src/tokenize.c

   1 /*
   2 ** 2001 September 15
   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 ** An tokenizer for SQL
  13 **
  14 ** This file contains C code that splits an SQL input string up into
  15 ** individual tokens and sends those tokens one-by-one over to the
  16 ** parser for analysis.
  17 */
  18 #include "sqliteInt.h"
  19 #include <stdlib.h>
  20
  21 /* Character classes for tokenizing
  22 **
  23 ** In the sqlite3GetToken() function, a switch() on aiClass[c] is implemented
  24 ** using a lookup table, whereas a switch() directly on c uses a binary search.
  25 ** The lookup table is much faster.  To maximize speed, and to ensure that
  26 ** a lookup table is used, all of the classes need to be small integers and
  27 ** all of them need to be used within the switch.
  28 */
  29 #define CC_X          0    /* The letter 'x', or start of BLOB literal */
  30 #define CC_KYWD0      1    /* First letter of a keyword */
  31 #define CC_KYWD       2    /* Alphabetics or '_'.  Usable in a keyword */
  32 #define CC_DIGIT      3    /* Digits */
  33 #define CC_DOLLAR     4    /* '$' */
  34 #define CC_VARALPHA   5    /* '@', '#', ':'.  Alphabetic SQL variables */
  35 #define CC_VARNUM     6    /* '?'.  Numeric SQL variables */
  36 #define CC_SPACE      7    /* Space characters */
  37 #define CC_QUOTE      8    /* '"', '\'', or '`'.  String literals, quoted ids */
  38 #define CC_QUOTE2     9    /* '['.   [...] style quoted ids */
  39 #define CC_PIPE      10    /* '|'.   Bitwise OR or concatenate */
  40 #define CC_MINUS     11    /* '-'.  Minus or SQL-style comment */
  41 #define CC_LT        12    /* '<'.  Part of < or <= or <> */
  42 #define CC_GT        13    /* '>'.  Part of > or >= */
  43 #define CC_EQ        14    /* '='.  Part of = or == */
  44 #define CC_BANG      15    /* '!'.  Part of != */
  45 #define CC_SLASH     16    /* '/'.  / or c-style comment */
  46 #define CC_LP        17    /* '(' */
  47 #define CC_RP        18    /* ')' */
  48 #define CC_SEMI      19    /* ';' */
  49 #define CC_PLUS      20    /* '+' */
  50 #define CC_STAR      21    /* '*' */
  51 #define CC_PERCENT   22    /* '%' */
  52 #define CC_COMMA     23    /* ',' */
  53 #define CC_AND       24    /* '&' */
  54 #define CC_TILDA     25    /* '~' */
  55 #define CC_DOT       26    /* '.' */
  56 #define CC_ID        27    /* unicode characters usable in IDs */
  57 #define CC_ILLEGAL   28    /* Illegal character */
  58 #define CC_NUL       29    /* 0x00 */
  59 #define CC_BOM       30    /* First byte of UTF8 BOM:  0xEF 0xBB 0xBF */
  60
  61 static const unsigned char aiClass[] = {
  62 #ifdef SQLITE_ASCII
  63 /*         x0  x1  x2  x3  x4  x5  x6  x7  x8  x9  xa  xb  xc  xd  xe  xf */
  64 /* 0x */   29, 28, 28, 28, 28, 28, 28, 28, 28,  7,  7, 28,  7,  7, 28, 28,
  65 /* 1x */   28, 28, 28, 28, 28, 28, 28, 28, 28, 28, 28, 28, 28, 28, 28, 28,
  66 /* 2x */    7, 15,  8,  5,  4, 22, 24,  8, 17, 18, 21, 20, 23, 11, 26, 16,
  67 /* 3x */    3,  3,  3,  3,  3,  3,  3,  3,  3,  3,  5, 19, 12, 14, 13,  6,
  68 /* 4x */    5,  1,  1,  1,  1,  1,  1,  1,  1,  1,  1,  1,  1,  1,  1,  1,
  69 /* 5x */    1,  1,  1,  1,  1,  1,  1,  1,  0,  2,  2,  9, 28, 28, 28,  2,
  70 /* 6x */    8,  1,  1,  1,  1,  1,  1,  1,  1,  1,  1,  1,  1,  1,  1,  1,
  71 /* 7x */    1,  1,  1,  1,  1,  1,  1,  1,  0,  2,  2, 28, 10, 28, 25, 28,
  72 /* 8x */   27, 27, 27, 27, 27, 27, 27, 27, 27, 27, 27, 27, 27, 27, 27, 27,
  73 /* 9x */   27, 27, 27, 27, 27, 27, 27, 27, 27, 27, 27, 27, 27, 27, 27, 27,
  74 /* Ax */   27, 27, 27, 27, 27, 27, 27, 27, 27, 27, 27, 27, 27, 27, 27, 27,
  75 /* Bx */   27, 27, 27, 27, 27, 27, 27, 27, 27, 27, 27, 27, 27, 27, 27, 27,
  76 /* Cx */   27, 27, 27, 27, 27, 27, 27, 27, 27, 27, 27, 27, 27, 27, 27, 27,
