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 "clang/Basic/ConvertUTF.h"
51 #ifdef CVTUTF_DEBUG
52 #include <stdio.h>
53 #endif
54
55 static const int halfShift = 10; /* used for shifting by 10 bits */
56
57 static const UTF32 halfBase = 0x0010000UL;
58 static const UTF32 halfMask = 0x3FFUL;
59
60 #define UNI_SUR_HIGH_START (UTF32)0xD800
61 #define UNI_SUR_HIGH_END (UTF32)0xDBFF
62 #define UNI_SUR_LOW_START (UTF32)0xDC00
63 #define UNI_SUR_LOW_END (UTF32)0xDFFF
64 #define false 0
65 #define true 1
66
67 /* --------------------------------------------------------------------- */
68
69 /*
70 * Index into the table below with the first byte of a UTF-8 sequence to
71 * get the number of trailing bytes that are supposed to follow it.
72 * Note that *legal* UTF-8 values can't have 4 or 5-bytes. The table is
73 * left as-is for anyone who may want to do such conversion, which was
74 * allowed in earlier algorithms.
75 */
76 static const char trailingBytesForUTF8[256] = {
77 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,
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 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,
84 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
85 };
86
87 /*
88 * Magic values subtracted from a buffer value during UTF8 conversion.
89 * This table contains as many values as there might be trailing bytes
90 * in a UTF-8 sequence.
91 */
92 static const UTF32 offsetsFromUTF8[6] = { 0x00000000UL, 0x00003080UL, 0x000E2080UL,
93 0x03C82080UL, 0xFA082080UL, 0x82082080UL };
94
95 /*
96 * Once the bits are split out into bytes of UTF-8, this is a mask OR-ed
97 * into the first byte, depending on how many bytes follow. There are
98 * as many entries in this table as there are UTF-8 sequence types.
99 * (I.e., one byte sequence, two byte... etc.). Remember that sequencs
100 * for *legal* UTF-8 will be 4 or fewer bytes total.
101 */
102 static const UTF8 firstByteMark[7] = { 0x00, 0x00, 0xC0, 0xE0, 0xF0, 0xF8, 0xFC };
103
104 /* --------------------------------------------------------------------- */
105
106 /* The interface converts a whole buffer to avoid function-call overhead.
107 * Constants have been gathered. Loops & conditionals have been removed as
108 * much as possible for efficiency, in favor of drop-through switches.
109 * (See "Note A" at the bottom of the file for equivalent code.)
110 * If your compiler supports it, the "isLegalUTF8" call can be turned
111 * into an inline function.
112 */
113
114 #ifdef CLANG_NEEDS_THESE_ONE_DAY
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 #endif
343
344 /* --------------------------------------------------------------------- */
345
346 /*
347 * Utility routine to tell whether a sequence of bytes is legal UTF-8.
348 * This must be called with the length pre-determined by the first byte.
349 * If not calling this from ConvertUTF8to*, then the length can be set by:
350 * length = trailingBytesForUTF8[*source]+1;
351 * and the sequence is illegal right away if there aren't that many bytes
352 * available.
353 * If presented with a length > 4, this returns false. The Unicode
354 * definition of UTF-8 goes up to 4-byte sequences.
355 */
356
isLegalUTF8(const UTF8 * source,int length)357 static Boolean isLegalUTF8(const UTF8 *source, int length) {
358 UTF8 a;
359 const UTF8 *srcptr = source+length;
360 switch (length) {
361 default: return false;
362 /* Everything else falls through when "true"... */
363 case 4: if ((a = (*--srcptr)) < 0x80 || a > 0xBF) return false;
364 case 3: if ((a = (*--srcptr)) < 0x80 || a > 0xBF) return false;
365 case 2: if ((a = (*--srcptr)) > 0xBF) return false;
366
367 switch (*source) {
368 /* no fall-through in this inner switch */
369 case 0xE0: if (a < 0xA0) return false; break;
370 case 0xED: if (a > 0x9F) return false; break;
371 case 0xF0: if (a < 0x90) return false; break;
372 case 0xF4: if (a > 0x8F) return false; break;
373 default: if (a < 0x80) return false;
374 }
375
376 case 1: if (*source >= 0x80 && *source < 0xC2) return false;
377 }
378 if (*source > 0xF4) return false;
379 return true;
380 }
381
382 /* --------------------------------------------------------------------- */
383
384 /*
385 * Exported function to return whether a UTF-8 sequence is legal or not.
