1 /*===--- ConvertUTF.c - Universal Character Names conversions ---------------===
2 *
3 * Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
4 * See https://llvm.org/LICENSE.txt for license information.
5 * SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
6 *
7 *===------------------------------------------------------------------------=*/
8 /*
9 * Copyright © 1991-2015 Unicode, Inc. All rights reserved.
10 * Distributed under the Terms of Use in
11 * http://www.unicode.org/copyright.html.
12 *
13 * Permission is hereby granted, free of charge, to any person obtaining
14 * a copy of the Unicode data files and any associated documentation
15 * (the "Data Files") or Unicode software and any associated documentation
16 * (the "Software") to deal in the Data Files or Software
17 * without restriction, including without limitation the rights to use,
18 * copy, modify, merge, publish, distribute, and/or sell copies of
19 * the Data Files or Software, and to permit persons to whom the Data Files
20 * or Software are furnished to do so, provided that
21 * (a) this copyright and permission notice appear with all copies
22 * of the Data Files or Software,
23 * (b) this copyright and permission notice appear in associated
24 * documentation, and
25 * (c) there is clear notice in each modified Data File or in the Software
26 * as well as in the documentation associated with the Data File(s) or
27 * Software that the data or software has been modified.
28 *
29 * THE DATA FILES AND SOFTWARE ARE PROVIDED "AS IS", WITHOUT WARRANTY OF
30 * ANY KIND, EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE
31 * WARRANTIES OF MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
32 * NONINFRINGEMENT OF THIRD PARTY RIGHTS.
33 * IN NO EVENT SHALL THE COPYRIGHT HOLDER OR HOLDERS INCLUDED IN THIS
34 * NOTICE BE LIABLE FOR ANY CLAIM, OR ANY SPECIAL INDIRECT OR CONSEQUENTIAL
35 * DAMAGES, OR ANY DAMAGES WHATSOEVER RESULTING FROM LOSS OF USE,
36 * DATA OR PROFITS, WHETHER IN AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER
37 * TORTIOUS ACTION, ARISING OUT OF OR IN CONNECTION WITH THE USE OR
38 * PERFORMANCE OF THE DATA FILES OR SOFTWARE.
39 *
40 * Except as contained in this notice, the name of a copyright holder
41 * shall not be used in advertising or otherwise to promote the sale,
42 * use or other dealings in these Data Files or Software without prior
43 * written authorization of the copyright holder.
44 */
45
46 /* ---------------------------------------------------------------------
47
48 Conversions between UTF32, UTF-16, and UTF-8. Source code file.
49 Author: Mark E. Davis, 1994.
50 Rev History: Rick McGowan, fixes & updates May 2001.
51 Sept 2001: fixed const & error conditions per
52 mods suggested by S. Parent & A. Lillich.
53 June 2002: Tim Dodd added detection and handling of incomplete
54 source sequences, enhanced error detection, added casts
55 to eliminate compiler warnings.
56 July 2003: slight mods to back out aggressive FFFE detection.
57 Jan 2004: updated switches in from-UTF8 conversions.
58 Oct 2004: updated to use UNI_MAX_LEGAL_UTF32 in UTF-32 conversions.
59
60 See the header file "ConvertUTF.h" for complete documentation.
61
62 ------------------------------------------------------------------------ */
63
64 #include "llvm/Support/ConvertUTF.h"
65 #ifdef CVTUTF_DEBUG
66 #include <stdio.h>
67 #endif
68 #include <assert.h>
69
70 /*
71 * This code extensively uses fall-through switches.
72 * Keep the compiler from warning about that.
