1 /* 2 * Copyright 2001-2004 Unicode, Inc. 3 * 4 * Disclaimer 5 * 6 * This source code is provided as is by Unicode, Inc. No claims are 7 * made as to fitness for any particular purpose. No warranties of any 8 * kind are expressed or implied. The recipient agrees to determine 9 * applicability of information provided. If this file has been 10 * purchased on magnetic or optical media from Unicode, Inc., the 11 * sole remedy for any claim will be exchange of defective media 12 * within 90 days of receipt. 13 * 14 * Limitations on Rights to Redistribute This Code 15 * 16 * Unicode, Inc. hereby grants the right to freely use the information 17 * supplied in this file in the creation of products supporting the 18 * Unicode Standard, and to make copies of this file in any form 19 * for internal or external distribution as long as this notice 20 * remains attached. 21 */ 22 23 /* --------------------------------------------------------------------- 24 25 Conversions between UTF32, UTF-16, and UTF-8. Source code file. 26 Author: Mark E. Davis, 1994. 27 Rev History: Rick McGowan, fixes & updates May 2001. 28 Sept 2001: fixed const & error conditions per 29 mods suggested by S. Parent & A. Lillich. 30 June 2002: Tim Dodd added detection and handling of incomplete 31 source sequences, enhanced error detection, added casts 32 to eliminate compiler warnings. 33 July 2003: slight mods to back out aggressive FFFE detection. 34 Jan 2004: updated switches in from-UTF8 conversions. 35 Oct 2004: updated to use UNI_MAX_LEGAL_UTF32 in UTF-32 conversions. 36 37 See the header file "ConvertUTF.h" for complete documentation. 38 39 ------------------------------------------------------------------------ */ 40 41 42 #include "ConvertUTF.h" 43 #ifdef CVTUTF_DEBUG 44 #include <stdio.h> 45 #endif 46 47 static const int halfShift = 10; /* used for shifting by 10 bits */ 48 49 static const UTF32 halfBase = 0x0010000UL; 50 static const UTF32 halfMask = 0x3FFUL; 51 52 #define UNI_SUR_HIGH_START (UTF32)0xD800 53 #define UNI_SUR_HIGH_END (UTF32)0xDBFF 54 #define UNI_SUR_LOW_START (UTF32)0xDC00 55 #define UNI_SUR_LOW_END (UTF32)0xDFFF 56 #define false 0 57 #define true 1 58 59 /* --------------------------------------------------------------------- */ 60 61 ConversionResult ConvertUTF32toUTF16 ( 62 const UTF32** sourceStart, const UTF32* sourceEnd, 63 UTF16** targetStart, UTF16* targetEnd, ConversionFlags flags) { 64 ConversionResult result = conversionOK; 65 const UTF32* source = *sourceStart; 66 UTF16* target = *targetStart; 67 while (source < sourceEnd) { 68 UTF32 ch; 69 if (target >= targetEnd) { 70 result = targetExhausted; break; 71 } 72 ch = *source++; 73 if (ch <= UNI_MAX_BMP) { /* Target is a character <= 0xFFFF */ 74 /* UTF-16 surrogate values are illegal in UTF-32; 0xffff or 0xfffe are both reserved values */ 75 if (ch >= UNI_SUR_HIGH_START && ch <= UNI_SUR_LOW_END) { 76 if (flags == strictConversion) { 77 --source; /* return to the illegal value itself */ 78 result = sourceIllegal; 79 break; 80 } else { 81 *target++ = UNI_REPLACEMENT_CHAR; 82 } 83 } else { 84 *target++ = (UTF16)ch; /* normal case */ 85 } 86 } else if (ch > UNI_MAX_LEGAL_UTF32) { 87 if (flags == strictConversion) { 88 result = sourceIllegal; 89 } else { 90 *target++ = UNI_REPLACEMENT_CHAR; 91 } 92 } else { 93 /* target is a character