386 * This is not used here; it's just exported.
387 */
isLegalUTF8Sequence(const UTF8 * source,const UTF8 * sourceEnd)388 Boolean isLegalUTF8Sequence(const UTF8 *source, const UTF8 *sourceEnd) {
389 int length = trailingBytesForUTF8[*source]+1;
390 if (length > sourceEnd - source) {
391 return false;
392 }
393 return isLegalUTF8(source, length);
394 }
395
396 /* --------------------------------------------------------------------- */
397
398 /*
399 * Exported function to return whether a UTF-8 string is legal or not.
400 * This is not used here; it's just exported.
401 */
isLegalUTF8String(const UTF8 * source,const UTF8 * sourceEnd)402 Boolean isLegalUTF8String(const UTF8 *source, const UTF8 *sourceEnd) {
403 while (source != sourceEnd) {
404 int length = trailingBytesForUTF8[*source] + 1;
405 if (length > sourceEnd - source || !isLegalUTF8(source, length))
406 return false;
407 source += length;
408 }
409 return true;
410 }
411
412 /* --------------------------------------------------------------------- */
413
ConvertUTF8toUTF16(const UTF8 ** sourceStart,const UTF8 * sourceEnd,UTF16 ** targetStart,UTF16 * targetEnd,ConversionFlags flags)414 ConversionResult ConvertUTF8toUTF16 (
415 const UTF8** sourceStart, const UTF8* sourceEnd,
416 UTF16** targetStart, UTF16* targetEnd, ConversionFlags flags) {
417 ConversionResult result = conversionOK;
418 const UTF8* source = *sourceStart;
419 UTF16* target = *targetStart;
420 while (source < sourceEnd) {
421 UTF32 ch = 0;
422 unsigned short extraBytesToRead = trailingBytesForUTF8[*source];
423 if (extraBytesToRead >= sourceEnd - source) {
424 result = sourceExhausted; break;
425 }
426 /* Do this check whether lenient or strict */
427 if (!isLegalUTF8(source, extraBytesToRead+1)) {
428 result = sourceIllegal;
429 break;
430 }
431 /*
432 * The cases all fall through. See "Note A" below.
433 */
434 switch (extraBytesToRead) {
435 case 5: ch += *source++; ch <<= 6; /* remember, illegal UTF-8 */
436 case 4: ch += *source++; ch <<= 6; /* remember, illegal UTF-8 */
437 case 3: ch += *source++; ch <<= 6;
438 case 2: ch += *source++; ch <<= 6;
439 case 1: ch += *source++; ch <<= 6;
440 case 0: ch += *source++;
441 }
442 ch -= offsetsFromUTF8[extraBytesToRead];
443
444 if (target >= targetEnd) {
445 source -= (extraBytesToRead+1); /* Back up source pointer! */
446 result = targetExhausted; break;
447 }
448 if (ch <= UNI_MAX_BMP) { /* Target is a character <= 0xFFFF */
449 /* UTF-16 surrogate values are illegal in UTF-32 */
450 if (ch >= UNI_SUR_HIGH_START && ch <= UNI_SUR_LOW_END) {
451 if (flags == strictConversion) {
452 source -= (extraBytesToRead+1); /* return to the illegal value itself */
453 result = sourceIllegal;
454 break;
455 } else {
456 *target++ = UNI_REPLACEMENT_CHAR;
457 }
458 } else {
459 *target++ = (UTF16)ch; /* normal case */
460 }
461 } else if (ch > UNI_MAX_UTF16) {
462 if (flags == strictConversion) {
463 result = sourceIllegal;
464 source -= (extraBytesToRead+1); /* return to the start */
465 break; /* Bail out; shouldn't continue */
466 } else {
467 *target++ = UNI_REPLACEMENT_CHAR;
468 }
469 } else {