73 */
74 #if defined(__clang__) && defined(__has_warning)
75 # if __has_warning("-Wimplicit-fallthrough")
76 # define ConvertUTF_DISABLE_WARNINGS \
77 _Pragma("clang diagnostic push") \
78 _Pragma("clang diagnostic ignored \"-Wimplicit-fallthrough\"")
79 # define ConvertUTF_RESTORE_WARNINGS \
80 _Pragma("clang diagnostic pop")
81 # endif
82 #elif defined(__GNUC__) && __GNUC__ > 6
83 # define ConvertUTF_DISABLE_WARNINGS \
84 _Pragma("GCC diagnostic push") \
85 _Pragma("GCC diagnostic ignored \"-Wimplicit-fallthrough\"")
86 # define ConvertUTF_RESTORE_WARNINGS \
87 _Pragma("GCC diagnostic pop")
88 #endif
89 #ifndef ConvertUTF_DISABLE_WARNINGS
90 # define ConvertUTF_DISABLE_WARNINGS
91 #endif
92 #ifndef ConvertUTF_RESTORE_WARNINGS
93 # define ConvertUTF_RESTORE_WARNINGS
94 #endif
95
96 ConvertUTF_DISABLE_WARNINGS
97
98 namespace llvm {
99
100 static const int halfShift = 10; /* used for shifting by 10 bits */
101
102 static const UTF32 halfBase = 0x0010000UL;
103 static const UTF32 halfMask = 0x3FFUL;
104
105 #define UNI_SUR_HIGH_START (UTF32)0xD800
106 #define UNI_SUR_HIGH_END (UTF32)0xDBFF
107 #define UNI_SUR_LOW_START (UTF32)0xDC00
108 #define UNI_SUR_LOW_END (UTF32)0xDFFF
109
110 /* --------------------------------------------------------------------- */
111
112 /*
113 * Index into the table below with the first byte of a UTF-8 sequence to
114 * get the number of trailing bytes that are supposed to follow it.
115 * Note that *legal* UTF-8 values can't have 4 or 5-bytes. The table is
116 * left as-is for anyone who may want to do such conversion, which was
117 * allowed in earlier algorithms.
118 */
119 static const char trailingBytesForUTF8[256] = {
120 0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0, 0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,
121 0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0, 0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,
122 0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0, 0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,
123 0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0, 0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,
124 0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0, 0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,
125 0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0, 0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,
126 1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1, 1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,
127 2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2, 3,3,3,3,3,3,3,3,4,4,4,4,5,5,5,5
128 };
129
130 /*
131 * Magic values subtracted from a buffer value during UTF8 conversion.
132 * This table contains as many values as there might be trailing bytes
133 * in a UTF-8 sequence.
134 */
135 static const UTF32 offsetsFromUTF8[6] = { 0x00000000UL, 0x00003080UL, 0x000E2080UL,
136 0x03C82080UL, 0xFA082080UL, 0x82082080UL };
137
138 /*
139 * Once the bits are split out into bytes of UTF-8, this is a mask OR-ed
140 * into the first byte, depending on how many bytes follow. There are
141 * as many entries in this table as there are UTF-8 sequence types.
142 * (I.e., one byte sequence, two byte... etc.). Remember that sequencs
143 * for *legal* UTF-8 will be 4 or fewer bytes total.
144 */
145 static const UTF8 firstByteMark[7] = { 0x00, 0x00, 0xC0, 0xE0, 0xF0, 0xF8, 0xFC };
146
147 /* --------------------------------------------------------------------- */
148
149 /* The interface converts a whole buffer to avoid function-call overhead.
150 * Constants have been gathered. Loops & conditionals have been removed as
151 * much as possible for efficiency, in favor of drop-through switches.
152 * (See "Note A" at the bottom of the file for equivalent code.)
153 * If your compiler supports it, the "isLegalUTF8" call can be turned
154 * into an inline function.