  77 /* Dx */   27, 27, 27, 27, 27, 27, 27, 27, 27, 27, 27, 27, 27, 27, 27, 27,
  78 /* Ex */   27, 27, 27, 27, 27, 27, 27, 27, 27, 27, 27, 27, 27, 27, 27, 30,
  79 /* Fx */   27, 27, 27, 27, 27, 27, 27, 27, 27, 27, 27, 27, 27, 27, 27, 27
  80 #endif
  81 #ifdef SQLITE_EBCDIC
  82 /*         x0  x1  x2  x3  x4  x5  x6  x7  x8  x9  xa  xb  xc  xd  xe  xf */
  83 /* 0x */   29, 28, 28, 28, 28,  7, 28, 28, 28, 28, 28, 28,  7,  7, 28, 28,
  84 /* 1x */   28, 28, 28, 28, 28,  7, 28, 28, 28, 28, 28, 28, 28, 28, 28, 28,
  85 /* 2x */   28, 28, 28, 28, 28,  7, 28, 28, 28, 28, 28, 28, 28, 28, 28, 28,
  86 /* 3x */   28, 28, 28, 28, 28, 28, 28, 28, 28, 28, 28, 28, 28, 28, 28, 28,
  87 /* 4x */    7, 28, 28, 28, 28, 28, 28, 28, 28, 28, 28, 26, 12, 17, 20, 10,
  88 /* 5x */   24, 28, 28, 28, 28, 28, 28, 28, 28, 28, 15,  4, 21, 18, 19, 28,
  89 /* 6x */   11, 16, 28, 28, 28, 28, 28, 28, 28, 28, 28, 23, 22,  2, 13,  6,
  90 /* 7x */   28, 28, 28, 28, 28, 28, 28, 28, 28,  8,  5,  5,  5,  8, 14,  8,
  91 /* 8x */   28,  1,  1,  1,  1,  1,  1,  1,  1,  1, 28, 28, 28, 28, 28, 28,
  92 /* 9x */   28,  1,  1,  1,  1,  1,  1,  1,  1,  1, 28, 28, 28, 28, 28, 28,
  93 /* Ax */   28, 25,  1,  1,  1,  1,  1,  0,  2,  2, 28, 28, 28, 28, 28, 28,
  94 /* Bx */   28, 28, 28, 28, 28, 28, 28, 28, 28, 28,  9, 28, 28, 28, 28, 28,
  95 /* Cx */   28,  1,  1,  1,  1,  1,  1,  1,  1,  1, 28, 28, 28, 28, 28, 28,
  96 /* Dx */   28,  1,  1,  1,  1,  1,  1,  1,  1,  1, 28, 28, 28, 28, 28, 28,
  97 /* Ex */   28, 28,  1,  1,  1,  1,  1,  0,  2,  2, 28, 28, 28, 28, 28, 28,
  98 /* Fx */    3,  3,  3,  3,  3,  3,  3,  3,  3,  3, 28, 28, 28, 28, 28, 28,
  99 #endif
 100 };
 101
 102 /*
 103 ** The charMap() macro maps alphabetic characters (only) into their
 104 ** lower-case ASCII equivalent.  On ASCII machines, this is just
 105 ** an upper-to-lower case map.  On EBCDIC machines we also need
 106 ** to adjust the encoding.  The mapping is only valid for alphabetics
 107 ** which are the only characters for which this feature is used.
 108 **
 109 ** Used by keywordhash.h
 110 */
 111 #ifdef SQLITE_ASCII
 112 # define charMap(X) sqlite3UpperToLower[(unsigned char)X]
 113 #endif
 114 #ifdef SQLITE_EBCDIC
 115 # define charMap(X) ebcdicToAscii[(unsigned char)X]
 116 const unsigned char ebcdicToAscii[] = {
 117 /* 0   1   2   3   4   5   6   7   8   9   A   B   C   D   E   F */
 118    0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  /* 0x */
 119    0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  /* 1x */
 120    0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  /* 2x */
 121    0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  /* 3x */
 122    0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  /* 4x */
 123    0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  /* 5x */
 124    0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0, 95,  0,  0,  /* 6x */
 125    0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  /* 7x */
 126    0, 97, 98, 99,100,101,102,103,104,105,  0,  0,  0,  0,  0,  0,  /* 8x */
 127    0,106,107,108,109,110,111,112,113,114,  0,  0,  0,  0,  0,  0,  /* 9x */
 128    0,  0,115,116,117,118,119,120,121,122,  0,  0,  0,  0,  0,  0,  /* Ax */
 129    0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  /* Bx */
 130    0, 97, 98, 99,100,101,102,103,104,105,  0,  0,  0,  0,  0,  0,  /* Cx */
 131    0,106,107,108,109,110,111,112,113,114,  0,  0,  0,  0,  0,  0,  /* Dx */
 132    0,  0,115,116,117,118,119,120,121,122,  0,  0,  0,  0,  0,  0,  /* Ex */
 133    0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  /* Fx */
 134 };
 135 #endif
 136
 137 /*
 138 ** The sqlite3KeywordCode function looks up an identifier to determine if
 139 ** it is a keyword.  If it is a keyword, the token code of that keyword is
 140 ** returned.  If the input is not a keyword, TK_ID is returned.
