src/utf.c

   1 /*
   2 ** 2004 April 13
   3 **
   4 ** The author disclaims copyright to this source code.  In place of
   5 ** a legal notice, here is a blessing:
   6 **
   7 **    May you do good and not evil.
   8 **    May you find forgiveness for yourself and forgive others.
   9 **    May you share freely, never taking more than you give.
  10 **
  11 *************************************************************************
  12 ** This file contains routines used to translate between UTF-8,
  13 ** UTF-16, UTF-16BE, and UTF-16LE.
  14 **
  15 ** Notes on UTF-8:
  16 **
  17 **   Byte-0    Byte-1    Byte-2    Byte-3    Value
  18 **  0xxxxxxx                                 00000000 00000000 0xxxxxxx
  19 **  110yyyyy  10xxxxxx                       00000000 00000yyy yyxxxxxx
  20 **  1110zzzz  10yyyyyy  10xxxxxx             00000000 zzzzyyyy yyxxxxxx
  21 **  11110uuu  10uuzzzz  10yyyyyy  10xxxxxx   000uuuuu zzzzyyyy yyxxxxxx
  22 **
  23 **
  24 ** Notes on UTF-16:  (with wwww+1==uuuuu)
  25 **
  26 **      Word-0               Word-1          Value
  27 **  110110ww wwzzzzyy   110111yy yyxxxxxx    000uuuuu zzzzyyyy yyxxxxxx
  28 **  zzzzyyyy yyxxxxxx                        00000000 zzzzyyyy yyxxxxxx
  29 **
  30 **
  31 ** BOM or Byte Order Mark:
  32 **     0xff 0xfe   little-endian utf-16 follows
  33 **     0xfe 0xff   big-endian utf-16 follows
  34 **
  35 */
  36 #include "sqliteInt.h"
  37 #include <assert.h>
  38 #include "vdbeInt.h"
  39
  40 #if !defined(SQLITE_AMALGAMATION) && SQLITE_BYTEORDER==0
  41 /*
  42 ** The following constant value is used by the SQLITE_BIGENDIAN and
  43 ** SQLITE_LITTLEENDIAN macros.
  44 */
  45 const int sqlite3one = 1;
  46 #endif /* SQLITE_AMALGAMATION && SQLITE_BYTEORDER==0 */
  47
  48 /*
  49 ** This lookup table is used to help decode the first byte of
  50 ** a multi-byte UTF8 character.
  51 */
  52 static const unsigned char sqlite3Utf8Trans1[] = {
  53   0x00, 0x01, 0x02, 0x03, 0x04, 0x05, 0x06, 0x07,
  54   0x08, 0x09, 0x0a, 0x0b, 0x0c, 0x0d, 0x0e, 0x0f,
  55   0x10, 0x11, 0x12, 0x13, 0x14, 0x15, 0x16, 0x17,
  56   0x18, 0x19, 0x1a, 0x1b, 0x1c, 0x1d, 0x1e, 0x1f,
  57   0x00, 0x01, 0x02, 0x03, 0x04, 0x05, 0x06, 0x07,
  58   0x08, 0x09, 0x0a, 0x0b, 0x0c, 0x0d, 0x0e, 0x0f,
  59   0x00, 0x01, 0x02, 0x03, 0x04, 0x05, 0x06, 0x07,
  60   0x00, 0x01, 0x02, 0x03, 0x00, 0x01, 0x00, 0x00,
  61 };
  62
  63
  64 #define WRITE_UTF8(zOut, c) {                          \
  65   if( c<0x00080 ){                                     \
  66     *zOut++ = (u8)(c&0xFF);                            \
  67   }                                                    \
  68   else if( c<0x00800 ){                                \
  69     *zOut++ = 0xC0 + (u8)((c>>6)&0x1F);                \
  70     *zOut++ = 0x80 + (u8)(c & 0x3F);                   \
  71   }                                                    \
  72   else if( c<0x10000 ){                                \
  73     *zOut++ = 0xE0 + (u8)((c>>12)&0x0F);               \
  74     *zOut++ = 0x80 + (u8)((c>>6) & 0x3F);              \
  75     *zOut++ = 0x80 + (u8)(c & 0x3F);                   \
  76   }else{                                               \
  77     *zOut++ = 0xF0 + (u8)((c>>18) & 0x07);             \
  78     *zOut++ = 0x80 + (u8)((c>>12) & 0x3F);             \
  79     *zOut++ = 0x80 + (u8)((c>>6) & 0x3F);              \
  80     *zOut++ = 0x80 + (u8)(c & 0x3F);                   \
  81   }                                                    \
  82 }
  83
  84 #define WRITE_UTF16LE(zOut, c) {                                    \