155 */
156
157
158 /* --------------------------------------------------------------------- */
159
ConvertUTF32toUTF16(const UTF32 ** sourceStart,const UTF32 * sourceEnd,UTF16 ** targetStart,UTF16 * targetEnd,ConversionFlags flags)160 ConversionResult ConvertUTF32toUTF16 (
161 const UTF32** sourceStart, const UTF32* sourceEnd,
162 UTF16** targetStart, UTF16* targetEnd, ConversionFlags flags) {
163 ConversionResult result = conversionOK;
164 const UTF32* source = *sourceStart;
165 UTF16* target = *targetStart;
166 while (source < sourceEnd) {
167 UTF32 ch;
168 if (target >= targetEnd) {
169 result = targetExhausted; break;
170 }
171 ch = *source++;
172 if (ch <= UNI_MAX_BMP) { /* Target is a character <= 0xFFFF */
173 /* UTF-16 surrogate values are illegal in UTF-32; 0xffff or 0xfffe are both reserved values */
174 if (ch >= UNI_SUR_HIGH_START && ch <= UNI_SUR_LOW_END) {
175 if (flags == strictConversion) {
176 --source; /* return to the illegal value itself */
177 result = sourceIllegal;
178 break;
179 } else {
180 *target++ = UNI_REPLACEMENT_CHAR;
181 }
182 } else {
183 *target++ = (UTF16)ch; /* normal case */
184 }
185 } else if (ch > UNI_MAX_LEGAL_UTF32) {
186 if (flags == strictConversion) {
187 result = sourceIllegal;
188 } else {
189 *target++ = UNI_REPLACEMENT_CHAR;
190 }
191 } else {
192 /* target is a character in range 0xFFFF - 0x10FFFF. */
193 if (target + 1 >= targetEnd) {
194 --source; /* Back up source pointer! */
195 result = targetExhausted; break;
196 }
197 ch -= halfBase;
198 *target++ = (UTF16)((ch >> halfShift) + UNI_SUR_HIGH_START);
199 *target++ = (UTF16)((ch & halfMask) + UNI_SUR_LOW_START);
200 }
201 }
202 *sourceStart = source;
203 *targetStart = target;
204 return result;
205 }
206
207 /* --------------------------------------------------------------------- */
208
ConvertUTF16toUTF32(const UTF16 ** sourceStart,const UTF16 * sourceEnd,UTF32 ** targetStart,UTF32 * targetEnd,ConversionFlags flags)209 ConversionResult ConvertUTF16toUTF32 (
210 const UTF16** sourceStart, const UTF16* sourceEnd,
211 UTF32** targetStart, UTF32* targetEnd, ConversionFlags flags) {
212 ConversionResult result = conversionOK;
213 const UTF16* source = *sourceStart;
214 UTF32* target = *targetStart;
215 UTF32 ch, ch2;
216 while (source < sourceEnd) {
217 const UTF16* oldSource = source; /* In case we have to back up because of target overflow. */
218 ch = *source++;
219 /* If we have a surrogate pair, convert to UTF32 first. */
220 if (ch >= UNI_SUR_HIGH_START && ch <= UNI_SUR_HIGH_END) {
221 /* If the 16 bits following the high surrogate are in the source buffer... */
222 if (source < sourceEnd) {
223 ch2 = *source;
224 /* If it's a low surrogate, convert to UTF32. */
225 if (ch2 >= UNI_SUR_LOW_START && ch2 <= UNI_SUR_LOW_END) {
226 ch = ((ch - UNI_SUR_HIGH_START) << halfShift)
227 + (ch2 - UNI_SUR_LOW_START) + halfBase;
228 ++source;
229 } else if (flags == strictConversion) { /* it's an unpaired high surrogate */
230 --source; /* return to the illegal value itself */
231 result = sourceIllegal;
232 break;
233 }
234 } else { /* We don't have the 16 bits following the high surrogate. */
235 --source; /* return to the high surrogate */
236 result = sourceExhausted;
237 break;
238 }
239 } else if (flags == strictConversion) {
240 /* UTF-16 surrogate values are illegal in UTF-32 */
241 if (ch >= UNI_SUR_LOW_START && ch <= UNI_SUR_LOW_END) {
242 --source; /* return to the illegal value itself */
243 result = sourceIllegal;
244 break;
245 }
246 }
247 if (target >= targetEnd) {
248 source = oldSource; /* Back up source pointer! */
249 result = targetExhausted; break;
250 }
251 *target++ = ch;
252 }
253 *sourceStart = source;
254 *targetStart = target;
255 #ifdef CVTUTF_DEBUG
256 if (result == sourceIllegal) {
257 fprintf(stderr, "ConvertUTF16toUTF32 illegal seq 0x%04x,%04x\n", ch, ch2);
258 fflush(stderr);
259 }
