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 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 */ 113 114 115 /* --------------------------------------------------------------------- */ 116 117 ConversionResult ConvertUTF32toUTF16 ( 118 const UTF32** sourceStart, const UTF32* sourceEnd, 119 UTF16** targetStart, UTF16* targetEnd, ConversionFlags flags) { 120 ConversionResult result = conversionOK; 121 const UTF32* source = *sourceStart; 122 UTF16* target = *targetStart; 123 while (source < sourceEnd) { 124 UTF32 ch; 125 if (target >= targetEnd) { 126 result = targetExhausted; break; 127 } 128 ch = *source++; 129 if (ch <= UNI_MAX_BMP) { /* Target is a character <= 0xFFFF */ 130 /* UTF-16 surrogate values are illegal in UTF-32; 0xffff or 0xfffe are both reserved values */ 131 if (ch >= UNI_SUR_HIGH_START && ch <= UNI_SUR_LOW_END) { 132 if (flags == strictConversion) { 133 --source; /* return to the illegal value itself */ 134 result = sourceIllegal; 135 break; 136 } else { 137 *target++ = UNI_REPLACEMENT_CHAR; 138 } 139 } else { 140 *target++ = (UTF16)ch; /* normal case */ 141 } 142 } else if (ch > UNI_MAX_LEGAL_UTF32) { 143 if (flags == strictConversion) { 144 result = sourceIllegal; 145 } else { 146 *target++ = UNI_REPLACEMENT_CHAR; 147 } 148 } else { 149 /* target is a character in range 0xFFFF - 0x10FFFF. */ 150 if (target + 1 >= targetEnd) { 151 --source; /* Back up source pointer! */ 152 result = targetExhausted; break; 153 } 154 ch -= halfBase; 155 *target++ = (UTF16)((ch >> halfShift) + UNI_SUR_HIGH_START); 156 *target++ = (UTF16)((ch & halfMask) + UNI_SUR_LOW_START); 157 } 158 } 159 *sourceStart = source; 160 *targetStart = target; 161 return result; 162 } 163 164 /* --------------------------------------------------------------------- */ 165 166 ConversionResult ConvertUTF16toUTF32 ( 167 const UTF16** sourceStart, const UTF16* sourceEnd, 168 UTF32** targetStart, UTF32* targetEnd, ConversionFlags flags) { 169 ConversionResult result = conversionOK; 170 const UTF16* source = *sourceStart; 171 UTF32* target = *targetStart; 172 UTF32 ch, ch2; 173 while (source < sourceEnd) { 174 const UTF16* oldSource = source; /* In case we have to back up because of target overflow. */ 175 ch = *source++; 176 /* If we have a surrogate pair, convert to UTF32 first. */ 177 if (ch >= UNI_SUR_HIGH_START && ch <= UNI_SUR_HIGH_END) { 178 /* If the 16 bits following the high surrogate are in the source buffer... */ 179 if (source < sourceEnd) { 180 ch2 = *source; 181 /* If it's a low surrogate, convert to UTF32. */ 182 if (ch2 >= UNI_SUR_LOW_START && ch2 <= UNI_SUR_LOW_END) { 183 ch = ((ch - UNI_SUR_HIGH_START) << halfShift) 184 + (ch2 - UNI_SUR_LOW_START) + halfBase; 185 ++source; 186 } else if (flags == strictConversion) { /* it's an unpaired high surrogate */ 187 --source; /* return to the illegal value itself */ 188 result = sourceIllegal; 189 break; 190 } 191 } else { /* We don't have the 16 bits following the high surrogate. */ 192 --source; /* return to the high surrogate */ 193 result = sourceExhausted; 194 break; 195 } 196 } else if (flags == strictConversion) { 197 /* UTF-16 surrogate values are illegal in UTF-32 */ 198 if (ch >= UNI_SUR_LOW_START && ch <= UNI_SUR_LOW_END) { 199 --source; /* return to the illegal value itself */ 200 result = sourceIllegal; 201 break; 202 } 203 } 204 if (target >= targetEnd) { 205 source = oldSource; /* Back up source pointer! */ 206 result = targetExhausted; break; 207 } 208 *target++ = ch; 209 } 210 *sourceStart = source; 211 *targetStart = target; 212 #ifdef CVTUTF_DEBUG 213 if (result == sourceIllegal) { 214 fprintf(stderr, "ConvertUTF16toUTF32 illegal seq 0x%04x,%04x\n", ch, ch2); 215 fflush(stderr); 216 } 217 #endif 218 return result; 219 } 220 ConversionResult ConvertUTF16toUTF8 ( 221 const UTF16** sourceStart, const UTF16* sourceEnd, 222 UTF8** targetStart, UTF8* targetEnd, ConversionFlags flags) { 223 ConversionResult result = conversionOK; 224 const UTF16* source = *sourceStart; 225 UTF8* target = *targetStart; 226 while (source < sourceEnd) { 227 UTF32 ch; 228 unsigned short bytesToWrite = 0; 229 const UTF32 byteMask = 0xBF; 230 const UTF32 byteMark = 0x80; 231 const UTF16* oldSource = source; /* In case we have to back up because of target overflow. */ 232 ch = *source++; 233 /* If we have a surrogate pair, convert to UTF32 first. */ 234 if (ch >= UNI_SUR_HIGH_START && ch <= UNI_SUR_HIGH_END) { 235 /* If the 16 bits following the high surrogate are in the source buffer... */ 236 if (source < sourceEnd) { 237 UTF32 ch2 = *source; 238 /* If it's a low surrogate, convert to UTF32. */ 239 if (ch2 >= UNI_SUR_LOW_START && ch2 <= UNI_SUR_LOW_END) { 240 ch = ((ch - UNI_SUR_HIGH_START) << halfShift) 241 + (ch2 - UNI_SUR_LOW_START) + halfBase; 242 ++source; 243 } else if (flags == strictConversion) { /* it's an unpaired high surrogate */ 244 --source; /* return to the illegal value itself */ 245 result = sourceIllegal; 246 break; 247 } 248 } else { /* We don't have the 16 bits following the high surrogate. */ 249 --source; /* return to the high surrogate */ 250 result = sourceExhausted; 251 break; 252 } 253 } else if (flags == strictConversion) { 254 /* UTF-16 surrogate values are illegal in UTF-32 */ 255 if (ch >= UNI_SUR_LOW_START && ch <= UNI_SUR_LOW_END) { 256 --source; /* return to the illegal value itself */ 257 result = sourceIllegal; 258 break; 259 } 260 } 261 /* Figure out how many bytes the result will require */ 262 if (ch < (UTF32)0x80) { bytesToWrite = 1; 263 } else if (ch < (UTF32)0x800) { bytesToWrite = 2; 264 } else if (ch < (UTF32)0x10000) { bytesToWrite = 3; 265 } else if (ch < (UTF32)0x110000) { bytesToWrite = 4; 266 } else { bytesToWrite = 3; 267 ch = UNI_REPLACEMENT_CHAR; 268 } 269 270 target += bytesToWrite; 271 if (target > targetEnd) { 272 source = oldSource; /* Back up source pointer! */ 273 target -= bytesToWrite; result = targetExhausted; break; 274 } 275 switch (bytesToWrite) { /* note: everything falls through. */ 276 case 4: *--target = (UTF8)((ch | byteMark) & byteMask); ch >>= 6; 277 case 3: *--target = (UTF8)((ch | byteMark) & byteMask); ch >>= 6; 278 case 2: *--target = (UTF8)((ch | byteMark) & byteMask); ch >>= 6; 279 case 1: *--target = (UTF8)(ch | firstByteMark[bytesToWrite]); 280 } 281 target += bytesToWrite; 282 } 283 *sourceStart = source; 284 *targetStart = target; 285 return result; 286 } 287 288 /* --------------------------------------------------------------------- */ 