470 /* target is a character in range 0xFFFF - 0x10FFFF. */
471 if (target + 1 >= targetEnd) {
472 source -= (extraBytesToRead+1); /* Back up source pointer! */
473 result = targetExhausted; break;
474 }
475 ch -= halfBase;
476 *target++ = (UTF16)((ch >> halfShift) + UNI_SUR_HIGH_START);
477 *target++ = (UTF16)((ch & halfMask) + UNI_SUR_LOW_START);
478 }
479 }
480 *sourceStart = source;
481 *targetStart = target;
482 return result;
483 }
484
485 /* --------------------------------------------------------------------- */
486
ConvertUTF8toUTF32(const UTF8 ** sourceStart,const UTF8 * sourceEnd,UTF32 ** targetStart,UTF32 * targetEnd,ConversionFlags flags)487 ConversionResult ConvertUTF8toUTF32 (
488 const UTF8** sourceStart, const UTF8* sourceEnd,
489 UTF32** targetStart, UTF32* targetEnd, ConversionFlags flags) {
490 ConversionResult result = conversionOK;
491 const UTF8* source = *sourceStart;
492 UTF32* target = *targetStart;
493 while (source < sourceEnd) {
494 UTF32 ch = 0;
495 unsigned short extraBytesToRead = trailingBytesForUTF8[*source];
496 if (extraBytesToRead >= sourceEnd - source) {
497 result = sourceExhausted; break;
498 }
499 /* Do this check whether lenient or strict */
500 if (!isLegalUTF8(source, extraBytesToRead+1)) {
501 result = sourceIllegal;
502 break;
503 }
504 /*
505 * The cases all fall through. See "Note A" below.
506 */
507 switch (extraBytesToRead) {
508 case 5: ch += *source++; ch <<= 6;
509 case 4: ch += *source++; ch <<= 6;
510 case 3: ch += *source++; ch <<= 6;
511 case 2: ch += *source++; ch <<= 6;
512 case 1: ch += *source++; ch <<= 6;
513 case 0: ch += *source++;
514 }
515 ch -= offsetsFromUTF8[extraBytesToRead];
516
517 if (target >= targetEnd) {
518 source -= (extraBytesToRead+1); /* Back up the source pointer! */
519 result = targetExhausted; break;
520 }
521 if (ch <= UNI_MAX_LEGAL_UTF32) {
522 /*
523 * UTF-16 surrogate values are illegal in UTF-32, and anything
524 * over Plane 17 (> 0x10FFFF) is illegal.
525 */
526 if (ch >= UNI_SUR_HIGH_START && ch <= UNI_SUR_LOW_END) {
527 if (flags == strictConversion) {
528 source -= (extraBytesToRead+1); /* return to the illegal value itself */
529 result = sourceIllegal;
530 break;
531 } else {
532 *target++ = UNI_REPLACEMENT_CHAR;
533 }
534 } else {
535 *target++ = ch;
536 }
537 } else { /* i.e., ch > UNI_MAX_LEGAL_UTF32 */
538 result = sourceIllegal;
539 *target++ = UNI_REPLACEMENT_CHAR;
540 }
541 }
542 *sourceStart = source;
543 *targetStart = target;
544 return result;
545 }
546
547 /* ---------------------------------------------------------------------
548
549 Note A.
550 The fall-through switches in UTF-8 reading code save a
551 temp variable, some decrements & conditionals. The switches
552 are equivalent to the following loop:
553 {
554 int tmpBytesToRead = extraBytesToRead+1;
555 do {
556 ch += *source++;
557 --tmpBytesToRead;
558 if (tmpBytesToRead) ch <<= 6;
559 } while (tmpBytesToRead > 0);
560 }
561 In UTF-8 writing code, the switches on "bytesToWrite" are
562 similarly unrolled loops.
563
564 --------------------------------------------------------------------- */
565