260 #endif
261 return result;
262 }
ConvertUTF16toUTF8(const UTF16 ** sourceStart,const UTF16 * sourceEnd,UTF8 ** targetStart,UTF8 * targetEnd,ConversionFlags flags)263 ConversionResult ConvertUTF16toUTF8 (
264 const UTF16** sourceStart, const UTF16* sourceEnd,
265 UTF8** targetStart, UTF8* targetEnd, ConversionFlags flags) {
266 ConversionResult result = conversionOK;
267 const UTF16* source = *sourceStart;
268 UTF8* target = *targetStart;
269 while (source < sourceEnd) {
270 UTF32 ch;
271 unsigned short bytesToWrite = 0;
272 const UTF32 byteMask = 0xBF;
273 const UTF32 byteMark = 0x80;
274 const UTF16* oldSource = source; /* In case we have to back up because of target overflow. */
275 ch = *source++;
276 /* If we have a surrogate pair, convert to UTF32 first. */
277 if (ch >= UNI_SUR_HIGH_START && ch <= UNI_SUR_HIGH_END) {
278 /* If the 16 bits following the high surrogate are in the source buffer... */
279 if (source < sourceEnd) {
280 UTF32 ch2 = *source;
281 /* If it's a low surrogate, convert to UTF32. */
282 if (ch2 >= UNI_SUR_LOW_START && ch2 <= UNI_SUR_LOW_END) {
283 ch = ((ch - UNI_SUR_HIGH_START) << halfShift)
284 + (ch2 - UNI_SUR_LOW_START) + halfBase;
285 ++source;
286 } else if (flags == strictConversion) { /* it's an unpaired high surrogate */
287 --source; /* return to the illegal value itself */
288 result = sourceIllegal;
289 break;
290 }
291 } else { /* We don't have the 16 bits following the high surrogate. */
292 --source; /* return to the high surrogate */
293 result = sourceExhausted;
294 break;
295 }
296 } else if (flags == strictConversion) {
297 /* UTF-16 surrogate values are illegal in UTF-32 */
298 if (ch >= UNI_SUR_LOW_START && ch <= UNI_SUR_LOW_END) {
299 --source; /* return to the illegal value itself */
300 result = sourceIllegal;
301 break;
302 }
303 }
304 /* Figure out how many bytes the result will require */
305 if (ch < (UTF32)0x80) { bytesToWrite = 1;
306 } else if (ch < (UTF32)0x800) { bytesToWrite = 2;
307 } else if (ch < (UTF32)0x10000) { bytesToWrite = 3;
308 } else if (ch < (UTF32)0x110000) { bytesToWrite = 4;
309 } else { bytesToWrite = 3;
310 ch = UNI_REPLACEMENT_CHAR;
311 }
312
313 target += bytesToWrite;
314 if (target > targetEnd) {
315 source = oldSource; /* Back up source pointer! */
316 target -= bytesToWrite; result = targetExhausted; break;
317 }
318 switch (bytesToWrite) { /* note: everything falls through. */
319 case 4: *--target = (UTF8)((ch | byteMark) & byteMask); ch >>= 6;
320 case 3: *--target = (UTF8)((ch | byteMark) & byteMask); ch >>= 6;
321 case 2: *--target = (UTF8)((ch | byteMark) & byteMask); ch >>= 6;
322 case 1: *--target = (UTF8)(ch | firstByteMark[bytesToWrite]);
323 }
324 target += bytesToWrite;
325 }
326 *sourceStart = source;
327 *targetStart = target;
328 return result;
329 }
330
331 /* --------------------------------------------------------------------- */
332
ConvertUTF32toUTF8(const UTF32 ** sourceStart,const UTF32 * sourceEnd,UTF8 ** targetStart,UTF8 * targetEnd,ConversionFlags flags)333 ConversionResult ConvertUTF32toUTF8 (
334 const UTF32** sourceStart, const UTF32* sourceEnd,
335 UTF8** targetStart, UTF8* targetEnd, ConversionFlags flags) {
336 ConversionResult result = conversionOK;
337 const UTF32* source = *sourceStart;
338 UTF8* target = *targetStart;
339 while (source < sourceEnd) {
340 UTF32 ch;
341 unsigned short bytesToWrite = 0;
342 const UTF32 byteMask = 0xBF;
343 const UTF32 byteMark = 0x80;
344 ch = *source++;
345 if (flags == strictConversion ) {
346 /* UTF-16 surrogate values are illegal in UTF-32 */
347 if (ch >= UNI_SUR_HIGH_START && ch <= UNI_SUR_LOW_END) {
348 --source; /* return to the illegal value itself */
349 result = sourceIllegal;
350 break;
351 }
352 }
353 /*
354 * Figure out how many bytes the result will require. Turn any
355 * illegally large UTF32 things (> Plane 17) into replacement chars.