289 290 ConversionResult ConvertUTF32toUTF8 ( 291 const UTF32** sourceStart, const UTF32* sourceEnd, 292 UTF8** targetStart, UTF8* targetEnd, ConversionFlags flags) { 293 ConversionResult result = conversionOK; 294 const UTF32* source = *sourceStart; 295 UTF8* target = *targetStart; 296 while (source < sourceEnd) { 297 UTF32 ch; 298 unsigned short bytesToWrite = 0; 299 const UTF32 byteMask = 0xBF; 300 const UTF32 byteMark = 0x80; 301 ch = *source++; 302 if (flags == strictConversion ) { 303 /* UTF-16 surrogate values are illegal in UTF-32 */ 304 if (ch >= UNI_SUR_HIGH_START && ch <= UNI_SUR_LOW_END) { 305 --source; /* return to the illegal value itself */ 306 result = sourceIllegal; 307 break; 308 } 309 } 310 /* 311 * Figure out how many bytes the result will require. Turn any 312 * illegally large UTF32 things (> Plane 17) into replacement chars. 313 */ 314 if (ch < (UTF32)0x80) { bytesToWrite = 1; 315 } else if (ch < (UTF32)0x800) { bytesToWrite = 2; 316 } else if (ch < (UTF32)0x10000) { bytesToWrite = 3; 317 } else if (ch <= UNI_MAX_LEGAL_UTF32) { bytesToWrite = 4; 318 } else { bytesToWrite = 3; 319 ch = UNI_REPLACEMENT_CHAR; 320 result = sourceIllegal; 321 } 322 323 target += bytesToWrite; 324 if (target > targetEnd) { 325 --source; /* Back up source pointer! */ 326 target -= bytesToWrite; result = targetExhausted; break; 327 } 328 switch (bytesToWrite) { /* note: everything falls through. */ 329 case 4: *--target = (UTF8)((ch | byteMark) & byteMask); ch >>= 6; 330 case 3: *--target = (UTF8)((ch | byteMark) & byteMask); ch >>= 6; 331 case 2: *--target = (UTF8)((ch | byteMark) & byteMask); ch >>= 6; 332 case 1: *--target = (UTF8) (ch | firstByteMark[bytesToWrite]); 333 } 334 target += bytesToWrite; 335 } 336 *sourceStart = source; 337 *targetStart = target; 338 return result; 339 } 340 341 /* --------------------------------------------------------------------- */ 342 343 /* 344 * Utility routine to tell whether a sequence of bytes is legal UTF-8. 345 * This must be called with the length pre-determined by the first byte. 346 * If not calling this from ConvertUTF8to*, then the length can be set by: 347 * length = trailingBytesForUTF8[*source]+1; 348 * and the sequence is illegal right away if there aren't that many bytes 349 * available. 350 * If presented with a length > 4, this returns false. The Unicode 351 * definition of UTF-8 goes up to 4-byte sequences. 352 */ 353 354 static Boolean isLegalUTF8(const UTF8 *source, int length) { 355 UTF8 a; 356 const UTF8 *srcptr = source+length; 357 switch (length) { 358 default: return false; 359 /* Everything else falls through when "true"... */ 360 case 4: if ((a = (*--srcptr)) < 0x80 || a > 0xBF) return false; 361 case 3: if ((a = (*--srcptr)) < 0x80 || a > 0xBF) return false; 362 case 2: if ((a = (*--srcptr)) < 0x80 || a > 0xBF) return false; 363 364 switch (*source) { 365 /* no fall-through in this inner switch */ 366 case 0xE0: if (a < 0xA0) return false; break; 367 case 0xED: if (a > 0x9F) return false; break; 368 case 0xF0: if (a < 0x90) return false; break; 369 case 0xF4: if (a > 0x8F) return false; break; 370 default: if (a < 0x80) return false; 371 } 372 373 case 1: if (*source >= 0x80 && *source < 0xC2) return false; 374 } 375 if (*source > 0xF4) return false; 376 return true; 377 } 378 379 /* --------------------------------------------------------------------- */ 380 381 /* 382 * Exported function to return whether a UTF-8 sequence is legal or not. 383 * This is not used here; it's just exported. 