DEPRECATE_TO_FOLD 3 | |
3046 | ||
3047 | U32 utf8n_flags = 0; | |
3048 | const char * name; | |
3049 | const char * alternative; | |
3050 | ||
3051 | PERL_ARGS_ASSERT_CHECK_AND_DEPRECATE; | |
3052 | ||
3053 | if (*e == NULL) { | |
3054 | utf8n_flags = _UTF8_NO_CONFIDENCE_IN_CURLEN; | |
3055 | *e = p + UTF8SKIP(p); | |
3056 | ||
3057 | /* For mathoms.c calls, we use the function name we know is stored | |
c44e9413 | 3058 | * there. It could be part of a larger path */ |
607313a1 | 3059 | if (type == DEPRECATE_TO_UPPER) { |
c44e9413 | 3060 | name = instr(file, "mathoms.c") |
607313a1 KW |
3061 | ? "to_utf8_upper" |
3062 | : "toUPPER_utf8"; | |
3063 | alternative = "toUPPER_utf8_safe"; | |
3064 | } | |
3065 | else if (type == DEPRECATE_TO_TITLE) { | |
c44e9413 | 3066 | name = instr(file, "mathoms.c") |
607313a1 KW |
3067 | ? "to_utf8_title" |
3068 | : "toTITLE_utf8"; | |
3069 | alternative = "toTITLE_utf8_safe"; | |
3070 | } | |
3071 | else if (type == DEPRECATE_TO_LOWER) { | |
c44e9413 | 3072 | name = instr(file, "mathoms.c") |
607313a1 KW |
3073 | ? "to_utf8_lower" |
3074 | : "toLOWER_utf8"; | |
3075 | alternative = "toLOWER_utf8_safe"; | |
3076 | } | |
3077 | else if (type == DEPRECATE_TO_FOLD) { | |
c44e9413 | 3078 | name = instr(file, "mathoms.c") |
607313a1 KW |
3079 | ? "to_utf8_fold" |
3080 | : "toFOLD_utf8"; | |
3081 | alternative = "toFOLD_utf8_safe"; | |
3082 | } | |
3083 | else Perl_croak(aTHX_ "panic: Unexpected case change type"); | |
3084 | ||
3085 | warn_on_first_deprecated_use(name, alternative, use_locale, file, line); | |
3086 | } | |
3087 | else { | |
3088 | assert (p < *e); | |
3089 | } | |
3090 | ||
3091 | return utf8n_flags; | |
3092 | } | |
3093 | ||
eaf412bf KW |
3094 | /* The process for changing the case is essentially the same for the four case |
3095 | * change types, except there are complications for folding. Otherwise the | |
3096 | * difference is only which case to change to. To make sure that they all do | |
3097 | * the same thing, the bodies of the functions are extracted out into the | |
3098 | * following two macros. The functions are written with the same variable | |
3099 | * names, and these are known and used inside these macros. It would be | |
3100 | * better, of course, to have inline functions to do it, but since different | |
3101 | * macros are called, depending on which case is being changed to, this is not | |
3102 | * feasible in C (to khw's knowledge). Two macros are created so that the fold | |
3103 | * function can start with the common start macro, then finish with its special | |
3104 | * handling; while the other three cases can just use the common end macro. | |
3105 | * | |
3106 | * The algorithm is to use the proper (passed in) macro or function to change | |
3107 | * the case for code points that are below 256. The macro is used if using | |
3108 | * locale rules for the case change; the function if not. If the code point is | |
3109 | * above 255, it is computed from the input UTF-8, and another macro is called | |
3110 | * to do the conversion. If necessary, the output is converted to UTF-8. If | |
3111 | * using a locale, we have to check that the change did not cross the 255/256 | |
3112 | * boundary, see check_locale_boundary_crossing() for further details. | |
3113 | * | |
3114 | * The macros are split with the correct case change for the below-256 case | |
3115 | * stored into 'result', and in the middle of an else clause for the above-255 | |
3116 | * case. At that point in the 'else', 'result' is not the final result, but is | |
3117 | * the input code point calculated from the UTF-8. The fold code needs to | |
3118 | * realize all this and take it from there. | |
3119 | * | |
3120 | * If you read the two macros as sequential, it's easier to understand what's | |
3121 | * going on. */ | |
3122 | #define CASE_CHANGE_BODY_START(locale_flags, LC_L1_change_macro, L1_func, \ | |
3123 | L1_func_extra_param) \ | |
a239b1e2 | 3124 | \ |
eaf412bf KW |
3125 | if (flags & (locale_flags)) { \ |
3126 | /* Treat a UTF-8 locale as not being in locale at all */ \ | |
3127 | if (IN_UTF8_CTYPE_LOCALE) { \ | |
3128 | flags &= ~(locale_flags); \ | |
3129 | } \ | |
3130 | else { \ | |
3131 | _CHECK_AND_WARN_PROBLEMATIC_LOCALE; \ | |
3132 | } \ | |
3133 | } \ | |
3134 | \ | |
3135 | if (UTF8_IS_INVARIANT(*p)) { \ | |
3136 | if (flags & (locale_flags)) { \ | |
3137 | result = LC_L1_change_macro(*p); \ | |
3138 | } \ | |
3139 | else { \ | |
3140 | return L1_func(*p, ustrp, lenp, L1_func_extra_param); \ | |
3141 | } \ | |
3142 | } \ | |
a239b1e2 | 3143 | else if UTF8_IS_NEXT_CHAR_DOWNGRADEABLE(p, e) { \ |
eaf412bf KW |
3144 | if (flags & (locale_flags)) { \ |
3145 | result = LC_L1_change_macro(EIGHT_BIT_UTF8_TO_NATIVE(*p, \ | |
3146 | *(p+1))); \ | |
3147 | } \ | |
3148 | else { \ | |
3149 | return L1_func(EIGHT_BIT_UTF8_TO_NATIVE(*p, *(p+1)), \ | |
3150 | ustrp, lenp, L1_func_extra_param); \ | |
3151 | } \ | |
3152 | } \ | |
fa8ab374 KW |
3153 | else { /* malformed UTF-8 or ord above 255 */ \ |
3154 | STRLEN len_result; \ | |
fa8ab374 KW |
3155 | result = utf8n_to_uvchr(p, e - p, &len_result, UTF8_CHECK_ONLY); \ |
3156 | if (len_result == (STRLEN) -1) { \ | |
607313a1 KW |
3157 | _force_out_malformed_utf8_message(p, e, utf8n_flags, \ |
3158 | 1 /* Die */ ); \ | |
fa8ab374 | 3159 | } |
eaf412bf KW |
3160 | |
3161 | #define CASE_CHANGE_BODY_END(locale_flags, change_macro) \ | |
3162 | result = change_macro(result, p, ustrp, lenp); \ | |
3163 | \ | |
3164 | if (flags & (locale_flags)) { \ | |
3165 | result = check_locale_boundary_crossing(p, result, ustrp, lenp); \ | |
3166 | } \ | |
3167 | return result; \ | |
3168 | } \ | |
3169 | \ | |
3170 | /* Here, used locale rules. Convert back to UTF-8 */ \ | |
3171 | if (UTF8_IS_INVARIANT(result)) { \ | |
3172 | *ustrp = (U8) result; \ | |
3173 | *lenp = 1; \ | |
3174 | } \ | |
3175 | else { \ | |
3176 | *ustrp = UTF8_EIGHT_BIT_HI((U8) result); \ | |
3177 | *(ustrp + 1) = UTF8_EIGHT_BIT_LO((U8) result); \ | |
3178 | *lenp = 2; \ | |
3179 | } \ | |
3180 | \ | |
3181 | return result; | |
3182 | ||
d3e79532 | 3183 | /* |
87cea99e | 3184 | =for apidoc to_utf8_upper |
d3e79532 | 3185 | |
a239b1e2 | 3186 | Instead use L</toUPPER_utf8_safe>. |
a1433954 | 3187 | |
d3e79532 JH |
3188 | =cut */ |
3189 | ||
051a06d4 | 3190 | /* Not currently externally documented, and subject to change: |
31f05a37 KW |
3191 | * <flags> is set iff iff the rules from the current underlying locale are to |
3192 | * be used. */ | |
051a06d4 | 3193 | |
2104c8d9 | 3194 | UV |
607313a1 KW |
3195 | Perl__to_utf8_upper_flags(pTHX_ const U8 *p, |
3196 | const U8 *e, | |
3197 | U8* ustrp, | |
3198 | STRLEN *lenp, | |
3199 | bool flags, | |
3200 | const char * const file, | |
3201 | const int line) | |
a0ed51b3 | 3202 | { |
051a06d4 | 3203 | UV result; |
607313a1 KW |
3204 | const U32 utf8n_flags = check_and_deprecate(p, &e, DEPRECATE_TO_UPPER, |
3205 | cBOOL(flags), file, line); | |
051a06d4 KW |
3206 | |
3207 | PERL_ARGS_ASSERT__TO_UTF8_UPPER_FLAGS; | |
7918f24d | 3208 | |
eaf412bf KW |
3209 | /* ~0 makes anything non-zero in 'flags' mean we are using locale rules */ |
3210 | /* 2nd char of uc(U+DF) is 'S' */ | |
3211 | CASE_CHANGE_BODY_START(~0, toUPPER_LC, _to_upper_title_latin1, 'S'); | |
3212 | CASE_CHANGE_BODY_END (~0, CALL_UPPER_CASE); | |
983ffd37 | 3213 | } |
a0ed51b3 | 3214 | |
d3e79532 | 3215 | /* |
87cea99e | 3216 | =for apidoc to_utf8_title |
d3e79532 | 3217 | |
a239b1e2 | 3218 | Instead use L</toTITLE_utf8_safe>. |
a1433954 | 3219 | |
d3e79532 JH |
3220 | =cut */ |
3221 | ||
051a06d4 | 3222 | /* Not currently externally documented, and subject to change: |
31f05a37 KW |
3223 | * <flags> is set iff the rules from the current underlying locale are to be |
3224 | * used. Since titlecase is not defined in POSIX, for other than a | |
3225 | * UTF-8 locale, uppercase is used instead for code points < 256. | |
445bf929 | 3226 | */ |
051a06d4 | 3227 | |
983ffd37 | 3228 | UV |
607313a1 KW |
3229 | Perl__to_utf8_title_flags(pTHX_ const U8 *p, |
3230 | const U8 *e, | |
3231 | U8* ustrp, | |
3232 | STRLEN *lenp, | |
3233 | bool flags, | |
3234 | const char * const file, | |
3235 | const int line) | |
983ffd37 | 3236 | { |
051a06d4 | 3237 | UV result; |
607313a1 KW |
3238 | const U32 utf8n_flags = check_and_deprecate(p, &e, DEPRECATE_TO_TITLE, |
3239 | cBOOL(flags), file, line); | |
051a06d4 KW |
3240 | |
3241 | PERL_ARGS_ASSERT__TO_UTF8_TITLE_FLAGS; | |
7918f24d | 3242 | |
eaf412bf KW |
3243 | /* 2nd char of ucfirst(U+DF) is 's' */ |
3244 | CASE_CHANGE_BODY_START(~0, toUPPER_LC, _to_upper_title_latin1, 's'); | |
3245 | CASE_CHANGE_BODY_END (~0, CALL_TITLE_CASE); | |
a0ed51b3 LW |
3246 | } |
3247 | ||
d3e79532 | 3248 | /* |
87cea99e | 3249 | =for apidoc to_utf8_lower |
d3e79532 | 3250 | |
a239b1e2 | 3251 | Instead use L</toLOWER_utf8_safe>. |
a1433954 | 3252 | |
d3e79532 JH |
3253 | =cut */ |
3254 | ||
051a06d4 | 3255 | /* Not currently externally documented, and subject to change: |
31f05a37 KW |
3256 | * <flags> is set iff iff the rules from the current underlying locale are to |
3257 | * be used. | |
3258 | */ | |
051a06d4 | 3259 | |
2104c8d9 | 3260 | UV |
607313a1 KW |
3261 | Perl__to_utf8_lower_flags(pTHX_ const U8 *p, |
3262 | const U8 *e, | |
3263 | U8* ustrp, | |
3264 | STRLEN *lenp, | |
3265 | bool flags, | |
3266 | const char * const file, | |
3267 | const int line) | |
a0ed51b3 | 3268 | { |
051a06d4 | 3269 | UV result; |
607313a1 KW |
3270 | const U32 utf8n_flags = check_and_deprecate(p, &e, DEPRECATE_TO_LOWER, |
3271 | cBOOL(flags), file, line); | |
051a06d4 | 3272 | |
051a06d4 | 3273 | PERL_ARGS_ASSERT__TO_UTF8_LOWER_FLAGS; |
7918f24d | 3274 | |
eaf412bf KW |
3275 | CASE_CHANGE_BODY_START(~0, toLOWER_LC, to_lower_latin1, 0 /* 0 is dummy */) |
3276 | CASE_CHANGE_BODY_END (~0, CALL_LOWER_CASE) | |
b4e400f9 JH |
3277 | } |
3278 | ||
d3e79532 | 3279 | /* |
87cea99e | 3280 | =for apidoc to_utf8_fold |
d3e79532 | 3281 | |
a239b1e2 | 3282 | Instead use L</toFOLD_utf8_safe>. |
a1433954 | 3283 | |
d3e79532 JH |
3284 | =cut */ |
3285 | ||
051a06d4 KW |
3286 | /* Not currently externally documented, and subject to change, |
3287 | * in <flags> | |
31f05a37 KW |
3288 | * bit FOLD_FLAGS_LOCALE is set iff the rules from the current underlying |
3289 | * locale are to be used. | |
051a06d4 KW |
3290 | * bit FOLD_FLAGS_FULL is set iff full case folds are to be used; |
3291 | * otherwise simple folds | |
a0270393 KW |
3292 | * bit FOLD_FLAGS_NOMIX_ASCII is set iff folds of non-ASCII to ASCII are |
3293 | * prohibited | |
445bf929 | 3294 | */ |
36bb2ab6 | 3295 | |
b4e400f9 | 3296 | UV |
607313a1 KW |
3297 | Perl__to_utf8_fold_flags(pTHX_ const U8 *p, |
3298 | const U8 *e, | |
3299 | U8* ustrp, | |
3300 | STRLEN *lenp, | |
3301 | U8 flags, | |
3302 | const char * const file, | |
3303 | const int line) | |
b4e400f9 | 3304 | { |
051a06d4 | 3305 | UV result; |
607313a1 KW |
3306 | const U32 utf8n_flags = check_and_deprecate(p, &e, DEPRECATE_TO_FOLD, |
3307 | cBOOL(flags), file, line); | |
051a06d4 | 3308 | |
36bb2ab6 | 3309 | PERL_ARGS_ASSERT__TO_UTF8_FOLD_FLAGS; |
7918f24d | 3310 | |
a0270393 KW |
3311 | /* These are mutually exclusive */ |
3312 | assert (! ((flags & FOLD_FLAGS_LOCALE) && (flags & FOLD_FLAGS_NOMIX_ASCII))); | |
3313 | ||
50ba90ff KW |
3314 | assert(p != ustrp); /* Otherwise overwrites */ |
3315 | ||
eaf412bf KW |
3316 | CASE_CHANGE_BODY_START(FOLD_FLAGS_LOCALE, toFOLD_LC, _to_fold_latin1, |
3317 | ((flags) & (FOLD_FLAGS_FULL | FOLD_FLAGS_NOMIX_ASCII))); | |
31f05a37 | 3318 | |
eaf412bf | 3319 | result = CALL_FOLD_CASE(result, p, ustrp, lenp, flags & FOLD_FLAGS_FULL); |
a1dde8de | 3320 | |
1ca267a5 KW |
3321 | if (flags & FOLD_FLAGS_LOCALE) { |
3322 | ||
76f2ffcd | 3323 | # define LONG_S_T LATIN_SMALL_LIGATURE_LONG_S_T_UTF8 |
0766489e KW |
3324 | const unsigned int long_s_t_len = sizeof(LONG_S_T) - 1; |
3325 | ||
3326 | # ifdef LATIN_CAPITAL_LETTER_SHARP_S_UTF8 | |
3327 | # define CAP_SHARP_S LATIN_CAPITAL_LETTER_SHARP_S_UTF8 | |
76f2ffcd KW |
3328 | |
3329 | const unsigned int cap_sharp_s_len = sizeof(CAP_SHARP_S) - 1; | |
76f2ffcd | 3330 | |
538e84ed KW |
3331 | /* Special case these two characters, as what normally gets |
3332 | * returned under locale doesn't work */ | |
76f2ffcd KW |
3333 | if (UTF8SKIP(p) == cap_sharp_s_len |
3334 | && memEQ((char *) p, CAP_SHARP_S, cap_sharp_s_len)) | |
1ca267a5 | 3335 | { |
ab0b796c KW |
3336 | /* diag_listed_as: Can't do %s("%s") on non-UTF-8 locale; resolved to "%s". */ |
3337 | Perl_ck_warner(aTHX_ packWARN(WARN_LOCALE), | |
3338 | "Can't do fc(\"\\x{1E9E}\") on non-UTF-8 locale; " | |
3339 | "resolved to \"\\x{17F}\\x{17F}\"."); | |
1ca267a5 KW |
3340 | goto return_long_s; |
3341 | } | |
0766489e KW |
3342 | else |
3343 | #endif | |
3344 | if (UTF8SKIP(p) == long_s_t_len | |
76f2ffcd | 3345 | && memEQ((char *) p, LONG_S_T, long_s_t_len)) |
9fc2026f | 3346 | { |
ab0b796c KW |
3347 | /* diag_listed_as: Can't do %s("%s") on non-UTF-8 locale; resolved to "%s". */ |