356 */
357 if (ch < (UTF32)0x80) { bytesToWrite = 1;
358 } else if (ch < (UTF32)0x800) { bytesToWrite = 2;
359 } else if (ch < (UTF32)0x10000) { bytesToWrite = 3;
360 } else if (ch <= UNI_MAX_LEGAL_UTF32) { bytesToWrite = 4;
361 } else { bytesToWrite = 3;
362 ch = UNI_REPLACEMENT_CHAR;
363 result = sourceIllegal;
364 }
365
366 target += bytesToWrite;
367 if (target > targetEnd) {
368 --source; /* Back up source pointer! */
369 target -= bytesToWrite; result = targetExhausted; break;
370 }
371 switch (bytesToWrite) { /* note: everything falls through. */
372 case 4: *--target = (UTF8)((ch | byteMark) & byteMask); ch >>= 6;
373 case 3: *--target = (UTF8)((ch | byteMark) & byteMask); ch >>= 6;
374 case 2: *--target = (UTF8)((ch | byteMark) & byteMask); ch >>= 6;
375 case 1: *--target = (UTF8) (ch | firstByteMark[bytesToWrite]);
376 }
377 target += bytesToWrite;
378 }
379 *sourceStart = source;
380 *targetStart = target;
381 return result;
382 }
383
384 /* --------------------------------------------------------------------- */
385
386 /*
387 * Utility routine to tell whether a sequence of bytes is legal UTF-8.
388 * This must be called with the length pre-determined by the first byte.
389 * If not calling this from ConvertUTF8to*, then the length can be set by:
390 * length = trailingBytesForUTF8[*source]+1;
391 * and the sequence is illegal right away if there aren't that many bytes
392 * available.
393 * If presented with a length > 4, this returns false. The Unicode
394 * definition of UTF-8 goes up to 4-byte sequences.
395 */
396
isLegalUTF8(const UTF8 * source,int length)397 static Boolean isLegalUTF8(const UTF8 *source, int length) {
398 UTF8 a;
399 const UTF8 *srcptr = source+length;
400 switch (length) {
401 default: return false;
402 /* Everything else falls through when "true"... */
403 case 4: if ((a = (*--srcptr)) < 0x80 || a > 0xBF) return false;
404 case 3: if ((a = (*--srcptr)) < 0x80 || a > 0xBF) return false;
405 case 2: if ((a = (*--srcptr)) < 0x80 || a > 0xBF) return false;
406
407 switch (*source) {
408 /* no fall-through in this inner switch */
409 case 0xE0: if (a < 0xA0) return false; break;
410 case 0xED: if (a > 0x9F) return false; break;
411 case 0xF0: if (a < 0x90) return false; break;
412 case 0xF4: if (a > 0x8F) return false; break;
413 default: if (a < 0x80) return false;
414 }
415
416 case 1: if (*source >= 0x80 && *source < 0xC2) return false;
417 }
418 if (*source > 0xF4) return false;
419 return true;
420 }
421
422 /* --------------------------------------------------------------------- */
423
424 /*
425 * Exported function to return whether a UTF-8 sequence is legal or not.
426 * This is not used here; it's just exported.
427 */
isLegalUTF8Sequence(const UTF8 * source,const UTF8 * sourceEnd)428 Boolean isLegalUTF8Sequence(const UTF8 *source, const UTF8 *sourceEnd) {
429 int length = trailingBytesForUTF8[*source]+1;
430 if (length > sourceEnd - source) {
431 return false;
432 }
433 return isLegalUTF8(source, length);
434 }
435
436 /*
437 * Exported function to return the size of the first utf-8 code unit sequence,
438 * Or 0 if the sequence is not valid;
439 */
getUTF8SequenceSize(const UTF8 * source,const UTF8 * sourceEnd)440 unsigned getUTF8SequenceSize(const UTF8 *source, const UTF8 *sourceEnd) {
441 int length = trailingBytesForUTF8[*source] + 1;
442 return (length <= sourceEnd - source && isLegalUTF8(source, length)) ? length
443 : 0;
444 }
445
446 /* --------------------------------------------------------------------- */
447
448 static unsigned
findMaximalSubpartOfIllFormedUTF8Sequence(const UTF8 * source,const UTF8 * sourceEnd)449 findMaximalSubpartOfIllFormedUTF8Sequence(const UTF8 *source,
450 const UTF8 *sourceEnd) {
451 UTF8 b1, b2, b3;
452
453 assert(!isLegalUTF8Sequence(source, sourceEnd));
454
455 /*
456 * Unicode 6.3.0, D93b:
457 *
458 * Maximal subpart of an ill-formed subsequence: The longest code unit
459 * subsequence starting at an unconvertible offset that is either:
460 * a. the initial subsequence of a well-formed code unit sequence, or
461 * b. a subsequence of length one.