  85   if( c<=0xFFFF ){                                                  \
  86     *zOut++ = (u8)(c&0x00FF);                                       \
  87     *zOut++ = (u8)((c>>8)&0x00FF);                                  \
  88   }else{                                                            \
  89     *zOut++ = (u8)(((c>>10)&0x003F) + (((c-0x10000)>>10)&0x00C0));  \
  90     *zOut++ = (u8)(0x00D8 + (((c-0x10000)>>18)&0x03));              \
  91     *zOut++ = (u8)(c&0x00FF);                                       \
  92     *zOut++ = (u8)(0x00DC + ((c>>8)&0x03));                         \
  93   }                                                                 \
  94 }
  95
  96 #define WRITE_UTF16BE(zOut, c) {                                    \
  97   if( c<=0xFFFF ){                                                  \
  98     *zOut++ = (u8)((c>>8)&0x00FF);                                  \
  99     *zOut++ = (u8)(c&0x00FF);                                       \
 100   }else{                                                            \
 101     *zOut++ = (u8)(0x00D8 + (((c-0x10000)>>18)&0x03));              \
 102     *zOut++ = (u8)(((c>>10)&0x003F) + (((c-0x10000)>>10)&0x00C0));  \
 103     *zOut++ = (u8)(0x00DC + ((c>>8)&0x03));                         \
 104     *zOut++ = (u8)(c&0x00FF);                                       \
 105   }                                                                 \
 106 }
 107
 108 /*
 109 ** Translate a single UTF-8 character.  Return the unicode value.
 110 **
 111 ** During translation, assume that the byte that zTerm points
 112 ** is a 0x00.
 113 **
 114 ** Write a pointer to the next unread byte back into *pzNext.
 115 **
 116 ** Notes On Invalid UTF-8:
 117 **
 118 **  *  This routine never allows a 7-bit character (0x00 through 0x7f) to
 119 **     be encoded as a multi-byte character.  Any multi-byte character that
 120 **     attempts to encode a value between 0x00 and 0x7f is rendered as 0xfffd.
 121 **
 122 **  *  This routine never allows a UTF16 surrogate value to be encoded.
 123 **     If a multi-byte character attempts to encode a value between
 124 **     0xd800 and 0xe000 then it is rendered as 0xfffd.
 125 **
 126 **  *  Bytes in the range of 0x80 through 0xbf which occur as the first
 127 **     byte of a character are interpreted as single-byte characters
 128 **     and rendered as themselves even though they are technically
 129 **     invalid characters.
 130 **
 131 **  *  This routine accepts over-length UTF8 encodings
 132 **     for unicode values 0x80 and greater.  It does not change over-length
 133 **     encodings to 0xfffd as some systems recommend.
 134 */
 135 #define READ_UTF8(zIn, zTerm, c)                           \
 136   c = *(zIn++);                                            \
 137   if( c>=0xc0 ){                                           \
 138     c = sqlite3Utf8Trans1[c-0xc0];                         \
 139     while( zIn!=zTerm && (*zIn & 0xc0)==0x80 ){            \
 140       c = (c<<6) + (0x3f & *(zIn++));                      \
 141     }                                                      \
 142     if( c<0x80                                             \
 143         || (c&0xFFFFF800)==0xD800                          \
 144         || (c&0xFFFFFFFE)==0xFFFE ){  c = 0xFFFD; }        \
 145   }
 146 u32 sqlite3Utf8Read(
 147   const unsigned char **pz    /* Pointer to string from which to read char */
 148 ){
 149   unsigned int c;
 150
 151   /* Same as READ_UTF8() above but without the zTerm parameter.