 141 **
 142 ** The implementation of this routine was generated by a program,
 143 ** mkkeywordhash.c, located in the tool subdirectory of the distribution.
 144 ** The output of the mkkeywordhash.c program is written into a file
 145 ** named keywordhash.h and then included into this source file by
 146 ** the #include below.
 147 */
 148 #include "keywordhash.h"
 149
 150
 151 /*
 152 ** If X is a character that can be used in an identifier then
 153 ** IdChar(X) will be true.  Otherwise it is false.
 154 **
 155 ** For ASCII, any character with the high-order bit set is
 156 ** allowed in an identifier.  For 7-bit characters,
 157 ** sqlite3IsIdChar[X] must be 1.
 158 **
 159 ** For EBCDIC, the rules are more complex but have the same
 160 ** end result.
 161 **
 162 ** Ticket #1066.  the SQL standard does not allow '$' in the
 163 ** middle of identifiers.  But many SQL implementations do.
 164 ** SQLite will allow '$' in identifiers for compatibility.
 165 ** But the feature is undocumented.
 166 */
 167 #ifdef SQLITE_ASCII
 168 #define IdChar(C)  ((sqlite3CtypeMap[(unsigned char)C]&0x46)!=0)
 169 #endif
 170 #ifdef SQLITE_EBCDIC
 171 const char sqlite3IsEbcdicIdChar[] = {
 172 /* x0 x1 x2 x3 x4 x5 x6 x7 x8 x9 xA xB xC xD xE xF */
 173     0, 0, 1, 1, 1, 1, 1, 1, 1, 1, 0, 0, 0, 0, 0, 0,  /* 4x */
 174     0, 1, 1, 1, 1, 1, 1, 1, 1, 1, 0, 1, 0, 0, 0, 0,  /* 5x */
 175     0, 0, 1, 1, 1, 1, 1, 1, 1, 1, 0, 0, 0, 1, 0, 0,  /* 6x */
 176     0, 1, 1, 1, 1, 1, 1, 1, 1, 0, 0, 0, 0, 0, 0, 0,  /* 7x */
 177     0, 1, 1, 1, 1, 1, 1, 1, 1, 1, 0, 0, 1, 1, 1, 0,  /* 8x */
 178     0, 1, 1, 1, 1, 1, 1, 1, 1, 1, 0, 0, 1, 0, 1, 0,  /* 9x */
 179     1, 0, 1, 1, 1, 1, 1, 1, 1, 1, 1, 0, 1, 1, 1, 0,  /* Ax */
 180     0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,  /* Bx */
 181     0, 1, 1, 1, 1, 1, 1, 1, 1, 1, 0, 1, 1, 1, 1, 1,  /* Cx */
 182     0, 1, 1, 1, 1, 1, 1, 1, 1, 1, 0, 1, 1, 1, 1, 1,  /* Dx */
 183     0, 0, 1, 1, 1, 1, 1, 1, 1, 1, 0, 1, 1, 1, 1, 1,  /* Ex */
 184     1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 0, 1, 1, 1, 1, 0,  /* Fx */
 185 };
 186 #define IdChar(C)  (((c=C)>=0x42 && sqlite3IsEbcdicIdChar[c-0x40]))
 187 #endif
 188
 189 /* Make the IdChar function accessible from ctime.c and alter.c */
 190 int sqlite3IsIdChar(u8 c){ return IdChar(c); }
 191
 192 #ifndef SQLITE_OMIT_WINDOWFUNC
 193 /*
 194 ** Return the id of the next token in string (*pz). Before returning, set
 195 ** (*pz) to point to the byte following the parsed token.
 196 */
 197 static int getToken(const unsigned char **pz){
 198   const unsigned char *z = *pz;
 199   int t;                          /* Token type to return */
 200   do {
 201     z += sqlite3GetToken(z, &t);
 202   }while( t==TK_SPACE );
 203   if( t==TK_ID
 204    || t==TK_STRING
 205    || t==TK_JOIN_KW
 206    || t==TK_WINDOW
 207    || t==TK_OVER
 208    || sqlite3ParserFallback(t)==TK_ID
 209   ){
 210     t = TK_ID;
 211   }
 212   *pz = z;
 213   return t;
 214 }
 215
 216 /*
 217 ** The following three functions are called immediately after the tokenizer
 218 ** reads the keywords WINDOW, OVER and FILTER, respectively, to determine
 219 ** whether the token should be treated as a keyword or an SQL identifier.
 220 ** This cannot be handled by the usual lemon %fallback method, due to
 221 ** the ambiguity in some constructions. e.g.
 222 **
 223 **   SELECT sum(x) OVER ...
 224 **
 225 ** In the above, "OVER" might be a keyword, or it might be an alias for the
 226 ** sum(x) expression. If a "%fallback ID OVER" directive were added to
 227 ** grammar, then SQLite would always treat "OVER" as an alias, making it
 228 ** impossible to call a window-function without a FILTER clause.
 229 **
 230 ** WINDOW is treated as a keyword if:
 231 **
 232 **   * the following token is an identifier, or a keyword that can fallback
 233 **     to being an identifier, and
 234 **   * the token after than one is TK_AS.
