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