in range 0xFFFF - 0x10FFFF. */ 94 if (target + 1 >= targetEnd) { 95 --source; /* Back up source pointer! */ 96 result = targetExhausted; break; 97 } 98 ch -= halfBase; 99 *target++ = (UTF16)((ch >> halfShift) + UNI_SUR_HIGH_START); 100 *target++ = (UTF16)((ch & halfMask) + UNI_SUR_LOW_START); 101 } 102 } 103 *sourceStart = source; 104 *targetStart = target; 105 return result; 106 } 107 108 /* --------------------------------------------------------------------- */ 109 110 ConversionResult ConvertUTF16toUTF32 ( 111 const UTF16** sourceStart, const UTF16* sourceEnd, 112 UTF32** targetStart, UTF32* targetEnd, ConversionFlags flags) { 113 ConversionResult result = conversionOK; 114 const UTF16* source = *sourceStart; 115 UTF32* target = *targetStart; 116 UTF32 ch, ch2; 117 while (source < sourceEnd) { 118 const UTF16* oldSource = source; /* In case we have to back up because of target overflow. */ 119 ch = *source++; 120 /* If we have a surrogate pair, convert to UTF32 first. */ 121 if (ch >= UNI_SUR_HIGH_START && ch <= UNI_SUR_HIGH_END) { 122 /* If the 16 bits following the high surrogate are in the source buffer... */ 123 if (source < sourceEnd) { 124 ch2 = *source; 125 /* If it's a low surrogate, convert to UTF32. */ 126 if (ch2 >= UNI_SUR_LOW_START && ch2 <= UNI_SUR_LOW_END) { 127 ch = ((ch - UNI_SUR_HIGH_START) << halfShift) 128 + (ch2 - UNI_SUR_LOW_START) + halfBase; 129 ++source; 130 } else if (flags == strictConversion) { /* it's an unpaired high surrogate */ 131 --source; /* return to the illegal value itself */ 132 result = sourceIllegal; 133 break; 134 } 135 } else { /* We don't have the 16 bits following the high surrogate. */ 136 --source; /* return to the high surrogate */ 137 result = sourceExhausted; 138 break; 139 } 140 } else if (flags == strictConversion) { 141 /* UTF-16 surrogate values are illegal in UTF-32 */ 142 if (ch >= UNI_SUR_LOW_START && ch <= UNI_SUR_LOW_END) { 143 --source; /* return to the illegal value itself */ 144 result = sourceIllegal; 145 break; 146 } 147 } 148 if (target >= targetEnd) { 149 source = oldSource; /* Back up source pointer! */ 150 result = targetExhausted; break; 151 } 152 *target++ = ch; 153 } 154 *sourceStart = source; 155 *targetStart = target; 156 #ifdef CVTUTF_DEBUG 157 if (result == sourceIllegal) { 158 fprintf(stderr, "ConvertUTF16toUTF32 illegal seq 0x%04x,%04x\n", ch, ch2); 159 fflush(stderr); 160 } 161 #endif 162 return result; 163 } 164 165 /* --------------------------------------------------------------------- */ 166 167 /* 168 * Index into the table below with the first byte of a UTF-8 sequence to 169 * get the number of trailing bytes that are supposed to follow it. 170 * Note that *legal* UTF-8 values can't have 4 or 5-bytes. The table is 171 * left as-is for anyone who may want to do such conversion, which was 172 * allowed in earlier algorithms. 173 */ 174 static const char trailingBytesForUTF8[256] = { 175 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, 176 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, 177 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, 178 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, 179 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, 180 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, 181 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, 182 2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2, 3,3,3,3,3,3,3,3,4,4,4,4,5,5,5,5 183 }; 184 185 /* 186 * Magic values subtracted from a buffer value during UTF8 conversion. 