3348 | Perl_ck_warner(aTHX_ packWARN(WARN_LOCALE), | |
3349 | "Can't do fc(\"\\x{FB05}\") on non-UTF-8 locale; " | |
3350 | "resolved to \"\\x{FB06}\"."); | |
9fc2026f KW |
3351 | goto return_ligature_st; |
3352 | } | |
74894415 KW |
3353 | |
3354 | #if UNICODE_MAJOR_VERSION == 3 \ | |
3355 | && UNICODE_DOT_VERSION == 0 \ | |
3356 | && UNICODE_DOT_DOT_VERSION == 1 | |
3357 | # define DOTTED_I LATIN_CAPITAL_LETTER_I_WITH_DOT_ABOVE_UTF8 | |
3358 | ||
3359 | /* And special case this on this Unicode version only, for the same | |
3360 | * reaons the other two are special cased. They would cross the | |
3361 | * 255/256 boundary which is forbidden under /l, and so the code | |
3362 | * wouldn't catch that they are equivalent (which they are only in | |
3363 | * this release) */ | |
3364 | else if (UTF8SKIP(p) == sizeof(DOTTED_I) - 1 | |
3365 | && memEQ((char *) p, DOTTED_I, sizeof(DOTTED_I) - 1)) | |
3366 | { | |
3367 | /* diag_listed_as: Can't do %s("%s") on non-UTF-8 locale; resolved to "%s". */ | |
3368 | Perl_ck_warner(aTHX_ packWARN(WARN_LOCALE), | |
3369 | "Can't do fc(\"\\x{0130}\") on non-UTF-8 locale; " | |
3370 | "resolved to \"\\x{0131}\"."); | |
3371 | goto return_dotless_i; | |
3372 | } | |
3373 | #endif | |
3374 | ||
357aadde | 3375 | return check_locale_boundary_crossing(p, result, ustrp, lenp); |
051a06d4 | 3376 | } |
a0270393 KW |
3377 | else if (! (flags & FOLD_FLAGS_NOMIX_ASCII)) { |
3378 | return result; | |
3379 | } | |
3380 | else { | |
4a4088c4 | 3381 | /* This is called when changing the case of a UTF-8-encoded |
9fc2026f KW |
3382 | * character above the ASCII range, and the result should not |
3383 | * contain an ASCII character. */ | |
a0270393 KW |
3384 | |
3385 | UV original; /* To store the first code point of <p> */ | |
3386 | ||
3387 | /* Look at every character in the result; if any cross the | |
3388 | * boundary, the whole thing is disallowed */ | |
3389 | U8* s = ustrp; | |
3390 | U8* e = ustrp + *lenp; | |
3391 | while (s < e) { | |
3392 | if (isASCII(*s)) { | |
3393 | /* Crossed, have to return the original */ | |
3394 | original = valid_utf8_to_uvchr(p, lenp); | |
1ca267a5 | 3395 | |
9fc2026f | 3396 | /* But in these instances, there is an alternative we can |
1ca267a5 | 3397 | * return that is valid */ |
0766489e KW |
3398 | if (original == LATIN_SMALL_LETTER_SHARP_S |
3399 | #ifdef LATIN_CAPITAL_LETTER_SHARP_S /* not defined in early Unicode releases */ | |
3400 | || original == LATIN_CAPITAL_LETTER_SHARP_S | |
3401 | #endif | |
3402 | ) { | |
1ca267a5 KW |
3403 | goto return_long_s; |
3404 | } | |
9fc2026f KW |
3405 | else if (original == LATIN_SMALL_LIGATURE_LONG_S_T) { |
3406 | goto return_ligature_st; | |
3407 | } | |
74894415 KW |
3408 | #if UNICODE_MAJOR_VERSION == 3 \ |
3409 | && UNICODE_DOT_VERSION == 0 \ | |
3410 | && UNICODE_DOT_DOT_VERSION == 1 | |
3411 | ||
3412 | else if (original == LATIN_CAPITAL_LETTER_I_WITH_DOT_ABOVE) { | |
3413 | goto return_dotless_i; | |
3414 | } | |
3415 | #endif | |
a0270393 KW |
3416 | Copy(p, ustrp, *lenp, char); |
3417 | return original; | |
3418 | } | |
3419 | s += UTF8SKIP(s); | |
3420 | } | |
051a06d4 | 3421 | |
a0270393 KW |
3422 | /* Here, no characters crossed, result is ok as-is */ |
3423 | return result; | |
3424 | } | |
051a06d4 KW |
3425 | } |
3426 | ||
4a4088c4 | 3427 | /* Here, used locale rules. Convert back to UTF-8 */ |
051a06d4 KW |
3428 | if (UTF8_IS_INVARIANT(result)) { |
3429 | *ustrp = (U8) result; | |
3430 | *lenp = 1; | |
3431 | } | |
3432 | else { | |
62cb07ea KW |
3433 | *ustrp = UTF8_EIGHT_BIT_HI((U8) result); |
3434 | *(ustrp + 1) = UTF8_EIGHT_BIT_LO((U8) result); | |
051a06d4 KW |
3435 | *lenp = 2; |
3436 | } | |
3437 | ||
051a06d4 | 3438 | return result; |
1ca267a5 KW |
3439 | |
3440 | return_long_s: | |
3441 | /* Certain folds to 'ss' are prohibited by the options, but they do allow | |
3442 | * folds to a string of two of these characters. By returning this | |
3443 | * instead, then, e.g., | |
3444 | * fc("\x{1E9E}") eq fc("\x{17F}\x{17F}") | |
3445 | * works. */ | |
3446 | ||
3447 | *lenp = 2 * sizeof(LATIN_SMALL_LETTER_LONG_S_UTF8) - 2; | |
3448 | Copy(LATIN_SMALL_LETTER_LONG_S_UTF8 LATIN_SMALL_LETTER_LONG_S_UTF8, | |
3449 | ustrp, *lenp, U8); | |
3450 | return LATIN_SMALL_LETTER_LONG_S; | |
9fc2026f KW |
3451 | |
3452 | return_ligature_st: | |
3453 | /* Two folds to 'st' are prohibited by the options; instead we pick one and | |
3454 | * have the other one fold to it */ | |
3455 | ||
3456 | *lenp = sizeof(LATIN_SMALL_LIGATURE_ST_UTF8) - 1; | |
3457 | Copy(LATIN_SMALL_LIGATURE_ST_UTF8, ustrp, *lenp, U8); | |
3458 | return LATIN_SMALL_LIGATURE_ST; | |
74894415 KW |
3459 | |
3460 | #if UNICODE_MAJOR_VERSION == 3 \ | |
3461 | && UNICODE_DOT_VERSION == 0 \ | |
3462 | && UNICODE_DOT_DOT_VERSION == 1 | |
3463 | ||
3464 | return_dotless_i: | |
3465 | *lenp = sizeof(LATIN_SMALL_LETTER_DOTLESS_I_UTF8) - 1; | |
3466 | Copy(LATIN_SMALL_LETTER_DOTLESS_I_UTF8, ustrp, *lenp, U8); | |
3467 | return LATIN_SMALL_LETTER_DOTLESS_I; | |
3468 | ||
3469 | #endif | |
3470 | ||
a0ed51b3 LW |
3471 | } |
3472 | ||
711a919c | 3473 | /* Note: |
f90a9a02 | 3474 | * Returns a "swash" which is a hash described in utf8.c:Perl_swash_fetch(). |
711a919c TS |
3475 | * C<pkg> is a pointer to a package name for SWASHNEW, should be "utf8". |
3476 | * For other parameters, see utf8::SWASHNEW in lib/utf8_heavy.pl. | |
3477 | */ | |
c4a5db0c | 3478 | |
a0ed51b3 | 3479 | SV* |
7fc63493 | 3480 | Perl_swash_init(pTHX_ const char* pkg, const char* name, SV *listsv, I32 minbits, I32 none) |
a0ed51b3 | 3481 | { |
c4a5db0c KW |
3482 | PERL_ARGS_ASSERT_SWASH_INIT; |
3483 | ||
3484 | /* Returns a copy of a swash initiated by the called function. This is the | |
3485 | * public interface, and returning a copy prevents others from doing | |
3486 | * mischief on the original */ | |
3487 | ||
5d3d13d1 | 3488 | return newSVsv(_core_swash_init(pkg, name, listsv, minbits, none, NULL, NULL)); |
c4a5db0c KW |
3489 | } |
3490 | ||
3491 | SV* | |
5d3d13d1 | 3492 | Perl__core_swash_init(pTHX_ const char* pkg, const char* name, SV *listsv, I32 minbits, I32 none, SV* invlist, U8* const flags_p) |
c4a5db0c | 3493 | { |
2c1f00b9 YO |
3494 | |
3495 | /*NOTE NOTE NOTE - If you want to use "return" in this routine you MUST | |
3496 | * use the following define */ | |
3497 | ||
3498 | #define CORE_SWASH_INIT_RETURN(x) \ | |
3499 | PL_curpm= old_PL_curpm; \ | |
3500 | return x | |
3501 | ||
c4a5db0c | 3502 | /* Initialize and return a swash, creating it if necessary. It does this |
87367d5f KW |
3503 | * by calling utf8_heavy.pl in the general case. The returned value may be |
3504 | * the swash's inversion list instead if the input parameters allow it. | |
3505 | * Which is returned should be immaterial to callers, as the only | |
923b6d4e KW |
3506 | * operations permitted on a swash, swash_fetch(), _get_swash_invlist(), |
3507 | * and swash_to_invlist() handle both these transparently. | |
c4a5db0c KW |
3508 | * |
3509 | * This interface should only be used by functions that won't destroy or | |
3510 | * adversely change the swash, as doing so affects all other uses of the | |
3511 | * swash in the program; the general public should use 'Perl_swash_init' | |
3512 | * instead. | |
3513 | * | |
3514 | * pkg is the name of the package that <name> should be in. | |
3515 | * name is the name of the swash to find. Typically it is a Unicode | |
3516 | * property name, including user-defined ones | |
3517 | * listsv is a string to initialize the swash with. It must be of the form | |
3518 | * documented as the subroutine return value in | |
3519 | * L<perlunicode/User-Defined Character Properties> | |
3520 | * minbits is the number of bits required to represent each data element. | |
3521 | * It is '1' for binary properties. | |
3522 | * none I (khw) do not understand this one, but it is used only in tr///. | |
9a53f6cf | 3523 | * invlist is an inversion list to initialize the swash with (or NULL) |
83199d38 KW |
3524 | * flags_p if non-NULL is the address of various input and output flag bits |
3525 | * to the routine, as follows: ('I' means is input to the routine; | |
3526 | * 'O' means output from the routine. Only flags marked O are | |
3527 | * meaningful on return.) | |
3528 | * _CORE_SWASH_INIT_USER_DEFINED_PROPERTY indicates if the swash | |
3529 | * came from a user-defined property. (I O) | |
5d3d13d1 KW |
3530 | * _CORE_SWASH_INIT_RETURN_IF_UNDEF indicates that instead of croaking |
3531 | * when the swash cannot be located, to simply return NULL. (I) | |
87367d5f KW |
3532 | * _CORE_SWASH_INIT_ACCEPT_INVLIST indicates that the caller will accept a |
3533 | * return of an inversion list instead of a swash hash if this routine | |
3534 | * thinks that would result in faster execution of swash_fetch() later | |
3535 | * on. (I) | |
9a53f6cf KW |
3536 | * |
3537 | * Thus there are three possible inputs to find the swash: <name>, | |
3538 | * <listsv>, and <invlist>. At least one must be specified. The result | |
3539 | * will be the union of the specified ones, although <listsv>'s various | |
aabbdbda KW |
3540 | * actions can intersect, etc. what <name> gives. To avoid going out to |
3541 | * disk at all, <invlist> should specify completely what the swash should | |
3542 | * have, and <listsv> should be &PL_sv_undef and <name> should be "". | |
9a53f6cf KW |
3543 | * |
3544 | * <invlist> is only valid for binary properties */ | |
c4a5db0c | 3545 | |
2c1f00b9 YO |
3546 | PMOP *old_PL_curpm= PL_curpm; /* save away the old PL_curpm */ |
3547 | ||
c4a5db0c | 3548 | SV* retval = &PL_sv_undef; |
83199d38 | 3549 | HV* swash_hv = NULL; |
87367d5f KW |
3550 | const int invlist_swash_boundary = |
3551 | (flags_p && *flags_p & _CORE_SWASH_INIT_ACCEPT_INVLIST) | |
3552 | ? 512 /* Based on some benchmarking, but not extensive, see commit | |
3553 | message */ | |
3554 | : -1; /* Never return just an inversion list */ | |
9a53f6cf KW |
3555 | |
3556 | assert(listsv != &PL_sv_undef || strNE(name, "") || invlist); | |
3557 | assert(! invlist || minbits == 1); | |
3558 | ||
2c1f00b9 YO |
3559 | PL_curpm= NULL; /* reset PL_curpm so that we dont get confused between the regex |
3560 | that triggered the swash init and the swash init perl logic itself. | |
3561 | See perl #122747 */ | |
3562 | ||
9a53f6cf KW |
3563 | /* If data was passed in to go out to utf8_heavy to find the swash of, do |
3564 | * so */ | |
3565 | if (listsv != &PL_sv_undef || strNE(name, "")) { | |
69794297 KW |
3566 | dSP; |
3567 | const size_t pkg_len = strlen(pkg); | |
3568 | const size_t name_len = strlen(name); | |
3569 | HV * const stash = gv_stashpvn(pkg, pkg_len, 0); | |
3570 | SV* errsv_save; | |
3571 | GV *method; | |
3572 | ||
3573 | PERL_ARGS_ASSERT__CORE_SWASH_INIT; | |
3574 | ||
3575 | PUSHSTACKi(PERLSI_MAGIC); | |
ce3b816e | 3576 | ENTER; |
69794297 | 3577 | SAVEHINTS(); |
2782061f | 3578 | save_re_context(); |
650f067c JL |
3579 | /* We might get here via a subroutine signature which uses a utf8 |
3580 | * parameter name, at which point PL_subname will have been set | |
3581 | * but not yet used. */ | |
3582 | save_item(PL_subname); | |
69794297 KW |
3583 | if (PL_parser && PL_parser->error_count) |
3584 | SAVEI8(PL_parser->error_count), PL_parser->error_count = 0; | |
3585 | method = gv_fetchmeth(stash, "SWASHNEW", 8, -1); | |
4a4088c4 | 3586 | if (!method) { /* demand load UTF-8 */ |
69794297 | 3587 | ENTER; |
db2c6cb3 FC |
3588 | if ((errsv_save = GvSV(PL_errgv))) SAVEFREESV(errsv_save); |
3589 | GvSV(PL_errgv) = NULL; | |
1a419e6b | 3590 | #ifndef NO_TAINT_SUPPORT |
69794297 KW |
3591 | /* It is assumed that callers of this routine are not passing in |
3592 | * any user derived data. */ | |
2782061f DM |
3593 | /* Need to do this after save_re_context() as it will set |
3594 | * PL_tainted to 1 while saving $1 etc (see the code after getrx: | |
3595 | * in Perl_magic_get). Even line to create errsv_save can turn on | |
3596 | * PL_tainted. */ | |
284167a5 S |
3597 | SAVEBOOL(TAINT_get); |
3598 | TAINT_NOT; | |
3599 | #endif | |
69794297 KW |
3600 | Perl_load_module(aTHX_ PERL_LOADMOD_NOIMPORT, newSVpvn(pkg,pkg_len), |
3601 | NULL); | |
eed484f9 | 3602 | { |
db2c6cb3 FC |
3603 | /* Not ERRSV, as there is no need to vivify a scalar we are |
3604 | about to discard. */ | |
3605 | SV * const errsv = GvSV(PL_errgv); | |
3606 | if (!SvTRUE(errsv)) { | |
3607 | GvSV(PL_errgv) = SvREFCNT_inc_simple(errsv_save); | |
3608 | SvREFCNT_dec(errsv); | |
3609 | } | |
eed484f9 | 3610 | } |
69794297 KW |
3611 | LEAVE; |
3612 | } | |
3613 | SPAGAIN; | |
3614 | PUSHMARK(SP); | |
3615 | EXTEND(SP,5); | |
3616 | mPUSHp(pkg, pkg_len); | |
3617 | mPUSHp(name, name_len); | |
3618 | PUSHs(listsv); | |
3619 | mPUSHi(minbits); | |
3620 | mPUSHi(none); | |
3621 | PUTBACK; | |
db2c6cb3 FC |
3622 | if ((errsv_save = GvSV(PL_errgv))) SAVEFREESV(errsv_save); |
3623 | GvSV(PL_errgv) = NULL; | |
69794297 KW |
3624 | /* If we already have a pointer to the method, no need to use |