 152   ** For this routine, we assume the UTF8 string is always zero-terminated.
 153   */
 154   c = *((*pz)++);
 155   if( c>=0xc0 ){
 156     c = sqlite3Utf8Trans1[c-0xc0];
 157     while( (*(*pz) & 0xc0)==0x80 ){
 158       c = (c<<6) + (0x3f & *((*pz)++));
 159     }
 160     if( c<0x80
 161         || (c&0xFFFFF800)==0xD800
 162         || (c&0xFFFFFFFE)==0xFFFE ){  c = 0xFFFD; }
 163   }
 164   return c;
 165 }
 166
 167
 168
 169
 170 /*
 171 ** If the TRANSLATE_TRACE macro is defined, the value of each Mem is
 172 ** printed on stderr on the way into and out of sqlite3VdbeMemTranslate().
 173 */
 174 /* #define TRANSLATE_TRACE 1 */
 175
 176 #ifndef SQLITE_OMIT_UTF16
 177 /*
 178 ** This routine transforms the internal text encoding used by pMem to
 179 ** desiredEnc. It is an error if the string is already of the desired
 180 ** encoding, or if *pMem does not contain a string value.
 181 */
 182 SQLITE_NOINLINE int sqlite3VdbeMemTranslate(Mem *pMem, u8 desiredEnc){
 183   sqlite3_int64 len;          /* Maximum length of output string in bytes */
 184   unsigned char *zOut;        /* Output buffer */
 185   unsigned char *zIn;         /* Input iterator */
 186   unsigned char *zTerm;       /* End of input */
 187   unsigned char *z;           /* Output iterator */
 188   unsigned int c;
 189
 190   assert( pMem->db==0 || sqlite3_mutex_held(pMem->db->mutex) );
 191   assert( pMem->flags&MEM_Str );
 192   assert( pMem->enc!=desiredEnc );
 193   assert( pMem->enc!=0 );
 194   assert( pMem->n>=0 );
 195
 196 #if defined(TRANSLATE_TRACE) && defined(SQLITE_DEBUG)
 197   {
 198     StrAccum acc;
 199     char zBuf[1000];
 200     sqlite3StrAccumInit(&acc, 0, zBuf, sizeof(zBuf), 0);
 201     sqlite3VdbeMemPrettyPrint(pMem, &acc);
 202     fprintf(stderr, "INPUT:  %s\n", sqlite3StrAccumFinish(&acc));
 203   }
 204 #endif
 205
 206   /* If the translation is between UTF-16 little and big endian, then
 207   ** all that is required is to swap the byte order. This case is handled
 208   ** differently from the others.
 209   */
 210   if( pMem->enc!=SQLITE_UTF8 && desiredEnc!=SQLITE_UTF8 ){
 211     u8 temp;
 212     int rc;
 213     rc = sqlite3VdbeMemMakeWriteable(pMem);
 214     if( rc!=SQLITE_OK ){
 215       assert( rc==SQLITE_NOMEM );
 216       return SQLITE_NOMEM_BKPT;
 217     }
 218     zIn = (u8*)pMem->z;
 219     zTerm = &zIn[pMem->n&~1];
 220     while( zIn<zTerm ){
 221       temp = *zIn;
 222       *zIn = *(zIn+1);
 223       zIn++;
 224       *zIn++ = temp;
 225     }
 226     pMem->enc = desiredEnc;
 227     goto translate_out;
 228   }
 229
 230   /* Set len to the maximum number of bytes required in the output buffer. */
 231   if( desiredEnc==SQLITE_UTF8 ){
 232     /* When converting from UTF-16, the maximum growth results from
 233     ** translating a 2-byte character to a 4-byte UTF-8 character.
 234     ** A single byte is required for the output string
 235     ** nul-terminator.
 236     */
 237     pMem->n &= ~1;
 238     len = 2 * (sqlite3_int64)pMem->n + 1;
 239   }else{
 240     /* When converting from UTF-8 to UTF-16 the maximum growth is caused
 241     ** when a 1-byte UTF-8 character is translated into a 2-byte UTF-16
 242     ** character. Two bytes are required in the output buffer for the
 243     ** nul-terminator.