 235 **
 236 ** OVER is a keyword if:
 237 **
 238 **   * the previous token was TK_RP, and
 239 **   * the next token is either TK_LP or an identifier.
 240 **
 241 ** FILTER is a keyword if:
 242 **
 243 **   * the previous token was TK_RP, and
 244 **   * the next token is TK_LP.
 245 */
 246 static int analyzeWindowKeyword(const unsigned char *z){
 247   int t;
 248   t = getToken(&z);
 249   if( t!=TK_ID ) return TK_ID;
 250   t = getToken(&z);
 251   if( t!=TK_AS ) return TK_ID;
 252   return TK_WINDOW;
 253 }
 254 static int analyzeOverKeyword(const unsigned char *z, int lastToken){
 255   if( lastToken==TK_RP ){
 256     int t = getToken(&z);
 257     if( t==TK_LP || t==TK_ID ) return TK_OVER;
 258   }
 259   return TK_ID;
 260 }
 261 static int analyzeFilterKeyword(const unsigned char *z, int lastToken){
 262   if( lastToken==TK_RP && getToken(&z)==TK_LP ){
 263     return TK_FILTER;
 264   }
 265   return TK_ID;
 266 }
 267 #endif /* SQLITE_OMIT_WINDOWFUNC */
 268
 269 /*
 270 ** Return the length (in bytes) of the token that begins at z[0].
 271 ** Store the token type in *tokenType before returning.
 272 */
 273 int sqlite3GetToken(const unsigned char *z, int *tokenType){
 274   int i, c;
 275   switch( aiClass[*z] ){  /* Switch on the character-class of the first byte
 276                           ** of the token. See the comment on the CC_ defines
 277                           ** above. */
 278     case CC_SPACE: {
 279       testcase( z[0]==' ' );
 280       testcase( z[0]=='\t' );
 281       testcase( z[0]=='\n' );
 282       testcase( z[0]=='\f' );
 283       testcase( z[0]=='\r' );
 284       for(i=1; sqlite3Isspace(z[i]); i++){}
 285       *tokenType = TK_SPACE;
 286       return i;
 287     }
 288     case CC_MINUS: {
 289       if( z[1]=='-' ){
 290         for(i=2; (c=z[i])!=0 && c!='\n'; i++){}
 291         *tokenType = TK_SPACE;   /* IMP: R-22934-25134 */
 292         return i;
 293       }
 294       *tokenType = TK_MINUS;
 295       return 1;
 296     }
 297     case CC_LP: {
 298       *tokenType = TK_LP;
 299       return 1;
 300     }
 301     case CC_RP: {
 302       *tokenType = TK_RP;
 303       return 1;
 304     }
 305     case CC_SEMI: {
 306       *tokenType = TK_SEMI;
 307       return 1;
 308     }
 309     case CC_PLUS: {
 310       *tokenType = TK_PLUS;
 311       return 1;
 312     }
 313     case CC_STAR: {
 314       *tokenType = TK_STAR;
 315       return 1;
 316     }
 317     case CC_SLASH: {
 318       if( z[1]!='*' || z[2]==0 ){
 319         *tokenType = TK_SLASH;
 320         return 1;
 321       }
 322       for(i=3, c=z[2]; (c!='*' || z[i]!='/') && (c=z[i])!=0; i++){}
 323       if( c ) i++;
 324       *tokenType = TK_SPACE;   /* IMP: R-22934-25134 */
 325       return i;
 326     }
 327     case CC_PERCENT: {
 328       *tokenType = TK_REM;
 329       return 1;
 330     }
 331     case CC_EQ: {
 332       *tokenType = TK_EQ;
 333       return 1 + (z[1]=='=');
 334     }
 335     case CC_LT: {
 336       if( (c=z[1])=='=' ){
 337         *tokenType = TK_LE;
 338         return 2;
 339       }else if( c=='>' ){
 340         *tokenType = TK_NE;
 341         return 2;
 342       }else if( c=='<' ){
 343         *tokenType = TK_LSHIFT;
 344         return 2;
 345       }else{
 346         *tokenType = TK_LT;
 347         return 1;
 348       }
 349     }
 350     case CC_GT: {
 351       if( (c=z[1])=='=' ){
 352         *tokenType = TK_GE;
 353         return 2;
 354       }else if( c=='>' ){
 355         *tokenType = TK_RSHIFT;
 356         return 2;
 357       }else{
 358         *tokenType = TK_GT;
 359         return 1;
 360       }
 361     }
 362     case CC_BANG: {
 363       if( z[1]!='=' ){
 364         *tokenType = TK_ILLEGAL;
 365         return 1;
 366       }else{
 367         *tokenType = TK_NE;
 368         return 2;
 369       }
 370     }
 371     case CC_PIPE: {
 372       if( z[1]!='|' ){
 373         *tokenType = TK_BITOR;
 374         return 1;
 375       }else{
 376         *tokenType = TK_CONCAT;
 377         return 2;
 378       }
 379     }
 380     case CC_COMMA: {
 381       *tokenType = TK_COMMA;