3625 | * call_method() to repeat the lookup. */ | |
c41800a8 KW |
3626 | if (method |
3627 | ? call_sv(MUTABLE_SV(method), G_SCALAR) | |
69794297 KW |
3628 | : call_sv(newSVpvs_flags("SWASHNEW", SVs_TEMP), G_SCALAR | G_METHOD)) |
3629 | { | |
3630 | retval = *PL_stack_sp--; | |
3631 | SvREFCNT_inc(retval); | |
3632 | } | |
eed484f9 | 3633 | { |
db2c6cb3 FC |
3634 | /* Not ERRSV. See above. */ |
3635 | SV * const errsv = GvSV(PL_errgv); | |
3636 | if (!SvTRUE(errsv)) { | |
3637 | GvSV(PL_errgv) = SvREFCNT_inc_simple(errsv_save); | |
3638 | SvREFCNT_dec(errsv); | |
3639 | } | |
eed484f9 | 3640 | } |
ce3b816e | 3641 | LEAVE; |
69794297 KW |
3642 | POPSTACK; |
3643 | if (IN_PERL_COMPILETIME) { | |
3644 | CopHINTS_set(PL_curcop, PL_hints); | |
3645 | } | |
3646 | if (!SvROK(retval) || SvTYPE(SvRV(retval)) != SVt_PVHV) { | |
d95e4a00 | 3647 | if (SvPOK(retval)) { |
69794297 KW |
3648 | |
3649 | /* If caller wants to handle missing properties, let them */ | |
5d3d13d1 | 3650 | if (flags_p && *flags_p & _CORE_SWASH_INIT_RETURN_IF_UNDEF) { |
2c1f00b9 | 3651 | CORE_SWASH_INIT_RETURN(NULL); |
69794297 KW |
3652 | } |
3653 | Perl_croak(aTHX_ | |
147e3846 | 3654 | "Can't find Unicode property definition \"%" SVf "\"", |
69794297 | 3655 | SVfARG(retval)); |
a25b5927 | 3656 | NOT_REACHED; /* NOTREACHED */ |
d95e4a00 | 3657 | } |
69794297 | 3658 | } |
9a53f6cf | 3659 | } /* End of calling the module to find the swash */ |
36eb48b4 | 3660 | |
83199d38 KW |
3661 | /* If this operation fetched a swash, and we will need it later, get it */ |
3662 | if (retval != &PL_sv_undef | |
3663 | && (minbits == 1 || (flags_p | |
3664 | && ! (*flags_p | |
3665 | & _CORE_SWASH_INIT_USER_DEFINED_PROPERTY)))) | |
3666 | { | |
3667 | swash_hv = MUTABLE_HV(SvRV(retval)); | |
3668 | ||
3669 | /* If we don't already know that there is a user-defined component to | |
3670 | * this swash, and the user has indicated they wish to know if there is | |
3671 | * one (by passing <flags_p>), find out */ | |
3672 | if (flags_p && ! (*flags_p & _CORE_SWASH_INIT_USER_DEFINED_PROPERTY)) { | |
3673 | SV** user_defined = hv_fetchs(swash_hv, "USER_DEFINED", FALSE); | |
3674 | if (user_defined && SvUV(*user_defined)) { | |
3675 | *flags_p |= _CORE_SWASH_INIT_USER_DEFINED_PROPERTY; | |
3676 | } | |
3677 | } | |
3678 | } | |
3679 | ||
36eb48b4 KW |
3680 | /* Make sure there is an inversion list for binary properties */ |
3681 | if (minbits == 1) { | |
3682 | SV** swash_invlistsvp = NULL; | |
3683 | SV* swash_invlist = NULL; | |
9a53f6cf | 3684 | bool invlist_in_swash_is_valid = FALSE; |
02c85471 FC |
3685 | bool swash_invlist_unclaimed = FALSE; /* whether swash_invlist has |
3686 | an unclaimed reference count */ | |
36eb48b4 | 3687 | |
9a53f6cf | 3688 | /* If this operation fetched a swash, get its already existing |
83199d38 | 3689 | * inversion list, or create one for it */ |
36eb48b4 | 3690 | |
83199d38 | 3691 | if (swash_hv) { |
5c9f4bd2 | 3692 | swash_invlistsvp = hv_fetchs(swash_hv, "V", FALSE); |
9a53f6cf KW |
3693 | if (swash_invlistsvp) { |
3694 | swash_invlist = *swash_invlistsvp; | |
3695 | invlist_in_swash_is_valid = TRUE; | |
3696 | } | |
3697 | else { | |
36eb48b4 | 3698 | swash_invlist = _swash_to_invlist(retval); |
02c85471 | 3699 | swash_invlist_unclaimed = TRUE; |
9a53f6cf KW |
3700 | } |
3701 | } | |
3702 | ||
3703 | /* If an inversion list was passed in, have to include it */ | |
3704 | if (invlist) { | |
3705 | ||
3706 | /* Any fetched swash will by now have an inversion list in it; | |
3707 | * otherwise <swash_invlist> will be NULL, indicating that we | |
3708 | * didn't fetch a swash */ | |
3709 | if (swash_invlist) { | |
3710 | ||
3711 | /* Add the passed-in inversion list, which invalidates the one | |
3712 | * already stored in the swash */ | |
3713 | invlist_in_swash_is_valid = FALSE; | |
eee4c920 | 3714 | SvREADONLY_off(swash_invlist); /* Turned on again below */ |
9a53f6cf KW |
3715 | _invlist_union(invlist, swash_invlist, &swash_invlist); |
3716 | } | |
3717 | else { | |
3718 | ||
87367d5f KW |
3719 | /* Here, there is no swash already. Set up a minimal one, if |
3720 | * we are going to return a swash */ | |
3721 | if ((int) _invlist_len(invlist) > invlist_swash_boundary) { | |
971d486f | 3722 | swash_hv = newHV(); |
4aca0fe6 | 3723 | retval = newRV_noinc(MUTABLE_SV(swash_hv)); |
87367d5f | 3724 | } |
9a53f6cf KW |
3725 | swash_invlist = invlist; |
3726 | } | |
9a53f6cf KW |
3727 | } |
3728 | ||
3729 | /* Here, we have computed the union of all the passed-in data. It may | |
3730 | * be that there was an inversion list in the swash which didn't get | |
538e84ed | 3731 | * touched; otherwise save the computed one */ |
87367d5f KW |
3732 | if (! invlist_in_swash_is_valid |
3733 | && (int) _invlist_len(swash_invlist) > invlist_swash_boundary) | |
3734 | { | |
5c9f4bd2 | 3735 | if (! hv_stores(MUTABLE_HV(SvRV(retval)), "V", swash_invlist)) |
69794297 KW |
3736 | { |
3737 | Perl_croak(aTHX_ "panic: hv_store() unexpectedly failed"); | |
3738 | } | |
cc34d8c5 FC |
3739 | /* We just stole a reference count. */ |
3740 | if (swash_invlist_unclaimed) swash_invlist_unclaimed = FALSE; | |
3741 | else SvREFCNT_inc_simple_void_NN(swash_invlist); | |
9a53f6cf | 3742 | } |
87367d5f | 3743 | |
ee3222e3 | 3744 | /* The result is immutable. Forbid attempts to change it. */ |
dbfdbd26 KW |
3745 | SvREADONLY_on(swash_invlist); |
3746 | ||
c41800a8 | 3747 | /* Use the inversion list stand-alone if small enough */ |
87367d5f KW |
3748 | if ((int) _invlist_len(swash_invlist) <= invlist_swash_boundary) { |
3749 | SvREFCNT_dec(retval); | |
02c85471 FC |
3750 | if (!swash_invlist_unclaimed) |
3751 | SvREFCNT_inc_simple_void_NN(swash_invlist); | |
3752 | retval = newRV_noinc(swash_invlist); | |
87367d5f | 3753 | } |
36eb48b4 KW |
3754 | } |
3755 | ||
2c1f00b9 YO |
3756 | CORE_SWASH_INIT_RETURN(retval); |
3757 | #undef CORE_SWASH_INIT_RETURN | |
a0ed51b3 LW |
3758 | } |
3759 | ||
035d37be JH |
3760 | |
3761 | /* This API is wrong for special case conversions since we may need to | |
3762 | * return several Unicode characters for a single Unicode character | |
3763 | * (see lib/unicore/SpecCase.txt) The SWASHGET in lib/utf8_heavy.pl is | |
3764 | * the lower-level routine, and it is similarly broken for returning | |
38684baa | 3765 | * multiple values. --jhi |
b9992569 | 3766 | * For those, you should use S__to_utf8_case() instead */ |
b0e3252e | 3767 | /* Now SWASHGET is recasted into S_swatch_get in this file. */ |
680c470c TS |
3768 | |
3769 | /* Note: | |
3770 | * Returns the value of property/mapping C<swash> for the first character | |
3771 | * of the string C<ptr>. If C<do_utf8> is true, the string C<ptr> is | |
4a4088c4 | 3772 | * assumed to be in well-formed UTF-8. If C<do_utf8> is false, the string C<ptr> |
3d0f8846 | 3773 | * is assumed to be in native 8-bit encoding. Caches the swatch in C<swash>. |
af2af982 KW |
3774 | * |
3775 | * A "swash" is a hash which contains initially the keys/values set up by | |
3776 | * SWASHNEW. The purpose is to be able to completely represent a Unicode | |
3777 | * property for all possible code points. Things are stored in a compact form | |
3778 | * (see utf8_heavy.pl) so that calculation is required to find the actual | |
3779 | * property value for a given code point. As code points are looked up, new | |
3780 | * key/value pairs are added to the hash, so that the calculation doesn't have | |
3781 | * to ever be re-done. Further, each calculation is done, not just for the | |
3782 | * desired one, but for a whole block of code points adjacent to that one. | |
3783 | * For binary properties on ASCII machines, the block is usually for 64 code | |
3784 | * points, starting with a code point evenly divisible by 64. Thus if the | |
3785 | * property value for code point 257 is requested, the code goes out and | |
3786 | * calculates the property values for all 64 code points between 256 and 319, | |
3787 | * and stores these as a single 64-bit long bit vector, called a "swatch", | |
3788 | * under the key for code point 256. The key is the UTF-8 encoding for code | |
3789 | * point 256, minus the final byte. Thus, if the length of the UTF-8 encoding | |
3790 | * for a code point is 13 bytes, the key will be 12 bytes long. If the value | |
3791 | * for code point 258 is then requested, this code realizes that it would be | |
3792 | * stored under the key for 256, and would find that value and extract the | |
3793 | * relevant bit, offset from 256. | |
3794 | * | |
3795 | * Non-binary properties are stored in as many bits as necessary to represent | |
3796 | * their values (32 currently, though the code is more general than that), not | |
fc273927 | 3797 | * as single bits, but the principle is the same: the value for each key is a |
af2af982 KW |
3798 | * vector that encompasses the property values for all code points whose UTF-8 |
3799 | * representations are represented by the key. That is, for all code points | |
3800 | * whose UTF-8 representations are length N bytes, and the key is the first N-1 | |
3801 | * bytes of that. | |
680c470c | 3802 | */ |
a0ed51b3 | 3803 | UV |
680c470c | 3804 | Perl_swash_fetch(pTHX_ SV *swash, const U8 *ptr, bool do_utf8) |
a0ed51b3 | 3805 | { |
ef8f7699 | 3806 | HV *const hv = MUTABLE_HV(SvRV(swash)); |
3568d838 JH |
3807 | U32 klen; |
3808 | U32 off; | |
9b56a019 | 3809 | STRLEN slen = 0; |
7d85a32c | 3810 | STRLEN needents; |
cfd0369c | 3811 | const U8 *tmps = NULL; |
979f2922 | 3812 | SV *swatch; |
08fb1ac5 | 3813 | const U8 c = *ptr; |
3568d838 | 3814 | |
7918f24d NC |
3815 | PERL_ARGS_ASSERT_SWASH_FETCH; |
3816 | ||
87367d5f KW |
3817 | /* If it really isn't a hash, it isn't really swash; must be an inversion |
3818 | * list */ | |
3819 | if (SvTYPE(hv) != SVt_PVHV) { | |
3820 | return _invlist_contains_cp((SV*)hv, | |
3821 | (do_utf8) | |
3822 | ? valid_utf8_to_uvchr(ptr, NULL) | |
3823 | : c); | |
3824 | } | |
3825 | ||
08fb1ac5 KW |
3826 | /* We store the values in a "swatch" which is a vec() value in a swash |
3827 | * hash. Code points 0-255 are a single vec() stored with key length | |
3828 | * (klen) 0. All other code points have a UTF-8 representation | |
3829 | * 0xAA..0xYY,0xZZ. A vec() is constructed containing all of them which | |
3830 | * share 0xAA..0xYY, which is the key in the hash to that vec. So the key | |
3831 | * length for them is the length of the encoded char - 1. ptr[klen] is the | |
3832 | * final byte in the sequence representing the character */ | |
3833 | if (!do_utf8 || UTF8_IS_INVARIANT(c)) { | |
3834 | klen = 0; | |
3835 | needents = 256; | |
3836 | off = c; | |
3568d838 | 3837 | } |
08fb1ac5 KW |
3838 | else if (UTF8_IS_DOWNGRADEABLE_START(c)) { |
3839 | klen = 0; | |
3840 | needents = 256; | |
a62b247b | 3841 | off = EIGHT_BIT_UTF8_TO_NATIVE(c, *(ptr + 1)); |
979f2922 TS |
3842 | } |
3843 | else { | |
08fb1ac5 KW |
3844 | klen = UTF8SKIP(ptr) - 1; |
3845 | ||
3846 | /* Each vec() stores 2**UTF_ACCUMULATION_SHIFT values. The offset into | |
3847 | * the vec is the final byte in the sequence. (In EBCDIC this is | |
3848 | * converted to I8 to get consecutive values.) To help you visualize | |
3849 | * all this: | |
3850 | * Straight 1047 After final byte | |
3851 | * UTF-8 UTF-EBCDIC I8 transform | |
3852 | * U+0400: \xD0\x80 \xB8\x41\x41 \xB8\x41\xA0 | |
3853 | * U+0401: \xD0\x81 \xB8\x41\x42 \xB8\x41\xA1 | |
3854 | * ... | |
3855 | * U+0409: \xD0\x89 \xB8\x41\x4A \xB8\x41\xA9 | |
3856 | * U+040A: \xD0\x8A \xB8\x41\x51 \xB8\x41\xAA | |
3857 | * ... | |
3858 | * U+0412: \xD0\x92 \xB8\x41\x59 \xB8\x41\xB2 | |
3859 | * U+0413: \xD0\x93 \xB8\x41\x62 \xB8\x41\xB3 | |
3860 | * ... | |
3861 | * U+041B: \xD0\x9B \xB8\x41\x6A \xB8\x41\xBB | |
3862 | * U+041C: \xD0\x9C \xB8\x41\x70 \xB8\x41\xBC | |
3863 | * ... | |
3864 | * U+041F: \xD0\x9F \xB8\x41\x73 \xB8\x41\xBF | |
3865 | * U+0420: \xD0\xA0 \xB8\x42\x41 \xB8\x42\x41 | |
3866 | * | |
3867 | * (There are no discontinuities in the elided (...) entries.) | |
3868 | * The UTF-8 key for these 33 code points is '\xD0' (which also is the | |
3869 | * key for the next 31, up through U+043F, whose UTF-8 final byte is | |
3870 | * \xBF). Thus in UTF-8, each key is for a vec() for 64 code points. | |
3871 | * The final UTF-8 byte, which ranges between \x80 and \xBF, is an | |
3872 | * index into the vec() swatch (after subtracting 0x80, which we | |
3873 | * actually do with an '&'). | |
3874 | * In UTF-EBCDIC, each key is for a 32 code point vec(). The first 32 | |
3875 | * code points above have key '\xB8\x41'. The final UTF-EBCDIC byte has | |
3876 | * dicontinuities which go away by transforming it into I8, and we | |
3877 | * effectively subtract 0xA0 to get the index. */ | |
979f2922 | 3878 | needents = (1 << UTF_ACCUMULATION_SHIFT); |
bc3632a8 | 3879 | off = NATIVE_UTF8_TO_I8(ptr[klen]) & UTF_CONTINUATION_MASK; |
979f2922 | 3880 | } |
7d85a32c | 3881 | |
a0ed51b3 | 3882 | /* |
4a4088c4 | 3883 | * This single-entry cache saves about 1/3 of the UTF-8 overhead in test |