187 * This table contains as many values as there might be trailing bytes 188 * in a UTF-8 sequence. 189 */ 190 static const UTF32 offsetsFromUTF8[6] = { 0x00000000UL, 0x00003080UL, 0x000E2080UL, 191 0x03C82080UL, 0xFA082080UL, 0x82082080UL }; 192 193 /* 194 * Once the bits are split out into bytes of UTF-8, this is a mask OR-ed 195 * into the first byte, depending on how many bytes follow. There are 196 * as many entries in this table as there are UTF-8 sequence types. 197 * (I.e., one byte sequence, two byte... etc.). Remember that sequencs 198 * for *legal* UTF-8 will be 4 or fewer bytes total. 199 */ 200 static const UTF8 firstByteMark[7] = { 0x00, 0x00, 0xC0, 0xE0, 0xF0, 0xF8, 0xFC }; 201 202 /* --------------------------------------------------------------------- */ 203 204 /* The interface converts a whole buffer to avoid function-call overhead. 205 * Constants have been gathered. Loops & conditionals have been removed as 206 * much as possible for efficiency, in favor of drop-through switches. 207 * (See "Note A" at the bottom of the file for equivalent code.) 208 * If your compiler supports it, the "isLegalUTF8" call can be turned 209 * into an inline function. 210 */ 211 212 /* --------------------------------------------------------------------- */ 213 214 ConversionResult ConvertUTF16toUTF8 ( 215 const UTF16** sourceStart, const UTF16* sourceEnd, 216 UTF8** targetStart, UTF8* targetEnd, ConversionFlags flags) { 217 ConversionResult result = conversionOK; 218 const UTF16* source = *sourceStart; 219 UTF8* target = *targetStart; 220 while (source < sourceEnd) { 221 UTF32 ch; 222 unsigned short bytesToWrite = 0; 223 const UTF32 byteMask = 0xBF; 224 const UTF32 byteMark = 0x80; 225 const UTF16* oldSource = source; /* In case we have to back up because of target overflow. */ 226 ch = *source++; 227 /* If we have a surrogate pair, convert to UTF32 first. */ 228 if (ch >= UNI_SUR_HIGH_START && ch <= UNI_SUR_HIGH_END) { 229 /* If the 16 bits following the high surrogate are in the source buffer... */ 230 if (source < sourceEnd) { 231 UTF32 ch2 = *source; 232 /* If it's a low surrogate, convert to UTF32. */ 233 if (ch2 >= UNI_SUR_LOW_START && ch2 <= UNI_SUR_LOW_END) { 234 ch = ((ch - UNI_SUR_HIGH_START) << halfShift) 235 + (ch2 - UNI_SUR_LOW_START) + halfBase; 236 ++source; 237 } else if (flags == strictConversion) { /* it's an unpaired high surrogate */ 238 --source; /* return to the illegal value itself */ 239 result = sourceIllegal; 240 break; 241 } 242 } else { /* We don't have the 16 bits following the high surrogate. */ 243 --source; /* return to the high surrogate */ 244 result = sourceExhausted; 245 break; 246 } 247 } else if (flags == strictConversion) { 248 /* UTF-16 surrogate values are illegal in UTF-32 */ 249 if (ch >= UNI_SUR_LOW_START && ch <= UNI_SUR_LOW_END) { 250 --source; /* return to the illegal value itself */ 251 result = sourceIllegal; 252 break; 253 } 254 } 255 256 // TPN: substitute all control characters except for NULL, TAB, LF or CR 257 if (ch && (ch != (UTF32)0x09) && (ch != (UTF32)0x0a) && (ch != (UTF32)0x0d) && (ch < (UTF32)0x20) ) { 258 ch = (UTF32)0x3f; 259 } 260 // TPN: filter out byte order marks and invalid character 0xFFFF 261 if((ch == (UTF32)0xFEFF) || (ch == (UTF32)0xFFFE)|| (ch == (UTF32)0xFFFF)) { 262 continue; 263 } 264 265 /* Figure out how many bytes the result will require */ 266 if (ch < (UTF32)0x80) { bytesToWrite = 1; 267 } else if (ch < (UTF32)0x800) { bytesToWrite = 2; 268 } else if (ch < (UTF32)0x10000) { bytesToWrite = 3; 269 } else if (ch < (UTF32)0x110000) { bytesToWrite = 4; 270 } else { bytesToWrite = 3; 271 ch = UNI_REPLACEMENT_CHAR; 272 } 273 274 target += bytesToWrite; 275 if (target > targetEnd) { 276 source = oldSource; /* Back up source pointer! */ 277 target -= bytesToWrite; result = targetExhausted; break; 278 } 279 switch (bytesToWrite) { /* note: everything falls through. */ 280 case 4: *--target = (UTF8)((ch | byteMark) & byteMask); ch >>= 6; 281 case 3: *--target = (UTF8)((ch | byteMark) & byteMask); ch >>= 6; 282 case 2: *--target = (UTF8)((ch | byteMark) & byteMask); ch >>= 6; 283 case 1: *--target = (UTF8)(ch | firstByteMark[bytesToWrite]); 284 } 285 target += bytesToWrite; 286 } 287 *sourceStart = source; 288 *targetStart = target; 289 return result; 290 } 291 292 /* --------------------------------------------------------------------- */ 293 294 /* 295 * Utility routine to tell whether a sequence of bytes is legal UTF-8. 296 * This must be called with the length pre-determined by the first byte. 297 * If not calling this from ConvertUTF8to*, then the length can be set by: 298 * length = trailingBytesForUTF8[*source]+1; 299 * and the sequence is illegal right away if there aren't that many bytes 300 * available. 301 * If presented with a length > 4, this returns false. The Unicode 302 * definition of UTF-8 goes up to 4-byte sequences. 303 */ 304 305 inline Boolean isLegalUTF8(const UTF8 *source, int length) { 306 UTF8 a; 307 const UTF8 *srcptr = source+length; 308 switch (length) { 309 default: return false; 310 /* Everything else falls through when "true"... */ 311 case 4: if ((a = (*--srcptr)) < 0x80 || a > 0xBF) return false; 312 case 3: if ((a = (*--srcptr)) < 0x80 || a > 0xBF) return false; 313 case 2: if ((a = (*--srcptr)) > 0xBF) return false; 314 315 switch (*source) { 316 /* no fall-through in this inner switch */ 317 case 0xE0: if (a < 0xA0) return false; break; 318 case 0xED: if (a > 0x9F) return false; break; 319 case 0xF0: if (a < 0x90) return false; break; 320 case 0xF4: if (a > 0x8F) return false; break; 321 default: if (a < 0x80) return false; 322 } 323 324 case 1: if (*source >= 0x80 && *source < 0xC2) return false; 325 } 326 if (*source > 0xF4) return false; 327 return true; 328 } 329 330 /* --------------------------------------------------------------------- */ 331 332 /* 333 * Exported function to return whether a UTF-8 sequence is legal or not. 334 * This is not used here; it's just exported. 