 382       return 1;
 383     }
 384     case CC_AND: {
 385       *tokenType = TK_BITAND;
 386       return 1;
 387     }
 388     case CC_TILDA: {
 389       *tokenType = TK_BITNOT;
 390       return 1;
 391     }
 392     case CC_QUOTE: {
 393       int delim = z[0];
 394       testcase( delim=='`' );
 395       testcase( delim=='\'' );
 396       testcase( delim=='"' );
 397       for(i=1; (c=z[i])!=0; i++){
 398         if( c==delim ){
 399           if( z[i+1]==delim ){
 400             i++;
 401           }else{
 402             break;
 403           }
 404         }
 405       }
 406       if( c=='\'' ){
 407         *tokenType = TK_STRING;
 408         return i+1;
 409       }else if( c!=0 ){
 410         *tokenType = TK_ID;
 411         return i+1;
 412       }else{
 413         *tokenType = TK_ILLEGAL;
 414         return i;
 415       }
 416     }
 417     case CC_DOT: {
 418 #ifndef SQLITE_OMIT_FLOATING_POINT
 419       if( !sqlite3Isdigit(z[1]) )
 420 #endif
 421       {
 422         *tokenType = TK_DOT;
 423         return 1;
 424       }
 425       /* If the next character is a digit, this is a floating point
 426       ** number that begins with ".".  Fall thru into the next case */
 427       /* no break */ deliberate_fall_through
 428     }
 429     case CC_DIGIT: {
 430       testcase( z[0]=='0' );  testcase( z[0]=='1' );  testcase( z[0]=='2' );
 431       testcase( z[0]=='3' );  testcase( z[0]=='4' );  testcase( z[0]=='5' );
 432       testcase( z[0]=='6' );  testcase( z[0]=='7' );  testcase( z[0]=='8' );
 433       testcase( z[0]=='9' );
 434       *tokenType = TK_INTEGER;
 435 #ifndef SQLITE_OMIT_HEX_INTEGER
 436       if( z[0]=='0' && (z[1]=='x' || z[1]=='X') && sqlite3Isxdigit(z[2]) ){
 437         for(i=3; sqlite3Isxdigit(z[i]); i++){}
 438         return i;
 439       }
 440 #endif
 441       for(i=0; sqlite3Isdigit(z[i]); i++){}
 442 #ifndef SQLITE_OMIT_FLOATING_POINT
 443       if( z[i]=='.' ){
 444         i++;
 445         while( sqlite3Isdigit(z[i]) ){ i++; }
 446         *tokenType = TK_FLOAT;
 447       }
 448       if( (z[i]=='e' || z[i]=='E') &&
 449            ( sqlite3Isdigit(z[i+1])
 450             || ((z[i+1]=='+' || z[i+1]=='-') && sqlite3Isdigit(z[i+2]))
 451            )
 452       ){
 453         i += 2;
 454         while( sqlite3Isdigit(z[i]) ){ i++; }
 455         *tokenType = TK_FLOAT;
 456       }
 457 #endif
 458       while( IdChar(z[i]) ){
 459         *tokenType = TK_ILLEGAL;
 460         i++;
 461       }
 462       return i;
 463     }
 464     case CC_QUOTE2: {
 465       for(i=1, c=z[0]; c!=']' && (c=z[i])!=0; i++){}
 466       *tokenType = c==']' ? TK_ID : TK_ILLEGAL;
 467       return i;
 468     }
 469     case CC_VARNUM: {
 470       *tokenType = TK_VARIABLE;
 471       for(i=1; sqlite3Isdigit(z[i]); i++){}
 472       return i;
 473     }
 474     case CC_DOLLAR:
 475     case CC_VARALPHA: {
 476       int n = 0;
 477       testcase( z[0]=='$' );  testcase( z[0]=='@' );
 478       testcase( z[0]==':' );  testcase( z[0]=='#' );
 479       *tokenType = TK_VARIABLE;
 480       for(i=1; (c=z[i])!=0; i++){
 481         if( IdChar(c) ){
 482           n++;
 483 #ifndef SQLITE_OMIT_TCL_VARIABLE
 484         }else if( c=='(' && n>0 ){
 485           do{
 486             i++;
 487           }while( (c=z[i])!=0 && !sqlite3Isspace(c) && c!=')' );
 488           if( c==')' ){
 489             i++;
 490           }else{
 491             *tokenType = TK_ILLEGAL;
 492           }
 493           break;
 494         }else if( c==':' && z[i+1]==':' ){
 495           i++;
 496 #endif
 497         }else{
 498           break;
 499         }
 500       }
 501       if( n==0 ) *tokenType = TK_ILLEGAL;
 502       return i;
 503     }
 504     case CC_KYWD0: {
 505       for(i=1; aiClass[z[i]]<=CC_KYWD; i++){}
 506       if( IdChar(z[i]) ){
 507         /* This token started out using characters that can appear in keywords,
 508         ** but z[i] is a character not allowed within keywords, so this must
 509         ** be an identifier instead */
 510         i++;
 511         break;
 512       }
 513       *tokenType = TK_ID;