a0ed51b3 LW |
3884 | * suite. (That is, only 7-8% overall over just a hash cache. Still, |
3885 | * it's nothing to sniff at.) Pity we usually come through at least | |
3886 | * two function calls to get here... | |
3887 | * | |
3888 | * NB: this code assumes that swatches are never modified, once generated! | |
3889 | */ | |
3890 | ||
3568d838 | 3891 | if (hv == PL_last_swash_hv && |
a0ed51b3 | 3892 | klen == PL_last_swash_klen && |
27da23d5 | 3893 | (!klen || memEQ((char *)ptr, (char *)PL_last_swash_key, klen)) ) |
a0ed51b3 LW |
3894 | { |
3895 | tmps = PL_last_swash_tmps; | |
3896 | slen = PL_last_swash_slen; | |
3897 | } | |
3898 | else { | |
3899 | /* Try our second-level swatch cache, kept in a hash. */ | |
e1ec3a88 | 3900 | SV** svp = hv_fetch(hv, (const char*)ptr, klen, FALSE); |
a0ed51b3 | 3901 | |
b0e3252e | 3902 | /* If not cached, generate it via swatch_get */ |
979f2922 | 3903 | if (!svp || !SvPOK(*svp) |
08fb1ac5 KW |
3904 | || !(tmps = (const U8*)SvPV_const(*svp, slen))) |
3905 | { | |
3906 | if (klen) { | |
3907 | const UV code_point = valid_utf8_to_uvchr(ptr, NULL); | |
3908 | swatch = swatch_get(swash, | |
3909 | code_point & ~((UV)needents - 1), | |
3910 | needents); | |
3911 | } | |
3912 | else { /* For the first 256 code points, the swatch has a key of | |
3913 | length 0 */ | |
3914 | swatch = swatch_get(swash, 0, needents); | |
3915 | } | |
979f2922 | 3916 | |
923e4eb5 | 3917 | if (IN_PERL_COMPILETIME) |
623e6609 | 3918 | CopHINTS_set(PL_curcop, PL_hints); |
a0ed51b3 | 3919 | |
979f2922 | 3920 | svp = hv_store(hv, (const char *)ptr, klen, swatch, 0); |
a0ed51b3 | 3921 | |
979f2922 TS |
3922 | if (!svp || !(tmps = (U8*)SvPV(*svp, slen)) |
3923 | || (slen << 3) < needents) | |
5637ef5b | 3924 | Perl_croak(aTHX_ "panic: swash_fetch got improper swatch, " |
147e3846 | 3925 | "svp=%p, tmps=%p, slen=%" UVuf ", needents=%" UVuf, |
5637ef5b | 3926 | svp, tmps, (UV)slen, (UV)needents); |
a0ed51b3 LW |
3927 | } |
3928 | ||
3929 | PL_last_swash_hv = hv; | |
16d8f38a | 3930 | assert(klen <= sizeof(PL_last_swash_key)); |
eac04b2e | 3931 | PL_last_swash_klen = (U8)klen; |
cfd0369c NC |
3932 | /* FIXME change interpvar.h? */ |
3933 | PL_last_swash_tmps = (U8 *) tmps; | |
a0ed51b3 LW |
3934 | PL_last_swash_slen = slen; |
3935 | if (klen) | |
3936 | Copy(ptr, PL_last_swash_key, klen, U8); | |
3937 | } | |
3938 | ||
9faf8d75 | 3939 | switch ((int)((slen << 3) / needents)) { |
a0ed51b3 | 3940 | case 1: |
e7aca353 | 3941 | return ((UV) tmps[off >> 3] & (1 << (off & 7))) != 0; |
a0ed51b3 | 3942 | case 8: |
e7aca353 | 3943 | return ((UV) tmps[off]); |
a0ed51b3 LW |
3944 | case 16: |
3945 | off <<= 1; | |
e7aca353 JH |
3946 | return |
3947 | ((UV) tmps[off ] << 8) + | |
3948 | ((UV) tmps[off + 1]); | |
a0ed51b3 LW |
3949 | case 32: |
3950 | off <<= 2; | |
e7aca353 JH |
3951 | return |
3952 | ((UV) tmps[off ] << 24) + | |
3953 | ((UV) tmps[off + 1] << 16) + | |
3954 | ((UV) tmps[off + 2] << 8) + | |
3955 | ((UV) tmps[off + 3]); | |
a0ed51b3 | 3956 | } |
5637ef5b | 3957 | Perl_croak(aTHX_ "panic: swash_fetch got swatch of unexpected bit width, " |
147e3846 | 3958 | "slen=%" UVuf ", needents=%" UVuf, (UV)slen, (UV)needents); |
670f1322 | 3959 | NORETURN_FUNCTION_END; |
a0ed51b3 | 3960 | } |
2b9d42f0 | 3961 | |
319009ee KW |
3962 | /* Read a single line of the main body of the swash input text. These are of |
3963 | * the form: | |
3964 | * 0053 0056 0073 | |
3965 | * where each number is hex. The first two numbers form the minimum and | |
3966 | * maximum of a range, and the third is the value associated with the range. | |
3967 | * Not all swashes should have a third number | |
3968 | * | |
3969 | * On input: l points to the beginning of the line to be examined; it points | |
3970 | * to somewhere in the string of the whole input text, and is | |
3971 | * terminated by a \n or the null string terminator. | |
3972 | * lend points to the null terminator of that string | |
3973 | * wants_value is non-zero if the swash expects a third number | |
3974 | * typestr is the name of the swash's mapping, like 'ToLower' | |
3975 | * On output: *min, *max, and *val are set to the values read from the line. | |
3976 | * returns a pointer just beyond the line examined. If there was no | |
3977 | * valid min number on the line, returns lend+1 | |
3978 | */ | |
3979 | ||
3980 | STATIC U8* | |
3981 | S_swash_scan_list_line(pTHX_ U8* l, U8* const lend, UV* min, UV* max, UV* val, | |
3982 | const bool wants_value, const U8* const typestr) | |
3983 | { | |
3984 | const int typeto = typestr[0] == 'T' && typestr[1] == 'o'; | |
3985 | STRLEN numlen; /* Length of the number */ | |
02470786 KW |
3986 | I32 flags = PERL_SCAN_SILENT_ILLDIGIT |
3987 | | PERL_SCAN_DISALLOW_PREFIX | |
3988 | | PERL_SCAN_SILENT_NON_PORTABLE; | |
319009ee KW |
3989 | |
3990 | /* nl points to the next \n in the scan */ | |
3991 | U8* const nl = (U8*)memchr(l, '\n', lend - l); | |
3992 | ||
95543e92 KW |
3993 | PERL_ARGS_ASSERT_SWASH_SCAN_LIST_LINE; |
3994 | ||
319009ee KW |
3995 | /* Get the first number on the line: the range minimum */ |
3996 | numlen = lend - l; | |
3997 | *min = grok_hex((char *)l, &numlen, &flags, NULL); | |
c88850db | 3998 | *max = *min; /* So can never return without setting max */ |
319009ee KW |
3999 | if (numlen) /* If found a hex number, position past it */ |
4000 | l += numlen; | |
4001 | else if (nl) { /* Else, go handle next line, if any */ | |
4002 | return nl + 1; /* 1 is length of "\n" */ | |
4003 | } | |
4004 | else { /* Else, no next line */ | |
4005 | return lend + 1; /* to LIST's end at which \n is not found */ | |
4006 | } | |
4007 | ||
4008 | /* The max range value follows, separated by a BLANK */ | |
4009 | if (isBLANK(*l)) { | |
4010 | ++l; | |
02470786 KW |
4011 | flags = PERL_SCAN_SILENT_ILLDIGIT |
4012 | | PERL_SCAN_DISALLOW_PREFIX | |
4013 | | PERL_SCAN_SILENT_NON_PORTABLE; | |
319009ee KW |
4014 | numlen = lend - l; |
4015 | *max = grok_hex((char *)l, &numlen, &flags, NULL); | |
4016 | if (numlen) | |
4017 | l += numlen; | |
4018 | else /* If no value here, it is a single element range */ | |
4019 | *max = *min; | |
4020 | ||
4021 | /* Non-binary tables have a third entry: what the first element of the | |
24303724 | 4022 | * range maps to. The map for those currently read here is in hex */ |
319009ee KW |
4023 | if (wants_value) { |
4024 | if (isBLANK(*l)) { | |
4025 | ++l; | |
f2a7d0fc KW |
4026 | flags = PERL_SCAN_SILENT_ILLDIGIT |
4027 | | PERL_SCAN_DISALLOW_PREFIX | |
4028 | | PERL_SCAN_SILENT_NON_PORTABLE; | |
4029 | numlen = lend - l; | |
4030 | *val = grok_hex((char *)l, &numlen, &flags, NULL); | |
4031 | if (numlen) | |
4032 | l += numlen; | |
4033 | else | |
4034 | *val = 0; | |
319009ee KW |
4035 | } |
4036 | else { | |
4037 | *val = 0; | |
4038 | if (typeto) { | |
dcbac5bb | 4039 | /* diag_listed_as: To%s: illegal mapping '%s' */ |
319009ee KW |
4040 | Perl_croak(aTHX_ "%s: illegal mapping '%s'", |
4041 | typestr, l); | |
4042 | } | |
4043 | } | |
4044 | } | |
4045 | else | |
4046 | *val = 0; /* bits == 1, then any val should be ignored */ | |
4047 | } | |
4048 | else { /* Nothing following range min, should be single element with no | |
4049 | mapping expected */ | |
319009ee KW |
4050 | if (wants_value) { |
4051 | *val = 0; | |
4052 | if (typeto) { | |
dcbac5bb | 4053 | /* diag_listed_as: To%s: illegal mapping '%s' */ |
319009ee KW |
4054 | Perl_croak(aTHX_ "%s: illegal mapping '%s'", typestr, l); |
4055 | } | |
4056 | } | |
4057 | else | |
4058 | *val = 0; /* bits == 1, then val should be ignored */ | |
4059 | } | |
4060 | ||
4061 | /* Position to next line if any, or EOF */ | |
4062 | if (nl) | |
4063 | l = nl + 1; | |
4064 | else | |
4065 | l = lend; | |
4066 | ||
4067 | return l; | |
4068 | } | |
4069 | ||
979f2922 TS |
4070 | /* Note: |
4071 | * Returns a swatch (a bit vector string) for a code point sequence | |
4072 | * that starts from the value C<start> and comprises the number C<span>. | |
4073 | * A C<swash> must be an object created by SWASHNEW (see lib/utf8_heavy.pl). | |
4074 | * Should be used via swash_fetch, which will cache the swatch in C<swash>. | |
4075 | */ | |
4076 | STATIC SV* | |
b0e3252e | 4077 | S_swatch_get(pTHX_ SV* swash, UV start, UV span) |
979f2922 TS |
4078 | { |
4079 | SV *swatch; | |
77f9f126 | 4080 | U8 *l, *lend, *x, *xend, *s, *send; |
979f2922 | 4081 | STRLEN lcur, xcur, scur; |
ef8f7699 | 4082 | HV *const hv = MUTABLE_HV(SvRV(swash)); |
5c9f4bd2 | 4083 | SV** const invlistsvp = hv_fetchs(hv, "V", FALSE); |
36eb48b4 | 4084 | |
88d45d28 KW |
4085 | SV** listsvp = NULL; /* The string containing the main body of the table */ |
4086 | SV** extssvp = NULL; | |
4087 | SV** invert_it_svp = NULL; | |
4088 | U8* typestr = NULL; | |
786861f5 KW |
4089 | STRLEN bits; |
4090 | STRLEN octets; /* if bits == 1, then octets == 0 */ | |
4091 | UV none; | |
4092 | UV end = start + span; | |
972dd592 | 4093 | |
36eb48b4 | 4094 | if (invlistsvp == NULL) { |
786861f5 KW |
4095 | SV** const bitssvp = hv_fetchs(hv, "BITS", FALSE); |
4096 | SV** const nonesvp = hv_fetchs(hv, "NONE", FALSE); | |
4097 | SV** const typesvp = hv_fetchs(hv, "TYPE", FALSE); | |
4098 | extssvp = hv_fetchs(hv, "EXTRAS", FALSE); | |
4099 | listsvp = hv_fetchs(hv, "LIST", FALSE); | |
4100 | invert_it_svp = hv_fetchs(hv, "INVERT_IT", FALSE); | |
4101 | ||
4102 | bits = SvUV(*bitssvp); | |
4103 | none = SvUV(*nonesvp); | |
4104 | typestr = (U8*)SvPV_nolen(*typesvp); | |
4105 | } | |
36eb48b4 KW |
4106 | else { |
4107 | bits = 1; | |
4108 | none = 0; | |
4109 | } | |
786861f5 | 4110 | octets = bits >> 3; /* if bits == 1, then octets == 0 */ |
979f2922 | 4111 | |
b0e3252e | 4112 | PERL_ARGS_ASSERT_SWATCH_GET; |
7918f24d | 4113 | |
979f2922 | 4114 | if (bits != 1 && bits != 8 && bits != 16 && bits != 32) { |
147e3846 | 4115 | Perl_croak(aTHX_ "panic: swatch_get doesn't expect bits %" UVuf, |
660a4616 | 4116 | (UV)bits); |
979f2922 TS |
4117 | } |
4118 | ||
84ea5ef6 KW |
4119 | /* If overflowed, use the max possible */ |
4120 | if (end < start) { | |
4121 | end = UV_MAX; | |
4122 | span = end - start; | |
4123 | } | |
4124 | ||
979f2922 | 4125 | /* create and initialize $swatch */ |
979f2922 | 4126 | scur = octets ? (span * octets) : (span + 7) / 8; |
e524fe40 NC |
4127 | swatch = newSV(scur); |
4128 | SvPOK_on(swatch); | |
979f2922 TS |
4129 | s = (U8*)SvPVX(swatch); |
4130 | if (octets && none) { | |
0bd48802 | 4131 | const U8* const e = s + scur; |
979f2922 TS |
4132 | while (s < e) { |
4133 | if (bits == 8) | |
4134 | *s++ = (U8)(none & 0xff); | |
4135 | else if (bits == 16) { | |
4136 | *s++ = (U8)((none >> 8) & 0xff); | |
4137 | *s++ = (U8)( none & 0xff); | |
4138 | } | |
4139 | else if (bits == 32) { | |
4140 | *s++ = (U8)((none >> 24) & 0xff); | |
4141 | *s++ = (U8)((none >> 16) & 0xff); | |
4142 | *s++ = (U8)((none >> 8) & 0xff); | |
4143 | *s++ = (U8)( none & 0xff); | |
4144 | } | |
4145 | } | |
4146 | *s = '\0'; | |
4147 | } | |
4148 | else { | |
4149 | (void)memzero((U8*)s, scur + 1); | |
4150 | } | |
4151 | SvCUR_set(swatch, scur); | |
4152 | s = (U8*)SvPVX(swatch); | |
4153 | ||
36eb48b4 KW |
4154 | if (invlistsvp) { /* If has an inversion list set up use that */ |
4155 | _invlist_populate_swatch(*invlistsvp, start, end, s); | |
4156 | return swatch; | |
4157 | } | |
4158 | ||
4159 | /* read $swash->{LIST} */ | |
979f2922 TS |
4160 | l = (U8*)SvPV(*listsvp, lcur); |
4161 | lend = l + lcur; | |
4162 | while (l < lend) { | |
8ed25d53 | 4163 | UV min, max, val, upper; |
95543e92 KW |
4164 | l = swash_scan_list_line(l, lend, &min, &max, &val, |
4165 | cBOOL(octets), typestr); | |
319009ee | 4166 | if (l > lend) { |
979f2922 TS |
4167 | break; |
4168 | } | |
4169 | ||
972dd592 | 4170 | /* If looking for something beyond this range, go try the next one */ |
979f2922 TS |
4171 | if (max < start) |
4172 | continue; | |
4173 | ||
8ed25d53 KW |
4174 | /* <end> is generally 1 beyond where we want to set things, but at the |
4175 | * platform's infinity, where we can't go any higher, we want to | |
4176 | * include the code point at <end> */ | |
4177 | upper = (max < end) | |
4178 | ? max | |
4179 | : (max != UV_MAX || end != UV_MAX) | |
4180 | ? end - 1 | |
4181 | : end; | |
4182 | ||
979f2922 | 4183 | if (octets) { |
35da51f7 | 4184 | UV key; |
979f2922 TS |
4185 | if (min < start) { |
4186 | if (!none || val < none) { | |
4187 | val += start - min; | |
4188 | } | |
4189 | min = start; | |
4190 | } | |
8ed25d53 | 4191 | for (key = min; key <= upper; key++) { |
979f2922 | 4192 | STRLEN offset; |
979f2922 TS |
4193 | /* offset must be non-negative (start <= min <= key < end) */ |
4194 | offset = octets * (key - start); | |
4195 | if (bits == 8) | |
4196 | s[offset] = (U8)(val & 0xff); | |
4197 | else if (bits == 16) { | |
4198 | s[offset ] = (U8)((val >> 8) & 0xff); | |
4199 | s[offset + 1] = (U8)( val & 0xff); | |
4200 | } | |
4201 | else if (bits == 32) { | |
4202 | s[offset ] = (U8)((val >> 24) & 0xff); | |
4203 | s[offset + 1] = (U8)((val >> 16) & 0xff); | |
