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 /* --------------------------------------------------------------------- */ 117 118 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 /* --------------------------------------------------------------------- */ 166 167 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 } 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 /* --------------------------------------------------------------------- */ 290 291 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 ConversionResult ConvertUTF8toUTF32 ( 345 const UTF8** sourceStart, const UTF8* sourceEnd, 346 UTF32** targetStart, UTF32* targetEnd, ConversionFlags flags) { 347 ConversionResult result = conversionOK; 348 const UTF8* source = *sourceStart; 349 UTF32* target = *targetStart; 350 while (source < sourceEnd) { 351 UTF32 ch = 0; 352 unsigned short extraBytesToRead = trailingBytesForUTF8[*source]; 353 if (source + extraBytesToRead >= sourceEnd) { 354 result = sourceExhausted; break; 355 } 356 /* Do this check whether lenient or strict */ 357 if (!isLegalUTF8(source, extraBytesToRead+1)) { 358 result = sourceIllegal; 359 break; 360 } 361 /* 362 * The cases all fall through. See "Note A" below. 363 */ 364 switch (extraBytesToRead) { 365 case 5: ch += *source++; ch <<= 6; 366 case 4: ch += *source++; ch <<= 6; 367 case 3: ch += *source++; ch <<= 6; 368 case 2: ch += *source++; ch <<= 6; 369 case 1: ch += *source++; ch <<= 6; 370 case 0: ch += *source++; 371 } 372 ch -= offsetsFromUTF8[extraBytesToRead]; 373 374 if (target >= targetEnd) { 375 source -= (extraBytesToRead+1); /* Back up the source pointer! */ 376 result = targetExhausted; break; 377 } 378 if (ch <= UNI_MAX_LEGAL_UTF32) { 379 /* 380 * UTF-16 surrogate values are illegal in UTF-32, and anything 381 * over Plane 17 (> 0x10FFFF) is illegal. 382 */ 383 if (ch >= UNI_SUR_HIGH_START && ch <= UNI_SUR_LOW_END) { 384 if (flags == strictConversion) { 385 source -= (extraBytesToRead+1); /* return to the illegal value itself */ 386 result = sourceIllegal; 387 break; 388 } else { 389 *target++ = UNI_REPLACEMENT_CHAR; 390 } 391 } else { 392 *target++ = ch; 393 } 394 } else { /* i.e., ch > UNI_MAX_LEGAL_UTF32 */ 395 result = sourceIllegal; 396 *target++ = UNI_REPLACEMENT_CHAR; 397 } 398 } 399 *sourceStart = source; 400 *targetStart = target; 401 return result; 402 } 403 #endif 404 405 /* --------------------------------------------------------------------- */ 406 407 /* 408 * Utility routine to tell whether a sequence of bytes is legal UTF-8. 409 * This must be called with the length pre-determined by the first byte. 410 * If not calling this from ConvertUTF8to*, then the length can be set by: 411 * length = trailingBytesForUTF8[*source]+1; 412 * and the sequence is illegal right away if there aren't that many bytes 413 * available. 414 * If presented with a length > 4, this returns false. The Unicode 415 * definition of UTF-8 goes up to 4-byte sequences. 416 */ 417 418 static Boolean isLegalUTF8(const UTF8 *source, int length) { 419 UTF8 a; 420 const UTF8 *srcptr = source+length; 421 switch (length) { 422 default: return false; 423 /* Everything else falls through when "true"... */ 424 case 4: if ((a = (*--srcptr)) < 0x80 || a > 0xBF) return false; 425 case 3: if ((a = (*--srcptr)) < 0x80 || a > 0xBF) return false; 426 case 2: if ((a = (*--srcptr)) > 0xBF) return false; 427 428 switch (*source) { 429 /* no fall-through in this inner switch */ 430 case 0xE0: if (a < 0xA0) return false; break; 431 case 0xED: if (a > 0x9F) return false; break; 432 case 0xF0: if (a < 0x90) return false; break; 433 case 0xF4: if (a > 0x8F) return false; break; 434 default: if (a < 0x80) return false; 435 } 436 437 case 1: if (*source >= 0x80 && *source < 0xC2) return false; 438 } 439 if (*source > 0xF4) return false; 440 return true; 441 } 442 443 /* --------------------------------------------------------------------- */ 444 445 /* 446 * Exported function to return whether a UTF-8 sequence is legal or not. 447 * This is not used here; it's just exported. 