335 */ 336 Boolean isLegalUTF8Sequence(const UTF8 *source, const UTF8 *sourceEnd) { 337 int length = trailingBytesForUTF8[*source]+1; 338 if (source+length > sourceEnd) { 339 return false; 340 } 341 return isLegalUTF8(source, length); 342 } 343 344 /* --------------------------------------------------------------------- */ 345 346 ConversionResult ConvertUTF8toUTF16 ( 347 const UTF8** sourceStart, const UTF8* sourceEnd, 348 UTF16** targetStart, UTF16* targetEnd, ConversionFlags flags) { 349 ConversionResult result = conversionOK; 350 const UTF8* source = *sourceStart; 351 UTF16* target = *targetStart; 352 while (source < sourceEnd) { 353 UTF32 ch = 0; 354 unsigned short extraBytesToRead = trailingBytesForUTF8[*source]; 355 if (source + extraBytesToRead >= sourceEnd) { 356 result = sourceExhausted; break; 357 } 358 /* Do this check whether lenient or strict */ 359 if (! isLegalUTF8(source, extraBytesToRead+1)) { 360 result = sourceIllegal; 361 break; 362 } 363 /* 364 * The cases all fall through. See "Note A" below. 365 */ 366 switch (extraBytesToRead) { 367 case 5: ch += *source++; ch <<= 6; /* remember, illegal UTF-8 */ 368 case 4: ch += *source++; ch <<= 6; /* remember, illegal UTF-8 */ 369 case 3: ch += *source++; ch <<= 6; 370 case 2: ch += *source++; ch <<= 6; 371 case 1: ch += *source++; ch <<= 6; 372 case 0: ch += *source++; 373 } 374 ch -= offsetsFromUTF8[extraBytesToRead]; 375 376 if (target >= targetEnd) { 377 source -= (extraBytesToRead+1); /* Back up source pointer! */ 378 result = targetExhausted; break; 379 } 380 if (ch <= UNI_MAX_BMP) { /* Target is a character <= 0xFFFF */ 381 /* UTF-16 surrogate values are illegal in UTF-32 */ 382 if (ch >= UNI_SUR_HIGH_START && ch <= UNI_SUR_LOW_END) { 383 if (flags == strictConversion) { 384 source -= (extraBytesToRead+1); /* return to the illegal value itself */ 385 result = sourceIllegal; 386 break; 387 } else { 388 *target++ = UNI_REPLACEMENT_CHAR; 389 } 390 } else { 391 *target++ = (UTF16)ch; /* normal case */ 392 } 393 } else if (ch > UNI_MAX_UTF16) { 394 if (flags == strictConversion) { 395 result = sourceIllegal; 396 source -= (extraBytesToRead+1); /* return to the start */ 397 break; /* Bail out; shouldn't continue */ 398 } else { 399 *target++ = UNI_REPLACEMENT_CHAR; 400 } 401 } else { 402 /* target is a character in range 0xFFFF - 0x10FFFF. */ 403 if (target + 1 >= targetEnd) { 404 source -= (extraBytesToRead+1); /* Back up source pointer! */ 405 result = targetExhausted; break; 406 } 407 ch -= halfBase; 408 *target++ = (UTF16)((ch >> halfShift) + UNI_SUR_HIGH_START); 409 *target++ = (UTF16)((ch & halfMask) + UNI_SUR_LOW_START); 410 } 411 } 412 *sourceStart = source; 413 *targetStart = target; 414 return result; 415 } 416 417 /* --------------------------------------------------------------------- */ 418 419 ConversionResult ConvertUTF32toUTF8 ( 420 const UTF32** sourceStart, const UTF32* sourceEnd, 421 UTF8** targetStart, UTF8* targetEnd, ConversionFlags flags) { 422 ConversionResult result = conversionOK; 423 const UTF32* source = *sourceStart; 424 UTF8* target = *targetStart; 425 while (source < sourceEnd) { 426 UTF32 ch; 427 unsigned short bytesToWrite = 0; 428 const UTF32 byteMask = 0xBF; 429 const UTF32 byteMark = 0x80; 430 ch = *source++; 431 if (flags == strictConversion ) { 432 /* UTF-16 surrogate values are illegal in UTF-32 */ 433 if (ch >= UNI_SUR_HIGH_START && ch <= UNI_SUR_LOW_END) { 434 --source; /* return to the illegal value itself */ 435 result = sourceIllegal; 436 break; 437 } 438 } 439 /* 440 * Figure out how many bytes the result will require. Turn any 441 * illegally large UTF32 things (> Plane 17) into replacement chars. 