4204 | s[offset + 2] = (U8)((val >> 8) & 0xff); | |
4205 | s[offset + 3] = (U8)( val & 0xff); | |
4206 | } | |
4207 | ||
4208 | if (!none || val < none) | |
4209 | ++val; | |
4210 | } | |
4211 | } | |
711a919c | 4212 | else { /* bits == 1, then val should be ignored */ |
35da51f7 | 4213 | UV key; |
979f2922 TS |
4214 | if (min < start) |
4215 | min = start; | |
6cb05c12 | 4216 | |
8ed25d53 | 4217 | for (key = min; key <= upper; key++) { |
0bd48802 | 4218 | const STRLEN offset = (STRLEN)(key - start); |
979f2922 TS |
4219 | s[offset >> 3] |= 1 << (offset & 7); |
4220 | } | |
4221 | } | |
4222 | } /* while */ | |
979f2922 | 4223 | |
9479a769 | 4224 | /* Invert if the data says it should be. Assumes that bits == 1 */ |
77f9f126 | 4225 | if (invert_it_svp && SvUV(*invert_it_svp)) { |
0bda3001 KW |
4226 | |
4227 | /* Unicode properties should come with all bits above PERL_UNICODE_MAX | |
4228 | * be 0, and their inversion should also be 0, as we don't succeed any | |
4229 | * Unicode property matches for non-Unicode code points */ | |
4230 | if (start <= PERL_UNICODE_MAX) { | |
4231 | ||
4232 | /* The code below assumes that we never cross the | |
4233 | * Unicode/above-Unicode boundary in a range, as otherwise we would | |
4234 | * have to figure out where to stop flipping the bits. Since this | |
4235 | * boundary is divisible by a large power of 2, and swatches comes | |
4236 | * in small powers of 2, this should be a valid assumption */ | |
4237 | assert(start + span - 1 <= PERL_UNICODE_MAX); | |
4238 | ||
507a8485 KW |
4239 | send = s + scur; |
4240 | while (s < send) { | |
4241 | *s = ~(*s); | |
4242 | s++; | |
4243 | } | |
0bda3001 | 4244 | } |
77f9f126 KW |
4245 | } |
4246 | ||
d73c39c5 KW |
4247 | /* read $swash->{EXTRAS} |
4248 | * This code also copied to swash_to_invlist() below */ | |
979f2922 TS |
4249 | x = (U8*)SvPV(*extssvp, xcur); |
4250 | xend = x + xcur; | |
4251 | while (x < xend) { | |
4252 | STRLEN namelen; | |
4253 | U8 *namestr; | |
4254 | SV** othersvp; | |
4255 | HV* otherhv; | |
4256 | STRLEN otherbits; | |
4257 | SV **otherbitssvp, *other; | |
711a919c | 4258 | U8 *s, *o, *nl; |
979f2922 TS |
4259 | STRLEN slen, olen; |
4260 | ||
35da51f7 | 4261 | const U8 opc = *x++; |
979f2922 TS |
4262 | if (opc == '\n') |
4263 | continue; | |
4264 | ||
4265 | nl = (U8*)memchr(x, '\n', xend - x); | |
4266 | ||
4267 | if (opc != '-' && opc != '+' && opc != '!' && opc != '&') { | |
4268 | if (nl) { | |
4269 | x = nl + 1; /* 1 is length of "\n" */ | |
4270 | continue; | |
4271 | } | |
4272 | else { | |
4273 | x = xend; /* to EXTRAS' end at which \n is not found */ | |
4274 | break; | |
4275 | } | |
4276 | } | |
4277 | ||
4278 | namestr = x; | |
4279 | if (nl) { | |
4280 | namelen = nl - namestr; | |
4281 | x = nl + 1; | |
4282 | } | |
4283 | else { | |
4284 | namelen = xend - namestr; | |
4285 | x = xend; | |
4286 | } | |
4287 | ||
4288 | othersvp = hv_fetch(hv, (char *)namestr, namelen, FALSE); | |
ef8f7699 | 4289 | otherhv = MUTABLE_HV(SvRV(*othersvp)); |
017a3ce5 | 4290 | otherbitssvp = hv_fetchs(otherhv, "BITS", FALSE); |
979f2922 TS |
4291 | otherbits = (STRLEN)SvUV(*otherbitssvp); |
4292 | if (bits < otherbits) | |
5637ef5b | 4293 | Perl_croak(aTHX_ "panic: swatch_get found swatch size mismatch, " |
147e3846 | 4294 | "bits=%" UVuf ", otherbits=%" UVuf, (UV)bits, (UV)otherbits); |
979f2922 TS |
4295 | |
4296 | /* The "other" swatch must be destroyed after. */ | |
b0e3252e | 4297 | other = swatch_get(*othersvp, start, span); |
979f2922 TS |
4298 | o = (U8*)SvPV(other, olen); |
4299 | ||
4300 | if (!olen) | |
b0e3252e | 4301 | Perl_croak(aTHX_ "panic: swatch_get got improper swatch"); |
979f2922 TS |
4302 | |
4303 | s = (U8*)SvPV(swatch, slen); | |
4304 | if (bits == 1 && otherbits == 1) { | |
4305 | if (slen != olen) | |
5637ef5b | 4306 | Perl_croak(aTHX_ "panic: swatch_get found swatch length " |
147e3846 | 4307 | "mismatch, slen=%" UVuf ", olen=%" UVuf, |
5637ef5b | 4308 | (UV)slen, (UV)olen); |
979f2922 TS |
4309 | |
4310 | switch (opc) { | |
4311 | case '+': | |
4312 | while (slen--) | |
4313 | *s++ |= *o++; | |
4314 | break; | |
4315 | case '!': | |
4316 | while (slen--) | |
4317 | *s++ |= ~*o++; | |
4318 | break; | |
4319 | case '-': | |
4320 | while (slen--) | |
4321 | *s++ &= ~*o++; | |
4322 | break; | |
4323 | case '&': | |
4324 | while (slen--) | |
4325 | *s++ &= *o++; | |
4326 | break; | |
4327 | default: | |
4328 | break; | |
4329 | } | |
4330 | } | |
711a919c | 4331 | else { |
979f2922 TS |
4332 | STRLEN otheroctets = otherbits >> 3; |
4333 | STRLEN offset = 0; | |
35da51f7 | 4334 | U8* const send = s + slen; |
979f2922 TS |
4335 | |
4336 | while (s < send) { | |
4337 | UV otherval = 0; | |
4338 | ||
4339 | if (otherbits == 1) { | |
4340 | otherval = (o[offset >> 3] >> (offset & 7)) & 1; | |
4341 | ++offset; | |
4342 | } | |
4343 | else { | |
4344 | STRLEN vlen = otheroctets; | |
4345 | otherval = *o++; | |
4346 | while (--vlen) { | |
4347 | otherval <<= 8; | |
4348 | otherval |= *o++; | |
4349 | } | |
4350 | } | |
4351 | ||
711a919c | 4352 | if (opc == '+' && otherval) |
6f207bd3 | 4353 | NOOP; /* replace with otherval */ |
979f2922 TS |
4354 | else if (opc == '!' && !otherval) |
4355 | otherval = 1; | |
4356 | else if (opc == '-' && otherval) | |
4357 | otherval = 0; | |
4358 | else if (opc == '&' && !otherval) | |
4359 | otherval = 0; | |
4360 | else { | |
711a919c | 4361 | s += octets; /* no replacement */ |
979f2922 TS |
4362 | continue; |
4363 | } | |
4364 | ||
4365 | if (bits == 8) | |
4366 | *s++ = (U8)( otherval & 0xff); | |
4367 | else if (bits == 16) { | |
4368 | *s++ = (U8)((otherval >> 8) & 0xff); | |
4369 | *s++ = (U8)( otherval & 0xff); | |
4370 | } | |
4371 | else if (bits == 32) { | |
4372 | *s++ = (U8)((otherval >> 24) & 0xff); | |
4373 | *s++ = (U8)((otherval >> 16) & 0xff); | |
4374 | *s++ = (U8)((otherval >> 8) & 0xff); | |
4375 | *s++ = (U8)( otherval & 0xff); | |
4376 | } | |
4377 | } | |
4378 | } | |
4379 | sv_free(other); /* through with it! */ | |
4380 | } /* while */ | |
4381 | return swatch; | |
4382 | } | |
4383 | ||
064c021d | 4384 | HV* |
4c2e1131 | 4385 | Perl__swash_inversion_hash(pTHX_ SV* const swash) |
064c021d KW |
4386 | { |
4387 | ||
79a2a0e8 | 4388 | /* Subject to change or removal. For use only in regcomp.c and regexec.c |
5662e334 KW |
4389 | * Can't be used on a property that is subject to user override, as it |
4390 | * relies on the value of SPECIALS in the swash which would be set by | |
4391 | * utf8_heavy.pl to the hash in the non-overriden file, and hence is not set | |
4392 | * for overridden properties | |
064c021d KW |
4393 | * |
4394 | * Returns a hash which is the inversion and closure of a swash mapping. | |
4395 | * For example, consider the input lines: | |
4396 | * 004B 006B | |
4397 | * 004C 006C | |
4398 | * 212A 006B | |
4399 | * | |
4a4088c4 | 4400 | * The returned hash would have two keys, the UTF-8 for 006B and the UTF-8 for |
064c021d | 4401 | * 006C. The value for each key is an array. For 006C, the array would |
4a4088c4 KW |
4402 | * have two elements, the UTF-8 for itself, and for 004C. For 006B, there |
4403 | * would be three elements in its array, the UTF-8 for 006B, 004B and 212A. | |
064c021d | 4404 | * |
538e84ed KW |
4405 | * Note that there are no elements in the hash for 004B, 004C, 212A. The |
4406 | * keys are only code points that are folded-to, so it isn't a full closure. | |
4407 | * | |
064c021d KW |
4408 | * Essentially, for any code point, it gives all the code points that map to |
4409 | * it, or the list of 'froms' for that point. | |
4410 | * | |
5662e334 KW |
4411 | * Currently it ignores any additions or deletions from other swashes, |
4412 | * looking at just the main body of the swash, and if there are SPECIALS | |
4413 | * in the swash, at that hash | |
4414 | * | |
4415 | * The specials hash can be extra code points, and most likely consists of | |
4416 | * maps from single code points to multiple ones (each expressed as a string | |
4a4088c4 | 4417 | * of UTF-8 characters). This function currently returns only 1-1 mappings. |
5662e334 KW |
4418 | * However consider this possible input in the specials hash: |
4419 | * "\xEF\xAC\x85" => "\x{0073}\x{0074}", # U+FB05 => 0073 0074 | |
4420 | * "\xEF\xAC\x86" => "\x{0073}\x{0074}", # U+FB06 => 0073 0074 | |
4421 | * | |
4422 | * Both FB05 and FB06 map to the same multi-char sequence, which we don't | |
4423 | * currently handle. But it also means that FB05 and FB06 are equivalent in | |
4424 | * a 1-1 mapping which we should handle, and this relationship may not be in | |
4425 | * the main table. Therefore this function examines all the multi-char | |
74894415 KW |
4426 | * sequences and adds the 1-1 mappings that come out of that. |
4427 | * | |
4428 | * XXX This function was originally intended to be multipurpose, but its | |
4429 | * only use is quite likely to remain for constructing the inversion of | |
4430 | * the CaseFolding (//i) property. If it were more general purpose for | |
4431 | * regex patterns, it would have to do the FB05/FB06 game for simple folds, | |
4432 | * because certain folds are prohibited under /iaa and /il. As an example, | |
4433 | * in Unicode 3.0.1 both U+0130 and U+0131 fold to 'i', and hence are both | |
4434 | * equivalent under /i. But under /iaa and /il, the folds to 'i' are | |
4435 | * prohibited, so we would not figure out that they fold to each other. | |
4436 | * Code could be written to automatically figure this out, similar to the | |
4437 | * code that does this for multi-character folds, but this is the only case | |
4438 | * where something like this is ever likely to happen, as all the single | |
7ee537e6 | 4439 | * char folds to the 0-255 range are now quite settled. Instead there is a |
74894415 KW |
4440 | * little special code that is compiled only for this Unicode version. This |
4441 | * is smaller and didn't require much coding time to do. But this makes | |
4442 | * this routine strongly tied to being used just for CaseFolding. If ever | |
4443 | * it should be generalized, this would have to be fixed */ | |
064c021d KW |
4444 | |
4445 | U8 *l, *lend; | |
4446 | STRLEN lcur; | |
4447 | HV *const hv = MUTABLE_HV(SvRV(swash)); | |
4448 | ||
923b6d4e KW |
4449 | /* The string containing the main body of the table. This will have its |
4450 | * assertion fail if the swash has been converted to its inversion list */ | |
064c021d KW |
4451 | SV** const listsvp = hv_fetchs(hv, "LIST", FALSE); |
4452 | ||
4453 | SV** const typesvp = hv_fetchs(hv, "TYPE", FALSE); | |
4454 | SV** const bitssvp = hv_fetchs(hv, "BITS", FALSE); | |
4455 | SV** const nonesvp = hv_fetchs(hv, "NONE", FALSE); | |
4456 | /*SV** const extssvp = hv_fetchs(hv, "EXTRAS", FALSE);*/ | |
4457 | const U8* const typestr = (U8*)SvPV_nolen(*typesvp); | |
4458 | const STRLEN bits = SvUV(*bitssvp); | |
4459 | const STRLEN octets = bits >> 3; /* if bits == 1, then octets == 0 */ | |
4460 | const UV none = SvUV(*nonesvp); | |
5662e334 | 4461 | SV **specials_p = hv_fetchs(hv, "SPECIALS", 0); |
064c021d KW |
4462 | |
4463 | HV* ret = newHV(); | |
4464 | ||
4465 | PERL_ARGS_ASSERT__SWASH_INVERSION_HASH; | |
4466 | ||
4467 | /* Must have at least 8 bits to get the mappings */ | |
4468 | if (bits != 8 && bits != 16 && bits != 32) { | |
147e3846 | 4469 | Perl_croak(aTHX_ "panic: swash_inversion_hash doesn't expect bits %" UVuf, |
064c021d KW |
4470 | (UV)bits); |
4471 | } | |
4472 | ||
5662e334 KW |
4473 | if (specials_p) { /* It might be "special" (sometimes, but not always, a |
4474 | mapping to more than one character */ | |
4475 | ||
4476 | /* Construct an inverse mapping hash for the specials */ | |
4477 | HV * const specials_hv = MUTABLE_HV(SvRV(*specials_p)); | |
4478 | HV * specials_inverse = newHV(); | |
4479 | char *char_from; /* the lhs of the map */ | |
4480 | I32 from_len; /* its byte length */ | |
4481 | char *char_to; /* the rhs of the map */ | |
4482 | I32 to_len; /* its byte length */ | |
4483 | SV *sv_to; /* and in a sv */ | |
4484 | AV* from_list; /* list of things that map to each 'to' */ | |
4485 | ||
4486 | hv_iterinit(specials_hv); | |
4487 | ||
4a4088c4 KW |
4488 | /* The keys are the characters (in UTF-8) that map to the corresponding |
4489 | * UTF-8 string value. Iterate through the list creating the inverse | |
5662e334 KW |
4490 | * list. */ |
4491 | while ((sv_to = hv_iternextsv(specials_hv, &char_from, &from_len))) { | |
4492 | SV** listp; | |
4493 | if (! SvPOK(sv_to)) { | |
5637ef5b NC |
4494 | Perl_croak(aTHX_ "panic: value returned from hv_iternextsv() " |
4495 | "unexpectedly is not a string, flags=%lu", | |
4496 | (unsigned long)SvFLAGS(sv_to)); | |
5662e334 | 4497 | } |
147e3846 | 4498 | /*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 |
4499 | |
4500 | /* Each key in the inverse list is a mapped-to value, and the key's | |
4a4088c4 | 4501 | * hash value is a list of the strings (each in UTF-8) that map to |
5662e334 KW |
4502 | * it. Those strings are all one character long */ |