 244     */
 245     len = 2 * (sqlite3_int64)pMem->n + 2;
 246   }
 247
 248   /* Set zIn to point at the start of the input buffer and zTerm to point 1
 249   ** byte past the end.
 250   **
 251   ** Variable zOut is set to point at the output buffer, space obtained
 252   ** from sqlite3_malloc().
 253   */
 254   zIn = (u8*)pMem->z;
 255   zTerm = &zIn[pMem->n];
 256   zOut = sqlite3DbMallocRaw(pMem->db, len);
 257   if( !zOut ){
 258     return SQLITE_NOMEM_BKPT;
 259   }
 260   z = zOut;
 261
 262   if( pMem->enc==SQLITE_UTF8 ){
 263     if( desiredEnc==SQLITE_UTF16LE ){
 264       /* UTF-8 -> UTF-16 Little-endian */
 265       while( zIn<zTerm ){
 266         READ_UTF8(zIn, zTerm, c);
 267         WRITE_UTF16LE(z, c);
 268       }
 269     }else{
 270       assert( desiredEnc==SQLITE_UTF16BE );
 271       /* UTF-8 -> UTF-16 Big-endian */
 272       while( zIn<zTerm ){
 273         READ_UTF8(zIn, zTerm, c);
 274         WRITE_UTF16BE(z, c);
 275       }
 276     }
 277     pMem->n = (int)(z - zOut);
 278     *z++ = 0;
 279   }else{
 280     assert( desiredEnc==SQLITE_UTF8 );
 281     if( pMem->enc==SQLITE_UTF16LE ){
 282       /* UTF-16 Little-endian -> UTF-8 */
 283       while( zIn<zTerm ){
 284         c = *(zIn++);
 285         c += (*(zIn++))<<8;
 286         if( c>=0xd800 && c<0xe000 ){
 287 #ifdef SQLITE_REPLACE_INVALID_UTF
 288           if( c>=0xdc00 || zIn>=zTerm ){
 289             c = 0xfffd;
 290           }else{
 291             int c2 = *(zIn++);
 292             c2 += (*(zIn++))<<8;
 293             if( c2<0xdc00 || c2>=0xe000 ){
 294               zIn -= 2;
 295               c = 0xfffd;
 296             }else{
 297               c = ((c&0x3ff)<<10) + (c2&0x3ff) + 0x10000;
 298             }
 299           }
 300 #else
 301           if( zIn<zTerm ){
 302             int c2 = (*zIn++);
 303             c2 += ((*zIn++)<<8);
 304             c = (c2&0x03FF) + ((c&0x003F)<<10) + (((c&0x03C0)+0x0040)<<10);
 305           }
 306 #endif
 307         }
 308         WRITE_UTF8(z, c);
 309       }
 310     }else{
 311       /* UTF-16 Big-endian -> UTF-8 */
 312       while( zIn<zTerm ){
 313         c = (*(zIn++))<<8;
 314         c += *(zIn++);
 315         if( c>=0xd800 && c<0xe000 ){
 316 #ifdef SQLITE_REPLACE_INVALID_UTF
 317           if( c>=0xdc00 || zIn>=zTerm ){
 318             c = 0xfffd;
 319           }else{
 320             int c2 = (*(zIn++))<<8;
 321             c2 += *(zIn++);
 322             if( c2<0xdc00 || c2>=0xe000 ){
 323               zIn -= 2;
 324               c = 0xfffd;
 325             }else{
 326               c = ((c&0x3ff)<<10) + (c2&0x3ff) + 0x10000;
 327             }
 328           }
 329 #else
 330           if( zIn<zTerm ){
 331             int c2 = ((*zIn++)<<8);
 332             c2 += (*zIn++);
 333             c = (c2&0x03FF) + ((c&0x003F)<<10) + (((c&0x03C0)+0x0040)<<10);
 334           }
 335 #endif
 336         }
 337         WRITE_UTF8(z, c);
 338       }
 339     }
 340     pMem->n = (int)(z - zOut);