 514       return keywordCode((char*)z, i, tokenType);
 515     }
 516     case CC_X: {
 517 #ifndef SQLITE_OMIT_BLOB_LITERAL
 518       testcase( z[0]=='x' ); testcase( z[0]=='X' );
 519       if( z[1]=='\'' ){
 520         *tokenType = TK_BLOB;
 521         for(i=2; sqlite3Isxdigit(z[i]); i++){}
 522         if( z[i]!='\'' || i%2 ){
 523           *tokenType = TK_ILLEGAL;
 524           while( z[i] && z[i]!='\'' ){ i++; }
 525         }
 526         if( z[i] ) i++;
 527         return i;
 528       }
 529 #endif
 530       /* If it is not a BLOB literal, then it must be an ID, since no
 531       ** SQL keywords start with the letter 'x'.  Fall through */
 532       /* no break */ deliberate_fall_through
 533     }
 534     case CC_KYWD:
 535     case CC_ID: {
 536       i = 1;
 537       break;
 538     }
 539     case CC_BOM: {
 540       if( z[1]==0xbb && z[2]==0xbf ){
 541         *tokenType = TK_SPACE;
 542         return 3;
 543       }
 544       i = 1;
 545       break;
 546     }
 547     case CC_NUL: {
 548       *tokenType = TK_ILLEGAL;
 549       return 0;
 550     }
 551     default: {
 552       *tokenType = TK_ILLEGAL;
 553       return 1;
 554     }
 555   }
 556   while( IdChar(z[i]) ){ i++; }
 557   *tokenType = TK_ID;
 558   return i;
 559 }
 560
 561 /*
 562 ** Run the parser on the given SQL string.  The parser structure is
 563 ** passed in.  An SQLITE_ status code is returned.  If an error occurs
 564 ** then an and attempt is made to write an error message into
 565 ** memory obtained from sqlite3_malloc() and to make *pzErrMsg point to that
 566 ** error message.
 567 */
 568 int sqlite3RunParser(Parse *pParse, const char *zSql, char **pzErrMsg){
 569   int nErr = 0;                   /* Number of errors encountered */
 570   void *pEngine;                  /* The LEMON-generated LALR(1) parser */
 571   int n = 0;                      /* Length of the next token token */
 572   int tokenType;                  /* type of the next token */
 573   int lastTokenParsed = -1;       /* type of the previous token */
 574   sqlite3 *db = pParse->db;       /* The database connection */
 575   int mxSqlLen;                   /* Max length of an SQL string */
 576   Parse *pParentParse = 0;        /* Outer parse context, if any */
 577 #ifdef sqlite3Parser_ENGINEALWAYSONSTACK
 578   yyParser sEngine;    /* Space to hold the Lemon-generated Parser object */
 579 #endif
 580   VVA_ONLY( u8 startedWithOom = db->mallocFailed );
 581
 582   assert( zSql!=0 );
 583   mxSqlLen = db->aLimit[SQLITE_LIMIT_SQL_LENGTH];
 584   if( db->nVdbeActive==0 ){
 585     AtomicStore(&db->u1.isInterrupted, 0);
 586   }
 587   pParse->rc = SQLITE_OK;
 588   pParse->zTail = zSql;
 589   assert( pzErrMsg!=0 );
 590 #ifdef SQLITE_DEBUG
 591   if( db->flags & SQLITE_ParserTrace ){
 592     printf("parser: [[[%s]]]\n", zSql);
 593     sqlite3ParserTrace(stdout, "parser: ");
 594   }else{
 595     sqlite3ParserTrace(0, 0);
 596   }
 597 #endif
 598 #ifdef sqlite3Parser_ENGINEALWAYSONSTACK
 599   pEngine = &sEngine;
 600   sqlite3ParserInit(pEngine, pParse);
 601 #else
 602   pEngine = sqlite3ParserAlloc(sqlite3Malloc, pParse);
 603   if( pEngine==0 ){
 604     sqlite3OomFault(db);
 605     return SQLITE_NOMEM_BKPT;
 606   }
 607 #endif
 608   assert( pParse->pNewTable==0 );
 609   assert( pParse->pNewTrigger==0 );
 610   assert( pParse->nVar==0 );
 611   assert( pParse->pVList==0 );
 612   pParentParse = db->pParse;
 613   db->pParse = pParse;
 614   while( 1 ){
 615     n = sqlite3GetToken((u8*)zSql, &tokenType);
 616     mxSqlLen -= n;
 617     if( mxSqlLen<0 ){
 618       pParse->rc = SQLITE_TOOBIG;
 619       break;
 620     }
 621 #ifndef SQLITE_OMIT_WINDOWFUNC
 622     if( tokenType>=TK_WINDOW ){
 623       assert( tokenType==TK_SPACE || tokenType==TK_OVER || tokenType==TK_FILTER
 624            || tokenType==TK_ILLEGAL || tokenType==TK_WINDOW
 625       );
 626 #else
 627     if( tokenType>=TK_SPACE ){
 628       assert( tokenType==TK_SPACE || tokenType==TK_ILLEGAL );
 629 #endif /* SQLITE_OMIT_WINDOWFUNC */
 630       if( AtomicLoad(&db->u1.isInterrupted) ){
 631         pParse->rc = SQLITE_INTERRUPT;