4503 | if ((listp = hv_fetch(specials_inverse, | |
4504 | SvPVX(sv_to), | |
4505 | SvCUR(sv_to), 0))) | |
4506 | { | |
4507 | from_list = (AV*) *listp; | |
4508 | } | |
4509 | else { /* No entry yet for it: create one */ | |
4510 | from_list = newAV(); | |
4511 | if (! hv_store(specials_inverse, | |
4512 | SvPVX(sv_to), | |
4513 | SvCUR(sv_to), | |
4514 | (SV*) from_list, 0)) | |
4515 | { | |
4516 | Perl_croak(aTHX_ "panic: hv_store() unexpectedly failed"); | |
4517 | } | |
4518 | } | |
4519 | ||
4520 | /* Here have the list associated with this 'to' (perhaps newly | |
4521 | * created and empty). Just add to it. Note that we ASSUME that | |
4522 | * the input is guaranteed to not have duplications, so we don't | |
4523 | * check for that. Duplications just slow down execution time. */ | |
4524 | av_push(from_list, newSVpvn_utf8(char_from, from_len, TRUE)); | |
4525 | } | |
4526 | ||
4527 | /* Here, 'specials_inverse' contains the inverse mapping. Go through | |
4528 | * it looking for cases like the FB05/FB06 examples above. There would | |
4529 | * be an entry in the hash like | |
4530 | * 'st' => [ FB05, FB06 ] | |
4531 | * In this example we will create two lists that get stored in the | |
4532 | * returned hash, 'ret': | |
4533 | * FB05 => [ FB05, FB06 ] | |
4534 | * FB06 => [ FB05, FB06 ] | |
4535 | * | |
4536 | * Note that there is nothing to do if the array only has one element. | |
4537 | * (In the normal 1-1 case handled below, we don't have to worry about | |
4538 | * two lists, as everything gets tied to the single list that is | |
4539 | * generated for the single character 'to'. But here, we are omitting | |
4540 | * that list, ('st' in the example), so must have multiple lists.) */ | |
4541 | while ((from_list = (AV *) hv_iternextsv(specials_inverse, | |
4542 | &char_to, &to_len))) | |
4543 | { | |
7e13d567 | 4544 | if (av_tindex_nomg(from_list) > 0) { |
c70927a6 | 4545 | SSize_t i; |
5662e334 KW |
4546 | |
4547 | /* We iterate over all combinations of i,j to place each code | |
4548 | * point on each list */ | |
7e13d567 | 4549 | for (i = 0; i <= av_tindex_nomg(from_list); i++) { |
c70927a6 | 4550 | SSize_t j; |
5662e334 KW |
4551 | AV* i_list = newAV(); |
4552 | SV** entryp = av_fetch(from_list, i, FALSE); | |
4553 | if (entryp == NULL) { | |
4554 | Perl_croak(aTHX_ "panic: av_fetch() unexpectedly failed"); | |
4555 | } | |
4556 | if (hv_fetch(ret, SvPVX(*entryp), SvCUR(*entryp), FALSE)) { | |
4557 | Perl_croak(aTHX_ "panic: unexpected entry for %s", SvPVX(*entryp)); | |
4558 | } | |
4559 | if (! hv_store(ret, SvPVX(*entryp), SvCUR(*entryp), | |
4560 | (SV*) i_list, FALSE)) | |
4561 | { | |
4562 | Perl_croak(aTHX_ "panic: hv_store() unexpectedly failed"); | |
4563 | } | |
4564 | ||
538e84ed | 4565 | /* For DEBUG_U: UV u = valid_utf8_to_uvchr((U8*) SvPVX(*entryp), 0);*/ |
7e13d567 | 4566 | for (j = 0; j <= av_tindex_nomg(from_list); j++) { |
5662e334 KW |
4567 | entryp = av_fetch(from_list, j, FALSE); |
4568 | if (entryp == NULL) { | |
4569 | Perl_croak(aTHX_ "panic: av_fetch() unexpectedly failed"); | |
4570 | } | |
4571 | ||
4572 | /* When i==j this adds itself to the list */ | |
4b88fb76 KW |
4573 | av_push(i_list, newSVuv(utf8_to_uvchr_buf( |
4574 | (U8*) SvPVX(*entryp), | |
4575 | (U8*) SvPVX(*entryp) + SvCUR(*entryp), | |
4576 | 0))); | |
147e3846 | 4577 | /*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 |
4578 | } |
4579 | } | |
4580 | } | |
4581 | } | |
4582 | SvREFCNT_dec(specials_inverse); /* done with it */ | |
4583 | } /* End of specials */ | |
4584 | ||
064c021d | 4585 | /* read $swash->{LIST} */ |
74894415 KW |
4586 | |
4587 | #if UNICODE_MAJOR_VERSION == 3 \ | |
4588 | && UNICODE_DOT_VERSION == 0 \ | |
4589 | && UNICODE_DOT_DOT_VERSION == 1 | |
4590 | ||
4591 | /* For this version only U+130 and U+131 are equivalent under qr//i. Add a | |
4592 | * rule so that things work under /iaa and /il */ | |
4593 | ||
4594 | SV * mod_listsv = sv_mortalcopy(*listsvp); | |
4595 | sv_catpv(mod_listsv, "130\t130\t131\n"); | |
4596 | l = (U8*)SvPV(mod_listsv, lcur); | |
4597 | ||
4598 | #else | |
4599 | ||
064c021d | 4600 | l = (U8*)SvPV(*listsvp, lcur); |
74894415 KW |
4601 | |
4602 | #endif | |
4603 | ||
064c021d KW |
4604 | lend = l + lcur; |
4605 | ||
4606 | /* Go through each input line */ | |
4607 | while (l < lend) { | |
4608 | UV min, max, val; | |
4609 | UV inverse; | |
95543e92 KW |
4610 | l = swash_scan_list_line(l, lend, &min, &max, &val, |
4611 | cBOOL(octets), typestr); | |
064c021d KW |
4612 | if (l > lend) { |
4613 | break; | |
4614 | } | |
4615 | ||
4616 | /* Each element in the range is to be inverted */ | |
4617 | for (inverse = min; inverse <= max; inverse++) { | |
4618 | AV* list; | |
064c021d KW |
4619 | SV** listp; |
4620 | IV i; | |
4621 | bool found_key = FALSE; | |
5662e334 | 4622 | bool found_inverse = FALSE; |
064c021d KW |
4623 | |
4624 | /* The key is the inverse mapping */ | |
4625 | char key[UTF8_MAXBYTES+1]; | |
c80e42f3 | 4626 | char* key_end = (char *) uvchr_to_utf8((U8*) key, val); |
064c021d KW |
4627 | STRLEN key_len = key_end - key; |
4628 | ||
064c021d KW |
4629 | /* Get the list for the map */ |
4630 | if ((listp = hv_fetch(ret, key, key_len, FALSE))) { | |
4631 | list = (AV*) *listp; | |
4632 | } | |
4633 | else { /* No entry yet for it: create one */ | |
4634 | list = newAV(); | |
4635 | if (! hv_store(ret, key, key_len, (SV*) list, FALSE)) { | |
4636 | Perl_croak(aTHX_ "panic: hv_store() unexpectedly failed"); | |
4637 | } | |
4638 | } | |
4639 | ||
5662e334 KW |
4640 | /* Look through list to see if this inverse mapping already is |
4641 | * listed, or if there is a mapping to itself already */ | |
7e13d567 | 4642 | for (i = 0; i <= av_tindex_nomg(list); i++) { |
064c021d KW |
4643 | SV** entryp = av_fetch(list, i, FALSE); |
4644 | SV* entry; | |
414db8a1 | 4645 | UV uv; |
064c021d KW |
4646 | if (entryp == NULL) { |
4647 | Perl_croak(aTHX_ "panic: av_fetch() unexpectedly failed"); | |
4648 | } | |
4649 | entry = *entryp; | |
414db8a1 | 4650 | uv = SvUV(entry); |
147e3846 | 4651 | /*DEBUG_U(PerlIO_printf(Perl_debug_log, "list for %" UVXf " contains %" UVXf "\n", val, uv));*/ |
414db8a1 | 4652 | if (uv == val) { |
064c021d | 4653 | found_key = TRUE; |
5662e334 | 4654 | } |
414db8a1 | 4655 | if (uv == inverse) { |
5662e334 KW |
4656 | found_inverse = TRUE; |
4657 | } | |
4658 | ||
4659 | /* No need to continue searching if found everything we are | |
4660 | * looking for */ | |
4661 | if (found_key && found_inverse) { | |
064c021d KW |
4662 | break; |
4663 | } | |
4664 | } | |
56ca34ca KW |
4665 | |
4666 | /* Make sure there is a mapping to itself on the list */ | |
064c021d | 4667 | if (! found_key) { |
d397ff6a | 4668 | av_push(list, newSVuv(val)); |
147e3846 | 4669 | /*DEBUG_U(PerlIO_printf(Perl_debug_log, "%s: %d: Adding %" UVXf " to list for %" UVXf "\n", __FILE__, __LINE__, val, val));*/ |
064c021d KW |
4670 | } |
4671 | ||
4672 | ||
4673 | /* Simply add the value to the list */ | |
5662e334 KW |
4674 | if (! found_inverse) { |
4675 | av_push(list, newSVuv(inverse)); | |
147e3846 | 4676 | /*DEBUG_U(PerlIO_printf(Perl_debug_log, "%s: %d: Adding %" UVXf " to list for %" UVXf "\n", __FILE__, __LINE__, inverse, val));*/ |
5662e334 | 4677 | } |
064c021d | 4678 | |
b0e3252e | 4679 | /* swatch_get() increments the value of val for each element in the |
064c021d KW |
4680 | * range. That makes more compact tables possible. You can |
4681 | * express the capitalization, for example, of all consecutive | |
4682 | * letters with a single line: 0061\t007A\t0041 This maps 0061 to | |
4683 | * 0041, 0062 to 0042, etc. I (khw) have never understood 'none', | |
bd3f2f94 | 4684 | * and it's not documented; it appears to be used only in |
b0e3252e | 4685 | * implementing tr//; I copied the semantics from swatch_get(), just |
bd3f2f94 | 4686 | * in case */ |
064c021d KW |
4687 | if (!none || val < none) { |
4688 | ++val; | |
4689 | } | |
4690 | } | |
4691 | } | |
4692 | ||
4693 | return ret; | |
4694 | } | |
4695 | ||
a25abddc | 4696 | SV* |
d764b54e KW |
4697 | Perl__swash_to_invlist(pTHX_ SV* const swash) |
4698 | { | |
4699 | ||
ed92f1b3 KW |
4700 | /* Subject to change or removal. For use only in one place in regcomp.c. |
4701 | * Ownership is given to one reference count in the returned SV* */ | |
d764b54e KW |
4702 | |
4703 | U8 *l, *lend; | |
4704 | char *loc; | |
4705 | STRLEN lcur; | |
4706 | HV *const hv = MUTABLE_HV(SvRV(swash)); | |
4707 | UV elements = 0; /* Number of elements in the inversion list */ | |
b443038a | 4708 | U8 empty[] = ""; |
923b6d4e KW |
4709 | SV** listsvp; |
4710 | SV** typesvp; | |
4711 | SV** bitssvp; | |
4712 | SV** extssvp; | |
4713 | SV** invert_it_svp; | |
d764b54e | 4714 | |
923b6d4e KW |
4715 | U8* typestr; |
4716 | STRLEN bits; | |
4717 | STRLEN octets; /* if bits == 1, then octets == 0 */ | |
d73c39c5 KW |
4718 | U8 *x, *xend; |
4719 | STRLEN xcur; | |
d764b54e | 4720 | |
a25abddc | 4721 | SV* invlist; |
d764b54e | 4722 | |
b81740c0 KW |
4723 | PERL_ARGS_ASSERT__SWASH_TO_INVLIST; |
4724 | ||
923b6d4e KW |
4725 | /* If not a hash, it must be the swash's inversion list instead */ |
4726 | if (SvTYPE(hv) != SVt_PVHV) { | |
ed92f1b3 | 4727 | return SvREFCNT_inc_simple_NN((SV*) hv); |
923b6d4e KW |
4728 | } |
4729 | ||
4730 | /* The string containing the main body of the table */ | |
4731 | listsvp = hv_fetchs(hv, "LIST", FALSE); | |
4732 | typesvp = hv_fetchs(hv, "TYPE", FALSE); | |
4733 | bitssvp = hv_fetchs(hv, "BITS", FALSE); | |
4734 | extssvp = hv_fetchs(hv, "EXTRAS", FALSE); | |
4735 | invert_it_svp = hv_fetchs(hv, "INVERT_IT", FALSE); | |
4736 | ||
4737 | typestr = (U8*)SvPV_nolen(*typesvp); | |
4738 | bits = SvUV(*bitssvp); | |
4739 | octets = bits >> 3; /* if bits == 1, then octets == 0 */ | |
4740 | ||
d764b54e | 4741 | /* read $swash->{LIST} */ |
b443038a KW |
4742 | if (SvPOK(*listsvp)) { |
4743 | l = (U8*)SvPV(*listsvp, lcur); | |
4744 | } | |
4745 | else { | |
4746 | /* LIST legitimately doesn't contain a string during compilation phases | |
4747 | * of Perl itself, before the Unicode tables are generated. In this | |
4748 | * case, just fake things up by creating an empty list */ | |
4749 | l = empty; | |
4750 | lcur = 0; | |
4751 | } | |
d764b54e KW |
4752 | loc = (char *) l; |
4753 | lend = l + lcur; | |
4754 | ||
31aa6e0b | 4755 | if (*l == 'V') { /* Inversion list format */ |
99944878 | 4756 | const char *after_atou = (char *) lend; |
31aa6e0b KW |
4757 | UV element0; |
4758 | UV* other_elements_ptr; | |
4759 | ||
4760 | /* The first number is a count of the rest */ | |
4761 | l++; | |
22ff3130 HS |
4762 | if (!grok_atoUV((const char *)l, &elements, &after_atou)) { |
4763 | Perl_croak(aTHX_ "panic: Expecting a valid count of elements at start of inversion list"); | |
4764 | } | |
eb092534 KW |
4765 | if (elements == 0) { |
4766 | invlist = _new_invlist(0); | |
4767 | } | |
4768 | else { | |
99944878 | 4769 | l = (U8 *) after_atou; |
1f9f7d4c KW |
4770 | |
4771 | /* Get the 0th element, which is needed to setup the inversion list */ | |
99944878 | 4772 | while (isSPACE(*l)) l++; |
22ff3130 HS |
4773 | if (!grok_atoUV((const char *)l, &element0, &after_atou)) { |
4774 | Perl_croak(aTHX_ "panic: Expecting a valid 0th element for inversion list"); | |
4775 | } | |
99944878 | 4776 | l = (U8 *) after_atou; |
1f9f7d4c KW |
4777 | invlist = _setup_canned_invlist(elements, element0, &other_elements_ptr); |
4778 | elements--; | |
4779 | ||
4780 | /* Then just populate the rest of the input */ | |
4781 | while (elements-- > 0) { | |
4782 | if (l > lend) { | |
147e3846 | 4783 | Perl_croak(aTHX_ "panic: Expecting %" UVuf " more elements than available", elements); |
1f9f7d4c | 4784 | } |
99944878 | 4785 | while (isSPACE(*l)) l++; |
22ff3130 HS |
4786 | if (!grok_atoUV((const char *)l, other_elements_ptr++, &after_atou)) { |
4787 | Perl_croak(aTHX_ "panic: Expecting a valid element in inversion list"); | |
4788 | } | |
99944878 | 4789 | l = (U8 *) after_atou; |
1f9f7d4c | 4790 | } |
eb092534 | 4791 | } |
31aa6e0b KW |
4792 | } |
4793 | else { | |
4794 | ||
1784d2f9 KW |
4795 | /* Scan the input to count the number of lines to preallocate array |
4796 | * size based on worst possible case, which is each line in the input | |
4797 | * creates 2 elements in the inversion list: 1) the beginning of a | |
4798 | * range in the list; 2) the beginning of a range not in the list. */ | |
4799 | while ((loc = (strchr(loc, '\n'))) != NULL) { | |
4800 | elements += 2; | |
4801 | loc++; | |
4802 | } | |
d764b54e | 4803 | |
1784d2f9 KW |
4804 | /* If the ending is somehow corrupt and isn't a new line, add another |
4805 | * element for the final range that isn't in the inversion list */ | |
4806 | if (! (*lend == '\n' | |
4807 | || (*lend == '\0' && (lcur == 0 || *(lend - 1) == '\n')))) | |
4808 | { | |
4809 | elements++; | |
4810 | } | |
d764b54e | 4811 | |
1784d2f9 | 4812 | invlist = _new_invlist(elements); |
d764b54e | 4813 | |
1784d2f9 KW |