448 */ 449 Boolean isLegalUTF8Sequence(const UTF8 *source, const UTF8 *sourceEnd) { 450 int length = trailingBytesForUTF8[*source]+1; 451 if (source+length > sourceEnd) { 452 return false; 453 } 454 return isLegalUTF8(source, length); 455 } 456 457 /* --------------------------------------------------------------------- */ 458 459 ConversionResult ConvertUTF8toUTF16 ( 460 const UTF8** sourceStart, const UTF8* sourceEnd, 461 UTF16** targetStart, UTF16* targetEnd, ConversionFlags flags) { 462 ConversionResult result = conversionOK; 463 const UTF8* source = *sourceStart; 464 UTF16* target = *targetStart; 465 while (source < sourceEnd) { 466 UTF32 ch = 0; 467 unsigned short extraBytesToRead = trailingBytesForUTF8[*source]; 468 if (source + extraBytesToRead >= sourceEnd) { 469 result = sourceExhausted; break; 470 } 471 /* Do this check whether lenient or strict */ 472 if (!isLegalUTF8(source, extraBytesToRead+1)) { 473 result = sourceIllegal; 474 break; 475 } 476 /* 477 * The cases all fall through. See "Note A" below. 478 */ 479 switch (extraBytesToRead) { 480 case 5: ch += *source++; ch <<= 6; /* remember, illegal UTF-8 */ 481 case 4: ch += *source++; ch <<= 6; /* remember, illegal UTF-8 */ 482 case 3: ch += *source++; ch <<= 6; 483 case 2: ch += *source++; ch <<= 6; 484 case 1: ch += *source++; ch <<= 6; 485 case 0: ch += *source++; 486 } 487 ch -= offsetsFromUTF8[extraBytesToRead]; 488 489 if (target >= targetEnd) { 490 source -= (extraBytesToRead+1); /* Back up source pointer! */ 491 result = targetExhausted; break; 492 } 493 if (ch <= UNI_MAX_BMP) { /* Target is a character <= 0xFFFF */ 494 /* UTF-16 surrogate values are illegal in UTF-32 */ 495 if (ch >= UNI_SUR_HIGH_START && ch <= UNI_SUR_LOW_END) { 496 if (flags == strictConversion) { 497 source -= (extraBytesToRead+1); /* return to the illegal value itself */ 498 result = sourceIllegal; 499 break; 500 } else { 501 *target++ = UNI_REPLACEMENT_CHAR; 502 } 503 } else { 504 *target++ = (UTF16)ch; /* normal case */ 505 } 506 } else if (ch > UNI_MAX_UTF16) { 507 if (flags == strictConversion) { 508 result = sourceIllegal; 509 source -= (extraBytesToRead+1); /* return to the start */ 510 break; /* Bail out; shouldn't continue */ 511 } else { 512 *target++ = UNI_REPLACEMENT_CHAR; 513 } 514 } else { 515 /* target is a character in range 0xFFFF - 0x10FFFF. */ 516 if (target + 1 >= targetEnd) { 517 source -= (extraBytesToRead+1); /* Back up source pointer! */ 518 result = targetExhausted; break; 519 } 520 ch -= halfBase; 521 *target++ = (UTF16)((ch >> halfShift) + UNI_SUR_HIGH_START); 522 *target++ = (UTF16)((ch & halfMask) + UNI_SUR_LOW_START); 523 } 524 } 525 *sourceStart = source; 526 *targetStart = target; 527 return result; 528 } 529 530 /* --------------------------------------------------------------------- 531 532 Note A. 533 The fall-through switches in UTF-8 reading code save a 534 temp variable, some decrements & conditionals. The switches 535 are equivalent to the following loop: 536 { 537 int tmpBytesToRead = extraBytesToRead+1; 538 do { 539 ch += *source++; 540 --tmpBytesToRead; 541 if (tmpBytesToRead) ch <<= 6; 542 } while (tmpBytesToRead > 0); 543 } 544 In UTF-8 writing code, the switches on "bytesToWrite" are 545 similarly unrolled loops. 546 547 --------------------------------------------------------------------- */ 548