462 */
463
464 if (source == sourceEnd)
465 return 0;
466
467 /*
468 * Perform case analysis. See Unicode 6.3.0, Table 3-7. Well-Formed UTF-8
469 * Byte Sequences.
470 */
471
472 b1 = *source;
473 ++source;
474 if (b1 >= 0xC2 && b1 <= 0xDF) {
475 /*
476 * First byte is valid, but we know that this code unit sequence is
477 * invalid, so the maximal subpart has to end after the first byte.
478 */
479 return 1;
480 }
481
482 if (source == sourceEnd)
483 return 1;
484
485 b2 = *source;
486 ++source;
487
488 if (b1 == 0xE0) {
489 return (b2 >= 0xA0 && b2 <= 0xBF) ? 2 : 1;
490 }
491 if (b1 >= 0xE1 && b1 <= 0xEC) {
492 return (b2 >= 0x80 && b2 <= 0xBF) ? 2 : 1;
493 }
494 if (b1 == 0xED) {
495 return (b2 >= 0x80 && b2 <= 0x9F) ? 2 : 1;
496 }
497 if (b1 >= 0xEE && b1 <= 0xEF) {
498 return (b2 >= 0x80 && b2 <= 0xBF) ? 2 : 1;
499 }
500 if (b1 == 0xF0) {
501 if (b2 >= 0x90 && b2 <= 0xBF) {
502 if (source == sourceEnd)
503 return 2;
504
505 b3 = *source;
506 return (b3 >= 0x80 && b3 <= 0xBF) ? 3 : 2;
507 }
508 return 1;
509 }
510 if (b1 >= 0xF1 && b1 <= 0xF3) {
511 if (b2 >= 0x80 && b2 <= 0xBF) {
512 if (source == sourceEnd)
513 return 2;
514
515 b3 = *source;
516 return (b3 >= 0x80 && b3 <= 0xBF) ? 3 : 2;
517 }
518 return 1;
519 }
520 if (b1 == 0xF4) {
521 if (b2 >= 0x80 && b2 <= 0x8F) {
522 if (source == sourceEnd)
523 return 2;
524
525 b3 = *source;
526 return (b3 >= 0x80 && b3 <= 0xBF) ? 3 : 2;
527 }
528 return 1;
529 }
530
531 assert((b1 >= 0x80 && b1 <= 0xC1) || b1 >= 0xF5);
532 /*
533 * There are no valid sequences that start with these bytes. Maximal subpart
534 * is defined to have length 1 in these cases.
535 */
536 return 1;
537 }
538
539 /* --------------------------------------------------------------------- */
540
541 /*
542 * Exported function to return the total number of bytes in a codepoint
543 * represented in UTF-8, given the value of the first byte.
544 */
getNumBytesForUTF8(UTF8 first)545 unsigned getNumBytesForUTF8(UTF8 first) {
546 return trailingBytesForUTF8[first] + 1;
547 }
548
549 /* --------------------------------------------------------------------- */
550
551 /*
552 * Exported function to return whether a UTF-8 string is legal or not.
553 * This is not used here; it's just exported.