 632         break;
 633       }
 634       if( tokenType==TK_SPACE ){
 635         zSql += n;
 636         continue;
 637       }
 638       if( zSql[0]==0 ){
 639         /* Upon reaching the end of input, call the parser two more times
 640         ** with tokens TK_SEMI and 0, in that order. */
 641         if( lastTokenParsed==TK_SEMI ){
 642           tokenType = 0;
 643         }else if( lastTokenParsed==0 ){
 644           break;
 645         }else{
 646           tokenType = TK_SEMI;
 647         }
 648         n = 0;
 649 #ifndef SQLITE_OMIT_WINDOWFUNC
 650       }else if( tokenType==TK_WINDOW ){
 651         assert( n==6 );
 652         tokenType = analyzeWindowKeyword((const u8*)&zSql[6]);
 653       }else if( tokenType==TK_OVER ){
 654         assert( n==4 );
 655         tokenType = analyzeOverKeyword((const u8*)&zSql[4], lastTokenParsed);
 656       }else if( tokenType==TK_FILTER ){
 657         assert( n==6 );
 658         tokenType = analyzeFilterKeyword((const u8*)&zSql[6], lastTokenParsed);
 659 #endif /* SQLITE_OMIT_WINDOWFUNC */
 660       }else{
 661         sqlite3ErrorMsg(pParse, "unrecognized token: \"%.*s\"", n, zSql);
 662         break;
 663       }
 664     }
 665     pParse->sLastToken.z = zSql;
 666     pParse->sLastToken.n = n;
 667     sqlite3Parser(pEngine, tokenType, pParse->sLastToken);
 668     lastTokenParsed = tokenType;
 669     zSql += n;
 670     assert( db->mallocFailed==0 || pParse->rc!=SQLITE_OK || startedWithOom );
 671     if( pParse->rc!=SQLITE_OK ) break;
 672   }
 673   assert( nErr==0 );
 674 #ifdef YYTRACKMAXSTACKDEPTH
 675   sqlite3_mutex_enter(sqlite3MallocMutex());
 676   sqlite3StatusHighwater(SQLITE_STATUS_PARSER_STACK,
 677       sqlite3ParserStackPeak(pEngine)
 678   );
 679   sqlite3_mutex_leave(sqlite3MallocMutex());
 680 #endif /* YYDEBUG */
 681 #ifdef sqlite3Parser_ENGINEALWAYSONSTACK
 682   sqlite3ParserFinalize(pEngine);
 683 #else
 684   sqlite3ParserFree(pEngine, sqlite3_free);
 685 #endif
 686   if( db->mallocFailed ){
 687     pParse->rc = SQLITE_NOMEM_BKPT;
 688   }
 689   if( pParse->rc!=SQLITE_OK && pParse->rc!=SQLITE_DONE && pParse->zErrMsg==0 ){
 690     pParse->zErrMsg = sqlite3MPrintf(db, "%s", sqlite3ErrStr(pParse->rc));
 691   }
 692   assert( pzErrMsg!=0 );
 693   if( pParse->zErrMsg ){
 694     *pzErrMsg = pParse->zErrMsg;
 695     sqlite3_log(pParse->rc, "%s in \"%s\"",
 696                 *pzErrMsg, pParse->zTail);
 697     pParse->zErrMsg = 0;
 698     nErr++;
 699   }
 700   pParse->zTail = zSql;
 701   if( pParse->pVdbe && pParse->nErr>0 && pParse->nested==0 ){
 702     sqlite3VdbeDelete(pParse->pVdbe);
 703     pParse->pVdbe = 0;
 704   }
 705 #ifndef SQLITE_OMIT_SHARED_CACHE
 706   if( pParse->nested==0 ){
 707     sqlite3DbFree(db, pParse->aTableLock);
 708     pParse->aTableLock = 0;
 709     pParse->nTableLock = 0;
 710   }
 711 #endif
 712 #ifndef SQLITE_OMIT_VIRTUALTABLE
 713   sqlite3_free(pParse->apVtabLock);
 714 #endif
 715
 716   if( !IN_SPECIAL_PARSE ){
 717     /* If the pParse->declareVtab flag is set, do not delete any table
 718     ** structure built up in pParse->pNewTable. The calling code (see vtab.c)
 719     ** will take responsibility for freeing the Table structure.
 720     */
 721     sqlite3DeleteTable(db, pParse->pNewTable);
 722   }
 723   if( !IN_RENAME_OBJECT ){
 724     sqlite3DeleteTrigger(db, pParse->pNewTrigger);
 725   }
 726   sqlite3DbFree(db, pParse->pVList);
 727   db->pParse = pParentParse;
 728   assert( nErr==0 || pParse->rc!=SQLITE_OK );
 729   return nErr;
 730 }
 731
 732
 733 #ifdef SQLITE_ENABLE_NORMALIZE
 734 /*
 735 ** Insert a single space character into pStr if the current string
 736 ** ends with an identifier
 737 */
 738 static void addSpaceSeparator(sqlite3_str *pStr){
 739   if( pStr->nChar && sqlite3IsIdChar(pStr->zText[pStr->nChar-1]) ){
 740     sqlite3_str_append(pStr, " ", 1);
 741   }
 742 }
 743
 744 /*
 745 ** Compute a normalization of the SQL given by zSql[0..nSql-1].  Return
 746 ** the normalization in space obtained from sqlite3DbMalloc().  Or return
 747 ** NULL if anything goes wrong or if zSql is NULL.