4814 | /* Now go through the input again, adding each range to the list */ |
4815 | while (l < lend) { | |
4816 | UV start, end; | |
4817 | UV val; /* Not used by this function */ | |
d764b54e | 4818 | |
95543e92 KW |
4819 | l = swash_scan_list_line(l, lend, &start, &end, &val, |
4820 | cBOOL(octets), typestr); | |
d764b54e | 4821 | |
1784d2f9 KW |
4822 | if (l > lend) { |
4823 | break; | |
4824 | } | |
4825 | ||
4826 | invlist = _add_range_to_invlist(invlist, start, end); | |
4827 | } | |
31aa6e0b | 4828 | } |
d764b54e | 4829 | |
77f9f126 KW |
4830 | /* Invert if the data says it should be */ |
4831 | if (invert_it_svp && SvUV(*invert_it_svp)) { | |
25151030 | 4832 | _invlist_invert(invlist); |
77f9f126 KW |
4833 | } |
4834 | ||
b0e3252e | 4835 | /* This code is copied from swatch_get() |
d73c39c5 KW |
4836 | * read $swash->{EXTRAS} */ |
4837 | x = (U8*)SvPV(*extssvp, xcur); | |
4838 | xend = x + xcur; | |
4839 | while (x < xend) { | |
4840 | STRLEN namelen; | |
4841 | U8 *namestr; | |
4842 | SV** othersvp; | |
4843 | HV* otherhv; | |
4844 | STRLEN otherbits; | |
4845 | SV **otherbitssvp, *other; | |
4846 | U8 *nl; | |
4847 | ||
4848 | const U8 opc = *x++; | |
4849 | if (opc == '\n') | |
4850 | continue; | |
4851 | ||
4852 | nl = (U8*)memchr(x, '\n', xend - x); | |
4853 | ||
4854 | if (opc != '-' && opc != '+' && opc != '!' && opc != '&') { | |
4855 | if (nl) { | |
4856 | x = nl + 1; /* 1 is length of "\n" */ | |
4857 | continue; | |
4858 | } | |
4859 | else { | |
4860 | x = xend; /* to EXTRAS' end at which \n is not found */ | |
4861 | break; | |
4862 | } | |
4863 | } | |
4864 | ||
4865 | namestr = x; | |
4866 | if (nl) { | |
4867 | namelen = nl - namestr; | |
4868 | x = nl + 1; | |
4869 | } | |
4870 | else { | |
4871 | namelen = xend - namestr; | |
4872 | x = xend; | |
4873 | } | |
4874 | ||
4875 | othersvp = hv_fetch(hv, (char *)namestr, namelen, FALSE); | |
4876 | otherhv = MUTABLE_HV(SvRV(*othersvp)); | |
4877 | otherbitssvp = hv_fetchs(otherhv, "BITS", FALSE); | |
4878 | otherbits = (STRLEN)SvUV(*otherbitssvp); | |
4879 | ||
4880 | if (bits != otherbits || bits != 1) { | |
5637ef5b | 4881 | Perl_croak(aTHX_ "panic: _swash_to_invlist only operates on boolean " |
147e3846 | 4882 | "properties, bits=%" UVuf ", otherbits=%" UVuf, |
5637ef5b | 4883 | (UV)bits, (UV)otherbits); |
d73c39c5 KW |
4884 | } |
4885 | ||
4886 | /* The "other" swatch must be destroyed after. */ | |
4887 | other = _swash_to_invlist((SV *)*othersvp); | |
4888 | ||
b0e3252e | 4889 | /* End of code copied from swatch_get() */ |
d73c39c5 KW |
4890 | switch (opc) { |
4891 | case '+': | |
4892 | _invlist_union(invlist, other, &invlist); | |
4893 | break; | |
4894 | case '!': | |
6c46377d | 4895 | _invlist_union_maybe_complement_2nd(invlist, other, TRUE, &invlist); |
d73c39c5 KW |
4896 | break; |
4897 | case '-': | |
4898 | _invlist_subtract(invlist, other, &invlist); | |
4899 | break; | |
4900 | case '&': | |
4901 | _invlist_intersection(invlist, other, &invlist); | |
4902 | break; | |
4903 | default: | |
4904 | break; | |
4905 | } | |
4906 | sv_free(other); /* through with it! */ | |
4907 | } | |
4908 | ||
dbfdbd26 | 4909 | SvREADONLY_on(invlist); |
d764b54e KW |
4910 | return invlist; |
4911 | } | |
4912 | ||
3fdfee00 KW |
4913 | SV* |
4914 | Perl__get_swash_invlist(pTHX_ SV* const swash) | |
4915 | { | |
872dd7e0 | 4916 | SV** ptr; |
3fdfee00 KW |
4917 | |
4918 | PERL_ARGS_ASSERT__GET_SWASH_INVLIST; | |
4919 | ||
87367d5f | 4920 | if (! SvROK(swash)) { |
872dd7e0 KW |
4921 | return NULL; |
4922 | } | |
4923 | ||
87367d5f KW |
4924 | /* If it really isn't a hash, it isn't really swash; must be an inversion |
4925 | * list */ | |
4926 | if (SvTYPE(SvRV(swash)) != SVt_PVHV) { | |
4927 | return SvRV(swash); | |
4928 | } | |
872dd7e0 | 4929 | |
87367d5f | 4930 | ptr = hv_fetchs(MUTABLE_HV(SvRV(swash)), "V", FALSE); |
3fdfee00 KW |
4931 | if (! ptr) { |
4932 | return NULL; | |
4933 | } | |
4934 | ||
4935 | return *ptr; | |
4936 | } | |
4937 | ||
0876b9a0 | 4938 | bool |
5aaab254 | 4939 | Perl_check_utf8_print(pTHX_ const U8* s, const STRLEN len) |
0876b9a0 KW |
4940 | { |
4941 | /* May change: warns if surrogates, non-character code points, or | |
af2af982 KW |
4942 | * non-Unicode code points are in s which has length len bytes. Returns |
4943 | * TRUE if none found; FALSE otherwise. The only other validity check is | |
760c7c2f KW |
4944 | * to make sure that this won't exceed the string's length. |
4945 | * | |
4946 | * Code points above the platform's C<IV_MAX> will raise a deprecation | |
4947 | * warning, unless those are turned off. */ | |
0876b9a0 KW |
4948 | |
4949 | const U8* const e = s + len; | |
4950 | bool ok = TRUE; | |
4951 | ||
4952 | PERL_ARGS_ASSERT_CHECK_UTF8_PRINT; | |
4953 | ||
4954 | while (s < e) { | |
4955 | if (UTF8SKIP(s) > len) { | |
4956 | Perl_ck_warner_d(aTHX_ packWARN(WARN_UTF8), | |
4957 | "%s in %s", unees, PL_op ? OP_DESC(PL_op) : "print"); | |
4958 | return FALSE; | |
4959 | } | |
ac6f1fbe | 4960 | if (UNLIKELY(isUTF8_POSSIBLY_PROBLEMATIC(*s))) { |
f2bf18cc | 4961 | if (UNLIKELY(UTF8_IS_SUPER(s, e))) { |
760c7c2f KW |
4962 | if ( ckWARN_d(WARN_NON_UNICODE) |
4963 | || ( ckWARN_d(WARN_DEPRECATED) | |
83dc0f42 KW |
4964 | #ifndef UV_IS_QUAD |
4965 | && UNLIKELY(is_utf8_cp_above_31_bits(s, e)) | |
4966 | #else /* Below is 64-bit words */ | |
760c7c2f KW |
4967 | /* 2**63 and up meet these conditions provided we have |
4968 | * a 64-bit word. */ | |
4969 | # ifdef EBCDIC | |
83dc0f42 KW |
4970 | && *s == 0xFE |
4971 | && NATIVE_UTF8_TO_I8(s[1]) >= 0xA8 | |
760c7c2f | 4972 | # else |
83dc0f42 KW |
4973 | && *s == 0xFF |
4974 | /* s[1] being above 0x80 overflows */ | |
760c7c2f KW |
4975 | && s[2] >= 0x88 |
4976 | # endif | |
760c7c2f KW |
4977 | #endif |
4978 | )) { | |
15ca5930 | 4979 | /* A side effect of this function will be to warn */ |
2db24202 | 4980 | (void) utf8n_to_uvchr(s, e - s, NULL, UTF8_WARN_SUPER); |
7ee537e6 KW |
4981 | ok = FALSE; |
4982 | } | |
0876b9a0 | 4983 | } |
f2bf18cc | 4984 | else if (UNLIKELY(UTF8_IS_SURROGATE(s, e))) { |
8457b38f | 4985 | if (ckWARN_d(WARN_SURROGATE)) { |
15ca5930 KW |
4986 | /* This has a different warning than the one the called |
4987 | * function would output, so can't just call it, unlike we | |
4988 | * do for the non-chars and above-unicodes */ | |
2db24202 | 4989 | UV uv = utf8_to_uvchr_buf(s, e, NULL); |
8457b38f | 4990 | Perl_warner(aTHX_ packWARN(WARN_SURROGATE), |
147e3846 | 4991 | "Unicode surrogate U+%04" UVXf " is illegal in UTF-8", uv); |
8457b38f KW |
4992 | ok = FALSE; |
4993 | } | |
0876b9a0 | 4994 | } |
f2bf18cc | 4995 | else if (UNLIKELY(UTF8_IS_NONCHAR(s, e)) && (ckWARN_d(WARN_NONCHAR))) { |
15ca5930 | 4996 | /* A side effect of this function will be to warn */ |
2db24202 | 4997 | (void) utf8n_to_uvchr(s, e - s, NULL, UTF8_WARN_NONCHAR); |
0876b9a0 KW |
4998 | ok = FALSE; |
4999 | } | |
5000 | } | |
5001 | s += UTF8SKIP(s); | |
5002 | } | |
5003 | ||
5004 | return ok; | |
5005 | } | |
5006 | ||
0f830e0b | 5007 | /* |
87cea99e | 5008 | =for apidoc pv_uni_display |
d2cc3551 | 5009 | |
a1433954 KW |
5010 | Build to the scalar C<dsv> a displayable version of the string C<spv>, |
5011 | length C<len>, the displayable version being at most C<pvlim> bytes long | |
796b6530 | 5012 | (if longer, the rest is truncated and C<"..."> will be appended). |
0a2ef054 | 5013 | |
796b6530 KW |
5014 | The C<flags> argument can have C<UNI_DISPLAY_ISPRINT> set to display |
5015 | C<isPRINT()>able characters as themselves, C<UNI_DISPLAY_BACKSLASH> | |
5016 | to display the C<\\[nrfta\\]> as the backslashed versions (like C<"\n">) | |
5017 | (C<UNI_DISPLAY_BACKSLASH> is preferred over C<UNI_DISPLAY_ISPRINT> for C<"\\">). | |
5018 | C<UNI_DISPLAY_QQ> (and its alias C<UNI_DISPLAY_REGEX>) have both | |
5019 | C<UNI_DISPLAY_BACKSLASH> and C<UNI_DISPLAY_ISPRINT> turned on. | |
0a2ef054 | 5020 | |
a1433954 | 5021 | The pointer to the PV of the C<dsv> is returned. |
d2cc3551 | 5022 | |
119bc988 KW |
5023 | See also L</sv_uni_display>. |
5024 | ||
d2cc3551 | 5025 | =cut */ |
e6b2e755 | 5026 | char * |
e1ec3a88 | 5027 | Perl_pv_uni_display(pTHX_ SV *dsv, const U8 *spv, STRLEN len, STRLEN pvlim, UV flags) |
e6b2e755 JH |
5028 | { |
5029 | int truncated = 0; | |
e1ec3a88 | 5030 | const char *s, *e; |
e6b2e755 | 5031 | |
7918f24d NC |
5032 | PERL_ARGS_ASSERT_PV_UNI_DISPLAY; |
5033 | ||
9e2aa2e7 | 5034 | SvPVCLEAR(dsv); |
7fddd944 | 5035 | SvUTF8_off(dsv); |
e1ec3a88 | 5036 | for (s = (const char *)spv, e = s + len; s < e; s += UTF8SKIP(s)) { |
e6b2e755 | 5037 | UV u; |
a49f32c6 NC |
5038 | /* This serves double duty as a flag and a character to print after |
5039 | a \ when flags & UNI_DISPLAY_BACKSLASH is true. | |
5040 | */ | |
5041 | char ok = 0; | |
c728cb41 | 5042 | |
e6b2e755 JH |
5043 | if (pvlim && SvCUR(dsv) >= pvlim) { |
5044 | truncated++; | |
5045 | break; | |
5046 | } | |
4b88fb76 | 5047 | u = utf8_to_uvchr_buf((U8*)s, (U8*)e, 0); |
c728cb41 | 5048 | if (u < 256) { |
a3b680e6 | 5049 | const unsigned char c = (unsigned char)u & 0xFF; |
0bd48802 | 5050 | if (flags & UNI_DISPLAY_BACKSLASH) { |
a49f32c6 | 5051 | switch (c) { |
c728cb41 | 5052 | case '\n': |
a49f32c6 | 5053 | ok = 'n'; break; |
c728cb41 | 5054 | case '\r': |
a49f32c6 | 5055 | ok = 'r'; break; |
c728cb41 | 5056 | case '\t': |
a49f32c6 | 5057 | ok = 't'; break; |
c728cb41 | 5058 | case '\f': |
a49f32c6 | 5059 | ok = 'f'; break; |
c728cb41 | 5060 | case '\a': |
a49f32c6 | 5061 | ok = 'a'; break; |
c728cb41 | 5062 | case '\\': |
a49f32c6 | 5063 | ok = '\\'; break; |
c728cb41 JH |
5064 | default: break; |
5065 | } | |
a49f32c6 | 5066 | if (ok) { |
88c9ea1e | 5067 | const char string = ok; |
76f68e9b | 5068 | sv_catpvs(dsv, "\\"); |
5e7aa789 | 5069 | sv_catpvn(dsv, &string, 1); |
a49f32c6 | 5070 | } |
c728cb41 | 5071 | } |
00e86452 | 5072 | /* isPRINT() is the locale-blind version. */ |
a49f32c6 | 5073 | if (!ok && (flags & UNI_DISPLAY_ISPRINT) && isPRINT(c)) { |
88c9ea1e | 5074 | const char string = c; |
5e7aa789 | 5075 | sv_catpvn(dsv, &string, 1); |
a49f32c6 | 5076 | ok = 1; |
0a2ef054 | 5077 | } |
c728cb41 JH |
5078 | } |
5079 | if (!ok) | |
147e3846 | 5080 | Perl_sv_catpvf(aTHX_ dsv, "\\x{%" UVxf "}", u); |
e6b2e755 JH |
5081 | } |
5082 | if (truncated) | |
396482e1 | 5083 | sv_catpvs(dsv, "..."); |
48ef279e | 5084 | |
e6b2e755 JH |
5085 | return SvPVX(dsv); |
5086 | } | |
2b9d42f0 | 5087 | |
d2cc3551 | 5088 | /* |
87cea99e | 5089 | =for apidoc sv_uni_display |
d2cc3551 | 5090 | |
a1433954 KW |
5091 | Build to the scalar C<dsv> a displayable version of the scalar C<sv>, |
5092 | the displayable version being at most C<pvlim> bytes long | |
d2cc3551 | 5093 | (if longer, the rest is truncated and "..." will be appended). |
0a2ef054 | 5094 | |
a1433954 | 5095 | The C<flags> argument is as in L</pv_uni_display>(). |
0a2ef054 | 5096 | |
a1433954 | 5097 | The pointer to the PV of the C<dsv> is returned. |
d2cc3551 | 5098 | |
d4c19fe8 AL |
5099 | =cut |
5100 | */ | |
e6b2e755 JH |
5101 | char * |
5102 | Perl_sv_uni_display(pTHX_ SV *dsv, SV *ssv, STRLEN pvlim, UV flags) | |
5103 | { | |
8cdde9f8 NC |
5104 | const char * const ptr = |
5105 | isREGEXP(ssv) ? RX_WRAPPED((REGEXP*)ssv) : SvPVX_const(ssv); | |
5106 | ||
7918f24d NC |
5107 | PERL_ARGS_ASSERT_SV_UNI_DISPLAY; |
5108 | ||
8cdde9f8 | 5109 | return Perl_pv_uni_display(aTHX_ dsv, (const U8*)ptr, |
cfd0369c | 5110 | SvCUR(ssv), pvlim, flags); |
701a277b JH |
5111 | } |
5112 | ||
d2cc3551 | 5113 | /* |
e6226b18 | 5114 | =for apidoc foldEQ_utf8 |
d2cc3551 | 5115 | |
a1433954 | 5116 | Returns true if the leading portions of the strings C<s1> and C<s2> (either or both |
e6226b18 | 5117 | of which may be in UTF-8) are the same case-insensitively; false otherwise. |
d51c1b21 | 5118 | How far into the strings to compare is determined by other input parameters. |
8b35872c | 5119 | |
a1433954 KW |
5120 | If C<u1> is true, the string C<s1> is assumed to be in UTF-8-encoded Unicode; |
5121 | otherwise it is assumed to be in native 8-bit encoding. Correspondingly for C<u2> | |
5122 | with respect to C<s2>. | |
8b35872c | 5123 | |
a1433954 KW |
5124 | If the byte length C<l1> is non-zero, it says how far into C<s1> to check for fold |
5125 | equality. In other words, C<s1>+C<l1> will be used as a goal to reach. The | |
8b35872c | 5126 | scan will not be considered to be a match unless the goal is reached, and |
a1433954 KW |
5127 | scanning won't continue past that goal. Correspondingly for C<l2> with respect to |
5128 | C<s2>. | |
5129 | ||
796b6530 | 5130 | If C<pe1> is non-C<NULL> and the pointer it points to is not C<NULL>, that pointer is |
03bb5c85 KW |
5131 | considered an end pointer to the position 1 byte past the maximum point |
5132 | in C<s1> beyond which scanning will not continue under any circumstances. | |
5133 | (This routine assumes that UTF-8 encoded input strings are not malformed; | |
5134 | malformed input can cause it to read past C<pe1>). | |
5135 | This means that if both C<l1> and C<pe1> are specified, and C<pe1> | |