554 */
isLegalUTF8String(const UTF8 ** source,const UTF8 * sourceEnd)555 Boolean isLegalUTF8String(const UTF8 **source, const UTF8 *sourceEnd) {
556 while (*source != sourceEnd) {
557 int length = trailingBytesForUTF8[**source] + 1;
558 if (length > sourceEnd - *source || !isLegalUTF8(*source, length))
559 return false;
560 *source += length;
561 }
562 return true;
563 }
564
565 /* --------------------------------------------------------------------- */
566
ConvertUTF8toUTF16(const UTF8 ** sourceStart,const UTF8 * sourceEnd,UTF16 ** targetStart,UTF16 * targetEnd,ConversionFlags flags)567 ConversionResult ConvertUTF8toUTF16 (
568 const UTF8** sourceStart, const UTF8* sourceEnd,
569 UTF16** targetStart, UTF16* targetEnd, ConversionFlags flags) {
570 ConversionResult result = conversionOK;
571 const UTF8* source = *sourceStart;
572 UTF16* target = *targetStart;
573 while (source < sourceEnd) {
574 UTF32 ch = 0;
575 unsigned short extraBytesToRead = trailingBytesForUTF8[*source];
576 if (extraBytesToRead >= sourceEnd - source) {
577 result = sourceExhausted; break;
578 }
579 /* Do this check whether lenient or strict */
580 if (!isLegalUTF8(source, extraBytesToRead+1)) {
581 result = sourceIllegal;
582 break;
583 }
584 /*
585 * The cases all fall through. See "Note A" below.
586 */
587 switch (extraBytesToRead) {
588 case 5: ch += *source++; ch <<= 6; /* remember, illegal UTF-8 */
589 case 4: ch += *source++; ch <<= 6; /* remember, illegal UTF-8 */
590 case 3: ch += *source++; ch <<= 6;
591 case 2: ch += *source++; ch <<= 6;
592 case 1: ch += *source++; ch <<= 6;
593 case 0: ch += *source++;
594 }
595 ch -= offsetsFromUTF8[extraBytesToRead];
596
597 if (target >= targetEnd) {
598 source -= (extraBytesToRead+1); /* Back up source pointer! */
599 result = targetExhausted; break;
600 }
601 if (ch <= UNI_MAX_BMP) { /* Target is a character <= 0xFFFF */
602 /* UTF-16 surrogate values are illegal in UTF-32 */
603 if (ch >= UNI_SUR_HIGH_START && ch <= UNI_SUR_LOW_END) {
604 if (flags == strictConversion) {
605 source -= (extraBytesToRead+1); /* return to the illegal value itself */
606 result = sourceIllegal;
607 break;
608 } else {
609 *target++ = UNI_REPLACEMENT_CHAR;
610 }
611 } else {
612 *target++ = (UTF16)ch; /* normal case */
613 }
614 } else if (ch > UNI_MAX_UTF16) {
615 if (flags == strictConversion) {
616 result = sourceIllegal;
617 source -= (extraBytesToRead+1); /* return to the start */
618 break; /* Bail out; shouldn't continue */
619 } else {
620 *target++ = UNI_REPLACEMENT_CHAR;
621 }
622 } else {
623 /* target is a character in range 0xFFFF - 0x10FFFF. */
624 if (target + 1 >= targetEnd) {
625 source -= (extraBytesToRead+1); /* Back up source pointer! */
626 result = targetExhausted; break;
627 }
628 ch -= halfBase;
629 *target++ = (UTF16)((ch >> halfShift) + UNI_SUR_HIGH_START);
630 *target++ = (UTF16)((ch & halfMask) + UNI_SUR_LOW_START);
631 }
632 }
633 *sourceStart = source;
634 *targetStart = target;
635 return result;
636 }
637
638 /* --------------------------------------------------------------------- */
639
ConvertUTF8toUTF32Impl(const UTF8 ** sourceStart,const UTF8 * sourceEnd,UTF32 ** targetStart,UTF32 * targetEnd,ConversionFlags flags,Boolean InputIsPartial)640 static ConversionResult ConvertUTF8toUTF32Impl(
641 const UTF8** sourceStart, const UTF8* sourceEnd,
642 UTF32** targetStart, UTF32* targetEnd, ConversionFlags flags,
643 Boolean InputIsPartial) {
644 ConversionResult result = conversionOK;
645 const UTF8* source = *sourceStart;
646 UTF32* target = *targetStart;
647 while (source < sourceEnd) {
648 UTF32 ch = 0;
649 unsigned short extraBytesToRead = trailingBytesForUTF8[*source];
650 if (extraBytesToRead >= sourceEnd - source) {
651 if (flags == strictConversion || InputIsPartial) {
652 result = sourceExhausted;
653 break;
654 } else {
655 result = sourceIllegal;
656
657 /*
658 * Replace the maximal subpart of ill-formed sequence with
659 * replacement character.