 748 */
 749 char *sqlite3Normalize(
 750   Vdbe *pVdbe,       /* VM being reprepared */
 751   const char *zSql   /* The original SQL string */
 752 ){
 753   sqlite3 *db;       /* The database connection */
 754   int i;             /* Next unread byte of zSql[] */
 755   int n;             /* length of current token */
 756   int tokenType;     /* type of current token */
 757   int prevType = 0;  /* Previous non-whitespace token */
 758   int nParen;        /* Number of nested levels of parentheses */
 759   int iStartIN;      /* Start of RHS of IN operator in z[] */
 760   int nParenAtIN;    /* Value of nParent at start of RHS of IN operator */
 761   u32 j;             /* Bytes of normalized SQL generated so far */
 762   sqlite3_str *pStr; /* The normalized SQL string under construction */
 763
 764   db = sqlite3VdbeDb(pVdbe);
 765   tokenType = -1;
 766   nParen = iStartIN = nParenAtIN = 0;
 767   pStr = sqlite3_str_new(db);
 768   assert( pStr!=0 );  /* sqlite3_str_new() never returns NULL */
 769   for(i=0; zSql[i] && pStr->accError==0; i+=n){
 770     if( tokenType!=TK_SPACE ){
 771       prevType = tokenType;
 772     }
 773     n = sqlite3GetToken((unsigned char*)zSql+i, &tokenType);
 774     if( NEVER(n<=0) ) break;
 775     switch( tokenType ){
 776       case TK_SPACE: {
 777         break;
 778       }
 779       case TK_NULL: {
 780         if( prevType==TK_IS || prevType==TK_NOT ){
 781           sqlite3_str_append(pStr, " NULL", 5);
 782           break;
 783         }
 784         /* Fall through */
 785       }
 786       case TK_STRING:
 787       case TK_INTEGER:
 788       case TK_FLOAT:
 789       case TK_VARIABLE:
 790       case TK_BLOB: {
 791         sqlite3_str_append(pStr, "?", 1);
 792         break;
 793       }
 794       case TK_LP: {
 795         nParen++;
 796         if( prevType==TK_IN ){
 797           iStartIN = pStr->nChar;
 798           nParenAtIN = nParen;
 799         }
 800         sqlite3_str_append(pStr, "(", 1);
 801         break;
 802       }
 803       case TK_RP: {
 804         if( iStartIN>0 && nParen==nParenAtIN ){
 805           assert( pStr->nChar>=(u32)iStartIN );
 806           pStr->nChar = iStartIN+1;
 807           sqlite3_str_append(pStr, "?,?,?", 5);
 808           iStartIN = 0;
 809         }
 810         nParen--;
 811         sqlite3_str_append(pStr, ")", 1);
 812         break;
 813       }
 814       case TK_ID: {
 815         iStartIN = 0;
 816         j = pStr->nChar;
 817         if( sqlite3Isquote(zSql[i]) ){
 818           char *zId = sqlite3DbStrNDup(db, zSql+i, n);
 819           int nId;
 820           int eType = 0;
 821           if( zId==0 ) break;
 822           sqlite3Dequote(zId);
 823           if( zSql[i]=='"' && sqlite3VdbeUsesDoubleQuotedString(pVdbe, zId) ){
 824             sqlite3_str_append(pStr, "?", 1);
 825             sqlite3DbFree(db, zId);
 826             break;
 827           }
 828           nId = sqlite3Strlen30(zId);
 829           if( sqlite3GetToken((u8*)zId, &eType)==nId && eType==TK_ID ){
 830             addSpaceSeparator(pStr);
 831             sqlite3_str_append(pStr, zId, nId);
 832           }else{
 833             sqlite3_str_appendf(pStr, "\"%w\"", zId);
 834           }
 835           sqlite3DbFree(db, zId);
 836         }else{
 837           addSpaceSeparator(pStr);
 838           sqlite3_str_append(pStr, zSql+i, n);
 839         }
 840         while( j<pStr->nChar ){
 841           pStr->zText[j] = sqlite3Tolower(pStr->zText[j]);
 842           j++;
 843         }
 844         break;
 845       }
 846       case TK_SELECT: {
 847         iStartIN = 0;
 848         /* fall through */
 849       }
 850       default: {
 851         if( sqlite3IsIdChar(zSql[i]) ) addSpaceSeparator(pStr);
 852         j = pStr->nChar;
 853         sqlite3_str_append(pStr, zSql+i, n);
 854         while( j<pStr->nChar ){
 855           pStr->zText[j] = sqlite3Toupper(pStr->zText[j]);
 856           j++;
 857         }
 858         break;
 859       }
 860     }
 861   }
 862   if( tokenType!=TK_SEMI ) sqlite3_str_append(pStr, ";", 1);
 863   return sqlite3_str_finish(pStr);
 864 }
 865 #endif /* SQLITE_ENABLE_NORMALIZE */