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