 341   }
 342   *z = 0;
 343   assert( (pMem->n+(desiredEnc==SQLITE_UTF8?1:2))<=len );
 344
 345   c = MEM_Str|MEM_Term|(pMem->flags&(MEM_AffMask|MEM_Subtype));
 346   sqlite3VdbeMemRelease(pMem);
 347   pMem->flags = c;
 348   pMem->enc = desiredEnc;
 349   pMem->z = (char*)zOut;
 350   pMem->zMalloc = pMem->z;
 351   pMem->szMalloc = sqlite3DbMallocSize(pMem->db, pMem->z);
 352
 353 translate_out:
 354 #if defined(TRANSLATE_TRACE) && defined(SQLITE_DEBUG)
 355   {
 356     StrAccum acc;
 357     char zBuf[1000];
 358     sqlite3StrAccumInit(&acc, 0, zBuf, sizeof(zBuf), 0);
 359     sqlite3VdbeMemPrettyPrint(pMem, &acc);
 360     fprintf(stderr, "OUTPUT: %s\n", sqlite3StrAccumFinish(&acc));
 361   }
 362 #endif
 363   return SQLITE_OK;
 364 }
 365 #endif /* SQLITE_OMIT_UTF16 */
 366
 367 #ifndef SQLITE_OMIT_UTF16
 368 /*
 369 ** This routine checks for a byte-order mark at the beginning of the
 370 ** UTF-16 string stored in *pMem. If one is present, it is removed and
 371 ** the encoding of the Mem adjusted. This routine does not do any
 372 ** byte-swapping, it just sets Mem.enc appropriately.
 373 **
 374 ** The allocation (static, dynamic etc.) and encoding of the Mem may be
 375 ** changed by this function.
 376 */
 377 int sqlite3VdbeMemHandleBom(Mem *pMem){
 378   int rc = SQLITE_OK;
 379   u8 bom = 0;
 380
 381   assert( pMem->n>=0 );
 382   if( pMem->n>1 ){
 383     u8 b1 = *(u8 *)pMem->z;
 384     u8 b2 = *(((u8 *)pMem->z) + 1);
 385     if( b1==0xFE && b2==0xFF ){
 386       bom = SQLITE_UTF16BE;
 387     }
 388     if( b1==0xFF && b2==0xFE ){
 389       bom = SQLITE_UTF16LE;
 390     }
 391   }
 392
 393   if( bom ){
 394     rc = sqlite3VdbeMemMakeWriteable(pMem);
 395     if( rc==SQLITE_OK ){
 396       pMem->n -= 2;
 397       memmove(pMem->z, &pMem->z[2], pMem->n);
 398       pMem->z[pMem->n] = '\0';
 399       pMem->z[pMem->n+1] = '\0';
 400       pMem->flags |= MEM_Term;
 401       pMem->enc = bom;
 402     }
 403   }
 404   return rc;
 405 }
 406 #endif /* SQLITE_OMIT_UTF16 */
 407
 408 /*
 409 ** pZ is a UTF-8 encoded unicode string. If nByte is less than zero,
 410 ** return the number of unicode characters in pZ up to (but not including)
 411 ** the first 0x00 byte. If nByte is not less than zero, return the
 412 ** number of unicode characters in the first nByte of pZ (or up to
 413 ** the first 0x00, whichever comes first).
 414 */
 415 int sqlite3Utf8CharLen(const char *zIn, int nByte){
 416   int r = 0;
 417   const u8 *z = (const u8*)zIn;
 418   const u8 *zTerm;
 419   if( nByte>=0 ){
 420     zTerm = &z[nByte];
 421   }else{
 422     zTerm = (const u8*)(-1);
 423   }
 424   assert( z<=zTerm );
 425   while( *z!=0 && z<zTerm ){
 426     SQLITE_SKIP_UTF8(z);
 427     r++;
 428   }
 429   return r;
 430 }
 431
 432 /* This test function is not currently used by the automated test-suite.
 433 ** Hence it is only available in debug builds.
 434 */
 435 #if defined(SQLITE_TEST) && defined(SQLITE_DEBUG)
 436 /*
 437 ** Translate UTF-8 to UTF-8.