384 */ 385 Boolean isLegalUTF8Sequence(const UTF8 *source, const UTF8 *sourceEnd) { 386 int length = trailingBytesForUTF8[*source]+1; 387 if (length > sourceEnd - source) { 388 return false; 389 } 390 return isLegalUTF8(source, length); 391 } 392 393 /* --------------------------------------------------------------------- */ 394 395 /* 396 * Exported function to return the total number of bytes in a codepoint 397 * represented in UTF-8, given the value of the first byte. 398 */ 399 unsigned getNumBytesForUTF8(UTF8 first) { 400 return trailingBytesForUTF8[first] + 1; 401 } 402 403 /* --------------------------------------------------------------------- */ 404 405 /* 406 * Exported function to return whether a UTF-8 string is legal or not. 407 * This is not used here; it's just exported. 408 */ 409 Boolean isLegalUTF8String(const UTF8 **source, const UTF8 *sourceEnd) { 410 while (*source != sourceEnd) { 411 int length = trailingBytesForUTF8[**source] + 1; 412 if (length > sourceEnd - *source || !isLegalUTF8(*source, length)) 413 return false; 414 *source += length; 415 } 416 return true; 417 } 418 419 /* --------------------------------------------------------------------- */ 420 421 ConversionResult ConvertUTF8toUTF16 ( 422 const UTF8** sourceStart, const UTF8* sourceEnd, 423 UTF16** targetStart, UTF16* targetEnd, ConversionFlags flags) { 424 ConversionResult result = conversionOK; 425 const UTF8* source = *sourceStart; 426 UTF16* target = *targetStart; 427 while (source < sourceEnd) { 428 UTF32 ch = 0; 429 unsigned short extraBytesToRead = trailingBytesForUTF8[*source]; 430 if (extraBytesToRead >= sourceEnd - source) { 431 result = sourceExhausted; break; 432 } 433 /* Do this check whether lenient or strict */ 434 if (!isLegalUTF8(source, extraBytesToRead+1)) { 435 result = sourceIllegal; 436 break; 437 } 438 /* 439 * The cases all fall through. See "Note A" below. 440 */ 441 switch (extraBytesToRead) { 442 case 5: ch += *source++; ch <<= 6; /* remember, illegal UTF-8 */ 443 case 4: ch += *source++; ch <<= 6; /* remember, illegal UTF-8 */ 444 case 3: ch += *source++; ch <<= 6; 445 case 2: ch += *source++; ch <<= 6; 446 case 1: ch += *source++; ch <<= 6; 447 case 0: ch += *source++; 448 } 449 ch -= offsetsFromUTF8[extraBytesToRead]; 450 451 if (target >= targetEnd) { 452 source -= (extraBytesToRead+1); /* Back up source pointer! */ 453 result = targetExhausted; break; 454 } 455 if (ch <= UNI_MAX_BMP) { /* Target is a character <= 0xFFFF */ 456 /* UTF-16 surrogate values are illegal in UTF-32 */ 457 if (ch >= UNI_SUR_HIGH_START && ch <= UNI_SUR_LOW_END) { 458 if (flags == strictConversion) { 459 source -= (extraBytesToRead+1); /* return to the illegal value itself */ 460 result = sourceIllegal; 461 break; 462 } else { 463 *target++ = UNI_REPLACEMENT_CHAR; 464 } 465 } else { 466 *target++ = (UTF16)ch; /* normal case */ 467 } 468 } else if (ch > UNI_MAX_UTF16) { 469 