660 */
661 source += findMaximalSubpartOfIllFormedUTF8Sequence(source,
662 sourceEnd);
663 *target++ = UNI_REPLACEMENT_CHAR;
664 continue;
665 }
666 }
667 if (target >= targetEnd) {
668 result = targetExhausted; break;
669 }
670
671 /* Do this check whether lenient or strict */
672 if (!isLegalUTF8(source, extraBytesToRead+1)) {
673 result = sourceIllegal;
674 if (flags == strictConversion) {
675 /* Abort conversion. */
676 break;
677 } else {
678 /*
679 * Replace the maximal subpart of ill-formed sequence with
680 * replacement character.
681 */
682 source += findMaximalSubpartOfIllFormedUTF8Sequence(source,
683 sourceEnd);
684 *target++ = UNI_REPLACEMENT_CHAR;
685 continue;
686 }
687 }
688 /*
689 * The cases all fall through. See "Note A" below.
690 */
691 switch (extraBytesToRead) {
692 case 5: ch += *source++; ch <<= 6;
693 case 4: ch += *source++; ch <<= 6;
694 case 3: ch += *source++; ch <<= 6;
695 case 2: ch += *source++; ch <<= 6;
696 case 1: ch += *source++; ch <<= 6;
697 case 0: ch += *source++;
698 }
699 ch -= offsetsFromUTF8[extraBytesToRead];
700
701 if (ch <= UNI_MAX_LEGAL_UTF32) {
702 /*
703 * UTF-16 surrogate values are illegal in UTF-32, and anything
704 * over Plane 17 (> 0x10FFFF) is illegal.
705 */
706 if (ch >= UNI_SUR_HIGH_START && ch <= UNI_SUR_LOW_END) {
707 if (flags == strictConversion) {
708 source -= (extraBytesToRead+1); /* return to the illegal value itself */
709 result = sourceIllegal;
710 break;
711 } else {
712 *target++ = UNI_REPLACEMENT_CHAR;
713 }
714 } else {
715 *target++ = ch;
716 }
717 } else { /* i.e., ch > UNI_MAX_LEGAL_UTF32 */
718 result = sourceIllegal;
719 *target++ = UNI_REPLACEMENT_CHAR;
720 }
721 }
722 *sourceStart = source;
723 *targetStart = target;
724 return result;
725 }
726
ConvertUTF8toUTF32Partial(const UTF8 ** sourceStart,const UTF8 * sourceEnd,UTF32 ** targetStart,UTF32 * targetEnd,ConversionFlags flags)727 ConversionResult ConvertUTF8toUTF32Partial(const UTF8 **sourceStart,
728 const UTF8 *sourceEnd,
729 UTF32 **targetStart,
730 UTF32 *targetEnd,
731 ConversionFlags flags) {
732 return ConvertUTF8toUTF32Impl(sourceStart, sourceEnd, targetStart, targetEnd,
733 flags, /*InputIsPartial=*/true);
734 }
735
ConvertUTF8toUTF32(const UTF8 ** sourceStart,const UTF8 * sourceEnd,UTF32 ** targetStart,UTF32 * targetEnd,ConversionFlags flags)736 ConversionResult ConvertUTF8toUTF32(const UTF8 **sourceStart,
737 const UTF8 *sourceEnd, UTF32 **targetStart,
738 UTF32 *targetEnd, ConversionFlags flags) {
739 return ConvertUTF8toUTF32Impl(sourceStart, sourceEnd, targetStart, targetEnd,
740 flags, /*InputIsPartial=*/false);
741 }
742
743 /* ---------------------------------------------------------------------
744
745 Note A.
746 The fall-through switches in UTF-8 reading code save a
747 temp variable, some decrements & conditionals. The switches
748 are equivalent to the following loop:
749 {
750 int tmpBytesToRead = extraBytesToRead+1;
751 do {
752 ch += *source++;
753 --tmpBytesToRead;
754 if (tmpBytesToRead) ch <<= 6;
755 } while (tmpBytesToRead > 0);
756 }
757 In UTF-8 writing code, the switches on "bytesToWrite" are
758 similarly unrolled loops.
759
760 --------------------------------------------------------------------- */
761
762 } // namespace llvm
763
764 ConvertUTF_RESTORE_WARNINGS
765