 438 **
 439 ** This has the effect of making sure that the string is well-formed
 440 ** UTF-8.  Miscoded characters are removed.
 441 **
 442 ** The translation is done in-place and aborted if the output
 443 ** overruns the input.
 444 */
 445 int sqlite3Utf8To8(unsigned char *zIn){
 446   unsigned char *zOut = zIn;
 447   unsigned char *zStart = zIn;
 448   u32 c;
 449
 450   while( zIn[0] && zOut<=zIn ){
 451     c = sqlite3Utf8Read((const u8**)&zIn);
 452     if( c!=0xfffd ){
 453       WRITE_UTF8(zOut, c);
 454     }
 455   }
 456   *zOut = 0;
 457   return (int)(zOut - zStart);
 458 }
 459 #endif
 460
 461 #ifndef SQLITE_OMIT_UTF16
 462 /*
 463 ** Convert a UTF-16 string in the native encoding into a UTF-8 string.
 464 ** Memory to hold the UTF-8 string is obtained from sqlite3_malloc and must
 465 ** be freed by the calling function.
 466 **
 467 ** NULL is returned if there is an allocation error.
 468 */
 469 char *sqlite3Utf16to8(sqlite3 *db, const void *z, int nByte, u8 enc){
 470   Mem m;
 471   memset(&m, 0, sizeof(m));
 472   m.db = db;
 473   sqlite3VdbeMemSetStr(&m, z, nByte, enc, SQLITE_STATIC);
 474   sqlite3VdbeChangeEncoding(&m, SQLITE_UTF8);
 475   if( db->mallocFailed ){
 476     sqlite3VdbeMemRelease(&m);
 477     m.z = 0;
 478   }
 479   assert( (m.flags & MEM_Term)!=0 || db->mallocFailed );
 480   assert( (m.flags & MEM_Str)!=0 || db->mallocFailed );
 481   assert( m.z || db->mallocFailed );
 482   return m.z;
 483 }
 484
 485 /*
 486 ** zIn is a UTF-16 encoded unicode string at least nChar characters long.
 487 ** Return the number of bytes in the first nChar unicode characters
 488 ** in pZ.  nChar must be non-negative.
 489 */
 490 int sqlite3Utf16ByteLen(const void *zIn, int nChar){
 491   int c;
 492   unsigned char const *z = zIn;
 493   int n = 0;
 494
 495   if( SQLITE_UTF16NATIVE==SQLITE_UTF16LE ) z++;
 496   while( n<nChar ){
 497     c = z[0];
 498     z += 2;
 499     if( c>=0xd8 && c<0xdc && z[0]>=0xdc && z[0]<0xe0 ) z += 2;
 500     n++;
 501   }
 502   return (int)(z-(unsigned char const *)zIn)
 503               - (SQLITE_UTF16NATIVE==SQLITE_UTF16LE);
 504 }
 505
 506 #if defined(SQLITE_TEST)
 507 /*
 508 ** This routine is called from the TCL test function "translate_selftest".
 509 ** It checks that the primitives for serializing and deserializing
 510 ** characters in each encoding are inverses of each other.
 511 */
 512 void sqlite3UtfSelfTest(void){
 513   unsigned int i, t;
 514   unsigned char zBuf[20];
 515   unsigned char *z;
 516   int n;
 517   unsigned int c;
 518
 519   for(i=0; i<0x00110000; i++){
 520     z = zBuf;
 521     WRITE_UTF8(z, i);
 522     n = (int)(z-zBuf);
 523     assert( n>0 && n<=4 );
 524     z[0] = 0;
 525     z = zBuf;
 526     c = sqlite3Utf8Read((const u8**)&z);
 527     t = i;
 528     if( i>=0xD800 && i<=0xDFFF ) t = 0xFFFD;
 529     if( (i&0xFFFFFFFE)==0xFFFE ) t = 0xFFFD;
 530     assert( c==t );
 531     assert( (z-zBuf)==n );
 532   }
 533 }
 534 #endif /* SQLITE_TEST */
 535 #endif /* SQLITE_OMIT_UTF16 */