if (flags == strictConversion) { 470 result = sourceIllegal; 471 source -= (extraBytesToRead+1); /* return to the start */ 472 break; /* Bail out; shouldn't continue */ 473 } else { 474 *target++ = UNI_REPLACEMENT_CHAR; 475 } 476 } else { 477 /* target is a character in range 0xFFFF - 0x10FFFF. */ 478 if (target + 1 >= targetEnd) { 479 source -= (extraBytesToRead+1); /* Back up source pointer! */ 480 result = targetExhausted; break; 481 } 482 ch -= halfBase; 483 *target++ = (UTF16)((ch >> halfShift) + UNI_SUR_HIGH_START); 484 *target++ = (UTF16)((ch & halfMask) + UNI_SUR_LOW_START); 485 } 486 } 487 *sourceStart = source; 488 *targetStart = target; 489 return result; 490 } 491 492 /* --------------------------------------------------------------------- */ 493 494 ConversionResult ConvertUTF8toUTF32 ( 495 const UTF8** sourceStart, const UTF8* sourceEnd, 496 UTF32** targetStart, UTF32* targetEnd, ConversionFlags flags) { 497 ConversionResult result = conversionOK; 498 const UTF8* source = *sourceStart; 499 UTF32* target = *targetStart; 500 while (source < sourceEnd) { 501 UTF32 ch = 0; 502 unsigned short extraBytesToRead = trailingBytesForUTF8[*source]; 503 if (extraBytesToRead >= sourceEnd - source) { 504 result = sourceExhausted; break; 505 } 506 /* Do this check whether lenient or strict */ 507 if (!isLegalUTF8(source, extraBytesToRead+1)) { 508 result = sourceIllegal; 509 break; 510 } 511 /* 512 * The cases all fall through. See "Note A" below. 513 */ 514 switch (extraBytesToRead) { 515 case 5: ch += *source++; ch <<= 6; 516 case 4: ch += *source++; ch <<= 6; 517 case 3: ch += *source++; ch <<= 6; 518 case 2: ch += *source++; ch <<= 6; 519 case 1: ch += *source++; ch <<= 6; 520 case 0: ch += *source++; 521 } 522 ch -= offsetsFromUTF8[extraBytesToRead]; 523 524 if (target >= targetEnd) { 525 source -= (extraBytesToRead+1); /* Back up the source pointer! */ 526 result = targetExhausted; break; 527 } 528 if (ch <= UNI_MAX_LEGAL_UTF32) { 529 /* 530 * UTF-16 surrogate values are illegal in UTF-32, and anything 531 * over Plane 17 (> 0x10FFFF) is illegal. 532 */ 533 if (ch >= UNI_SUR_HIGH_START && ch <= UNI_SUR_LOW_END) { 534 if (flags == strictConversion) { 535 source -= (extraBytesToRead+1); /* return to the illegal value itself */ 536 result = sourceIllegal; 537 break; 538 } else { 539 *target++ = UNI_REPLACEMENT_CHAR; 540 } 541 } else { 542 *target++ = ch; 543 } 544 } else { /* i.e., ch > UNI_MAX_LEGAL_UTF32 */ 545 result = sourceIllegal; 546 *target++ = UNI_REPLACEMENT_CHAR; 547 } 548 } 549 *sourceStart = source; 550 *targetStart = target; 551 return result; 552 } 553 554 /* --------------------------------------------------------------------- 555 556 Note A. 557 The fall-through switches in UTF-8 reading code save a 558 temp variable, some decrements & conditionals. The switches 559 are equivalent to the following loop: 560 { 561 int tmpBytesToRead = extraBytesToRead+1; 562 do { 563 ch += *source++; 564 --tmpBytesToRead; 565 if (tmpBytesToRead) ch <<= 6; 566 } while (tmpBytesToRead > 0); 567 } 568 In UTF-8 writing code, the switches on "bytesToWrite" are 569 similarly unrolled loops. 570 571 --------------------------------------------------------------------- */ 572