442 */ 443 if (ch < (UTF32)0x80) { bytesToWrite = 1; 444 } else if (ch < (UTF32)0x800) { bytesToWrite = 2; 445 } else if (ch < (UTF32)0x10000) { bytesToWrite = 3; 446 } else if (ch <= UNI_MAX_LEGAL_UTF32) { bytesToWrite = 4; 447 } else { bytesToWrite = 3; 448 ch = UNI_REPLACEMENT_CHAR; 449 result = sourceIllegal; 450 } 451 452 target += bytesToWrite; 453 if (target > targetEnd) { 454 --source; /* Back up source pointer! */ 455 target -= bytesToWrite; result = targetExhausted; break; 456 } 457 switch (bytesToWrite) { /* note: everything falls through. */ 458 case 4: *--target = (UTF8)((ch | byteMark) & byteMask); ch >>= 6; 459 case 3: *--target = (UTF8)((ch | byteMark) & byteMask); ch >>= 6; 460 case 2: *--target = (UTF8)((ch | byteMark) & byteMask); ch >>= 6; 461 case 1: *--target = (UTF8) (ch | firstByteMark[bytesToWrite]); 462 } 463 target += bytesToWrite; 464 } 465 *sourceStart = source; 466 *targetStart = target; 467 return result; 468 } 469 470 /* --------------------------------------------------------------------- */ 471 472 ConversionResult ConvertUTF8toUTF32 ( 473 const UTF8** sourceStart, const UTF8* sourceEnd, 474 UTF32** targetStart, UTF32* targetEnd, ConversionFlags flags) { 475 ConversionResult result = conversionOK; 476 const UTF8* source = *sourceStart; 477 UTF32* target = *targetStart; 478 while (source < sourceEnd) { 479 UTF32 ch = 0; 480 unsigned short extraBytesToRead = trailingBytesForUTF8[*source]; 481 if (source + extraBytesToRead >= sourceEnd) { 482 result = sourceExhausted; break; 483 } 484 /* Do this check whether lenient or strict */ 485 if (! isLegalUTF8(source, extraBytesToRead+1)) { 486 result = sourceIllegal; 487 break; 488 } 489 /* 490 * The cases all fall through. See "Note A" below. 491 */ 492 switch (extraBytesToRead) { 493 case 5: ch += *source++; ch <<= 6; 494 case 4: ch += *source++; ch <<= 6; 495 case 3: ch += *source++; ch <<= 6; 496 case 2: ch += *source++; ch <<= 6; 497 case 1: ch += *source++; ch <<= 6; 498 case 0: ch += *source++; 499 } 500 ch -= offsetsFromUTF8[extraBytesToRead]; 501 502 if (target >= targetEnd) { 503 source -= (extraBytesToRead+1); /* Back up the source pointer! */ 504 result = targetExhausted; break; 505 } 506 if (ch <= UNI_MAX_LEGAL_UTF32) { 507 /* 508 * UTF-16 surrogate values are illegal in UTF-32, and anything 509 * over Plane 17 (> 0x10FFFF) is illegal. 510 */ 511 if (ch >= UNI_SUR_HIGH_START && ch <= UNI_SUR_LOW_END) { 512 if (flags == strictConversion) { 513 source -= (extraBytesToRead+1); /* return to the illegal value itself */ 514 result = sourceIllegal; 515 break; 516 } else { 517 *target++ = UNI_REPLACEMENT_CHAR; 518 } 519 } else { 520 *target++ = ch; 521 } 522 } else { /* i.e., ch > UNI_MAX_LEGAL_UTF32 */ 523 result = sourceIllegal; 524 *target++ = UNI_REPLACEMENT_CHAR; 525 } 526 } 527 *sourceStart = source; 528 *targetStart = target; 529 return result; 530 } 531 532 /* --------------------------------------------------------------------- 533 534 Note A. 535 The fall-through switches in UTF-8 reading code save a 536 temp variable, some decrements & conditionals. The switches 537 are equivalent to the following loop: 538 { 539 int tmpBytesToRead = extraBytesToRead+1; 540 do { 541 ch += *source++; 542 --tmpBytesToRead; 543 if (tmpBytesToRead) ch <<= 6; 544 } while (tmpBytesToRead > 0); 545 } 546 In UTF-8 writing code, the switches on "bytesToWrite" are 547 similarly unrolled loops. 548 549 --------------------------------------------------------------------- */ 550