a1433954 KW |
5136 | is less than C<s1>+C<l1>, the match will never be successful because it can |
5137 | never | |
d51c1b21 | 5138 | get as far as its goal (and in fact is asserted against). Correspondingly for |
a1433954 | 5139 | C<pe2> with respect to C<s2>. |
8b35872c | 5140 | |
a1433954 KW |
5141 | At least one of C<s1> and C<s2> must have a goal (at least one of C<l1> and |
5142 | C<l2> must be non-zero), and if both do, both have to be | |
8b35872c KW |
5143 | reached for a successful match. Also, if the fold of a character is multiple |
5144 | characters, all of them must be matched (see tr21 reference below for | |
5145 | 'folding'). | |
5146 | ||
796b6530 | 5147 | Upon a successful match, if C<pe1> is non-C<NULL>, |
a1433954 KW |
5148 | it will be set to point to the beginning of the I<next> character of C<s1> |
5149 | beyond what was matched. Correspondingly for C<pe2> and C<s2>. | |
d2cc3551 JH |
5150 | |
5151 | For case-insensitiveness, the "casefolding" of Unicode is used | |
5152 | instead of upper/lowercasing both the characters, see | |
a1433954 | 5153 | L<http://www.unicode.org/unicode/reports/tr21/> (Case Mappings). |
d2cc3551 JH |
5154 | |
5155 | =cut */ | |
a33c29bc KW |
5156 | |
5157 | /* A flags parameter has been added which may change, and hence isn't | |
5158 | * externally documented. Currently it is: | |
5159 | * 0 for as-documented above | |
5160 | * FOLDEQ_UTF8_NOMIX_ASCII meaning that if a non-ASCII character folds to an | |
5161 | ASCII one, to not match | |
31f05a37 KW |
5162 | * FOLDEQ_LOCALE is set iff the rules from the current underlying |
5163 | * locale are to be used. | |
5164 | * FOLDEQ_S1_ALREADY_FOLDED s1 has already been folded before calling this | |
aa8ebe62 KW |
5165 | * routine. This allows that step to be skipped. |
5166 | * Currently, this requires s1 to be encoded as UTF-8 | |
5167 | * (u1 must be true), which is asserted for. | |
d635b710 KW |
5168 | * FOLDEQ_S1_FOLDS_SANE With either NOMIX_ASCII or LOCALE, no folds may |
5169 | * cross certain boundaries. Hence, the caller should | |
5170 | * let this function do the folding instead of | |
5171 | * pre-folding. This code contains an assertion to | |
5172 | * that effect. However, if the caller knows what | |
5173 | * it's doing, it can pass this flag to indicate that, | |
5174 | * and the assertion is skipped. | |
31f05a37 | 5175 | * FOLDEQ_S2_ALREADY_FOLDED Similarly. |
d635b710 | 5176 | * FOLDEQ_S2_FOLDS_SANE |
a33c29bc | 5177 | */ |
701a277b | 5178 | I32 |
5aaab254 | 5179 | 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 | 5180 | { |
eb578fdb KW |
5181 | const U8 *p1 = (const U8*)s1; /* Point to current char */ |
5182 | const U8 *p2 = (const U8*)s2; | |
5183 | const U8 *g1 = NULL; /* goal for s1 */ | |
5184 | const U8 *g2 = NULL; | |
5185 | const U8 *e1 = NULL; /* Don't scan s1 past this */ | |
5186 | U8 *f1 = NULL; /* Point to current folded */ | |
5187 | const U8 *e2 = NULL; | |
5188 | U8 *f2 = NULL; | |
48ef279e | 5189 | STRLEN n1 = 0, n2 = 0; /* Number of bytes in current char */ |
8b35872c KW |
5190 | U8 foldbuf1[UTF8_MAXBYTES_CASE+1]; |
5191 | U8 foldbuf2[UTF8_MAXBYTES_CASE+1]; | |
1d39b2cd | 5192 | U8 flags_for_folder = FOLD_FLAGS_FULL; |
8b35872c | 5193 | |
eda9cac1 | 5194 | PERL_ARGS_ASSERT_FOLDEQ_UTF8_FLAGS; |
8b35872c | 5195 | |
cea315b6 | 5196 | assert( ! ((flags & (FOLDEQ_UTF8_NOMIX_ASCII | FOLDEQ_LOCALE)) |
d635b710 KW |
5197 | && (((flags & FOLDEQ_S1_ALREADY_FOLDED) |
5198 | && !(flags & FOLDEQ_S1_FOLDS_SANE)) | |
5199 | || ((flags & FOLDEQ_S2_ALREADY_FOLDED) | |
5200 | && !(flags & FOLDEQ_S2_FOLDS_SANE))))); | |
b08f1bd5 KW |
5201 | /* The algorithm is to trial the folds without regard to the flags on |
5202 | * the first line of the above assert(), and then see if the result | |
5203 | * violates them. This means that the inputs can't be pre-folded to a | |
5204 | * violating result, hence the assert. This could be changed, with the | |
5205 | * addition of extra tests here for the already-folded case, which would | |
5206 | * slow it down. That cost is more than any possible gain for when these | |
5207 | * flags are specified, as the flags indicate /il or /iaa matching which | |
5208 | * is less common than /iu, and I (khw) also believe that real-world /il | |
5209 | * and /iaa matches are most likely to involve code points 0-255, and this | |
5210 | * function only under rare conditions gets called for 0-255. */ | |
18f762c3 | 5211 | |
1d39b2cd KW |
5212 | if (flags & FOLDEQ_LOCALE) { |
5213 | if (IN_UTF8_CTYPE_LOCALE) { | |
5214 | flags &= ~FOLDEQ_LOCALE; | |
5215 | } | |
5216 | else { | |
5217 | flags_for_folder |= FOLD_FLAGS_LOCALE; | |
5218 | } | |
31f05a37 KW |
5219 | } |
5220 | ||
8b35872c | 5221 | if (pe1) { |
48ef279e | 5222 | e1 = *(U8**)pe1; |
8b35872c KW |
5223 | } |
5224 | ||
5225 | if (l1) { | |
48ef279e | 5226 | g1 = (const U8*)s1 + l1; |
8b35872c KW |
5227 | } |
5228 | ||
5229 | if (pe2) { | |
48ef279e | 5230 | e2 = *(U8**)pe2; |
8b35872c KW |
5231 | } |
5232 | ||
5233 | if (l2) { | |
48ef279e | 5234 | g2 = (const U8*)s2 + l2; |
8b35872c KW |
5235 | } |
5236 | ||
5237 | /* Must have at least one goal */ | |
5238 | assert(g1 || g2); | |
5239 | ||
5240 | if (g1) { | |
5241 | ||
48ef279e KW |
5242 | /* Will never match if goal is out-of-bounds */ |
5243 | assert(! e1 || e1 >= g1); | |
8b35872c | 5244 | |
48ef279e KW |
5245 | /* Here, there isn't an end pointer, or it is beyond the goal. We |
5246 | * only go as far as the goal */ | |
5247 | e1 = g1; | |
8b35872c | 5248 | } |
313b38e5 NC |
5249 | else { |
5250 | assert(e1); /* Must have an end for looking at s1 */ | |
5251 | } | |
8b35872c KW |
5252 | |
5253 | /* Same for goal for s2 */ | |
5254 | if (g2) { | |
48ef279e KW |
5255 | assert(! e2 || e2 >= g2); |
5256 | e2 = g2; | |
8b35872c | 5257 | } |
313b38e5 NC |
5258 | else { |
5259 | assert(e2); | |
5260 | } | |
8b35872c | 5261 | |
18f762c3 KW |
5262 | /* If both operands are already folded, we could just do a memEQ on the |
5263 | * whole strings at once, but it would be better if the caller realized | |
5264 | * this and didn't even call us */ | |
5265 | ||
8b35872c KW |
5266 | /* Look through both strings, a character at a time */ |
5267 | while (p1 < e1 && p2 < e2) { | |
5268 | ||
d51c1b21 | 5269 | /* If at the beginning of a new character in s1, get its fold to use |
1d39b2cd | 5270 | * and the length of the fold. */ |
48ef279e | 5271 | if (n1 == 0) { |
18f762c3 KW |
5272 | if (flags & FOLDEQ_S1_ALREADY_FOLDED) { |
5273 | f1 = (U8 *) p1; | |
aa8ebe62 | 5274 | assert(u1); |
18f762c3 | 5275 | n1 = UTF8SKIP(f1); |
18f762c3 KW |
5276 | } |
5277 | else { | |
1d39b2cd KW |
5278 | if (isASCII(*p1) && ! (flags & FOLDEQ_LOCALE)) { |
5279 | ||
5280 | /* We have to forbid mixing ASCII with non-ASCII if the | |
5281 | * flags so indicate. And, we can short circuit having to | |
5282 | * call the general functions for this common ASCII case, | |
5283 | * all of whose non-locale folds are also ASCII, and hence | |
5284 | * UTF-8 invariants, so the UTF8ness of the strings is not | |
5285 | * relevant. */ | |
5286 | if ((flags & FOLDEQ_UTF8_NOMIX_ASCII) && ! isASCII(*p2)) { | |
5287 | return 0; | |
5288 | } | |
5289 | n1 = 1; | |
5290 | *foldbuf1 = toFOLD(*p1); | |
5291 | } | |
5292 | else if (u1) { | |
a1a5ec35 | 5293 | _toFOLD_utf8_flags(p1, e1, foldbuf1, &n1, flags_for_folder); |
1d39b2cd | 5294 | } |
4a4088c4 | 5295 | else { /* Not UTF-8, get UTF-8 fold */ |
1d39b2cd KW |
5296 | _to_uni_fold_flags(*p1, foldbuf1, &n1, flags_for_folder); |
5297 | } | |
5298 | f1 = foldbuf1; | |
5299 | } | |
48ef279e | 5300 | } |
8b35872c | 5301 | |
48ef279e | 5302 | if (n2 == 0) { /* Same for s2 */ |
18f762c3 KW |
5303 | if (flags & FOLDEQ_S2_ALREADY_FOLDED) { |
5304 | f2 = (U8 *) p2; | |
aa8ebe62 | 5305 | assert(u2); |
18f762c3 KW |
5306 | n2 = UTF8SKIP(f2); |
5307 | } | |
5308 | else { | |
1d39b2cd KW |
5309 | if (isASCII(*p2) && ! (flags & FOLDEQ_LOCALE)) { |
5310 | if ((flags & FOLDEQ_UTF8_NOMIX_ASCII) && ! isASCII(*p1)) { | |
5311 | return 0; | |
5312 | } | |
5313 | n2 = 1; | |
5314 | *foldbuf2 = toFOLD(*p2); | |
5315 | } | |
5316 | else if (u2) { | |
a1a5ec35 | 5317 | _toFOLD_utf8_flags(p2, e2, foldbuf2, &n2, flags_for_folder); |
1d39b2cd KW |
5318 | } |
5319 | else { | |
5320 | _to_uni_fold_flags(*p2, foldbuf2, &n2, flags_for_folder); | |
5321 | } | |
5322 | f2 = foldbuf2; | |
18f762c3 | 5323 | } |
48ef279e | 5324 | } |
8b35872c | 5325 | |
5001101e | 5326 | /* Here f1 and f2 point to the beginning of the strings to compare. |
227968da | 5327 | * These strings are the folds of the next character from each input |
4a4088c4 | 5328 | * string, stored in UTF-8. */ |
5e64d0fa | 5329 | |
48ef279e KW |
5330 | /* While there is more to look for in both folds, see if they |
5331 | * continue to match */ | |
5332 | while (n1 && n2) { | |
5333 | U8 fold_length = UTF8SKIP(f1); | |
5334 | if (fold_length != UTF8SKIP(f2) | |
5335 | || (fold_length == 1 && *f1 != *f2) /* Short circuit memNE | |
5336 | function call for single | |
a6d5f321 | 5337 | byte */ |
48ef279e KW |
5338 | || memNE((char*)f1, (char*)f2, fold_length)) |
5339 | { | |
e6226b18 | 5340 | return 0; /* mismatch */ |
48ef279e KW |
5341 | } |
5342 | ||
5343 | /* Here, they matched, advance past them */ | |
5344 | n1 -= fold_length; | |
5345 | f1 += fold_length; | |
5346 | n2 -= fold_length; | |
5347 | f2 += fold_length; | |
5348 | } | |
8b35872c | 5349 | |
48ef279e KW |
5350 | /* When reach the end of any fold, advance the input past it */ |
5351 | if (n1 == 0) { | |
5352 | p1 += u1 ? UTF8SKIP(p1) : 1; | |
5353 | } | |
5354 | if (n2 == 0) { | |
5355 | p2 += u2 ? UTF8SKIP(p2) : 1; | |
5356 | } | |
8b35872c KW |
5357 | } /* End of loop through both strings */ |
5358 | ||
5359 | /* A match is defined by each scan that specified an explicit length | |
5360 | * reaching its final goal, and the other not having matched a partial | |
5361 | * character (which can happen when the fold of a character is more than one | |
5362 | * character). */ | |
5363 | if (! ((g1 == 0 || p1 == g1) && (g2 == 0 || p2 == g2)) || n1 || n2) { | |
e6226b18 | 5364 | return 0; |
8b35872c KW |
5365 | } |
5366 | ||
5367 | /* Successful match. Set output pointers */ | |
5368 | if (pe1) { | |
48ef279e | 5369 | *pe1 = (char*)p1; |
8b35872c KW |
5370 | } |
5371 | if (pe2) { | |
48ef279e | 5372 | *pe2 = (char*)p2; |
8b35872c | 5373 | } |
e6226b18 | 5374 | return 1; |
e6b2e755 | 5375 | } |
701a277b | 5376 | |
f2645549 | 5377 | /* XXX The next two functions should likely be moved to mathoms.c once all |
37e7596b KW |
5378 | * occurrences of them are removed from the core; some cpan-upstream modules |
5379 | * still use them */ | |
5380 | ||
5381 | U8 * | |
5382 | Perl_uvuni_to_utf8(pTHX_ U8 *d, UV uv) | |
5383 | { | |
5384 | PERL_ARGS_ASSERT_UVUNI_TO_UTF8; | |
5385 | ||
5386 | return Perl_uvoffuni_to_utf8_flags(aTHX_ d, uv, 0); | |
5387 | } | |
5388 | ||
e505af10 KW |
5389 | /* |
5390 | =for apidoc utf8n_to_uvuni | |
5391 | ||
5392 | Instead use L</utf8_to_uvchr_buf>, or rarely, L</utf8n_to_uvchr>. | |
5393 | ||
5394 | This function was useful for code that wanted to handle both EBCDIC and | |
5395 | ASCII platforms with Unicode properties, but starting in Perl v5.20, the | |
5396 | distinctions between the platforms have mostly been made invisible to most | |
5397 | code, so this function is quite unlikely to be what you want. If you do need | |
5398 | this precise functionality, use instead | |
5399 | C<L<NATIVE_TO_UNI(utf8_to_uvchr_buf(...))|/utf8_to_uvchr_buf>> | |
5400 | or C<L<NATIVE_TO_UNI(utf8n_to_uvchr(...))|/utf8n_to_uvchr>>. | |
5401 | ||
5402 | =cut | |
5403 | */ | |
5404 | ||
37e7596b KW |
5405 | UV |
5406 | Perl_utf8n_to_uvuni(pTHX_ const U8 *s, STRLEN curlen, STRLEN *retlen, U32 flags) | |
5407 | { | |
5408 | PERL_ARGS_ASSERT_UTF8N_TO_UVUNI; | |
5409 | ||
5410 | return NATIVE_TO_UNI(utf8n_to_uvchr(s, curlen, retlen, flags)); | |
5411 | } | |
5412 | ||
5413 | /* | |
5414 | =for apidoc uvuni_to_utf8_flags | |
5415 | ||
5416 | Instead you almost certainly want to use L</uvchr_to_utf8> or | |
efa9cd84 | 5417 | L</uvchr_to_utf8_flags>. |
37e7596b KW |
5418 | |
5419 | This function is a deprecated synonym for L</uvoffuni_to_utf8_flags>, | |
5420 | which itself, while not deprecated, should be used only in isolated | |
5421 | circumstances. These functions were useful for code that wanted to handle | |
5422 | both EBCDIC and ASCII platforms with Unicode properties, but starting in Perl | |
5423 | v5.20, the distinctions between the platforms have mostly been made invisible | |
5424 | to most code, so this function is quite unlikely to be what you want. | |
5425 | ||
5426 | =cut | |
5427 | */ | |
5428 | ||
5429 | U8 * | |
5430 | Perl_uvuni_to_utf8_flags(pTHX_ U8 *d, UV uv, UV flags) | |
5431 | { | |
5432 | PERL_ARGS_ASSERT_UVUNI_TO_UTF8_FLAGS; | |
5433 | ||
5434 | return uvoffuni_to_utf8_flags(d, uv, flags); | |
5435 | } | |
5436 | ||
5437 | /* | |
14d04a33 | 5438 | * ex: set ts=8 sts=4 sw=4 et: |
37442d52 | 5439 | */ |