xref: /sqlite-3.40.0/src/tokenize.c (revision 9cffb0ff)
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 
60 static const unsigned char aiClass[] = {
61 #ifdef SQLITE_ASCII
62 /*         x0  x1  x2  x3  x4  x5  x6  x7  x8  x9  xa  xb  xc  xd  xe  xf */
63 /* 0x */   29, 28, 28, 28, 28, 28, 28, 28, 28,  7,  7, 28,  7,  7, 28, 28,
64 /* 1x */   28, 28, 28, 28, 28, 28, 28, 28, 28, 28, 28, 28, 28, 28, 28, 28,
65 /* 2x */    7, 15,  8,  5,  4, 22, 24,  8, 17, 18, 21, 20, 23, 11, 26, 16,
66 /* 3x */    3,  3,  3,  3,  3,  3,  3,  3,  3,  3,  5, 19, 12, 14, 13,  6,
67 /* 4x */    5,  1,  1,  1,  1,  1,  1,  1,  1,  1,  1,  1,  1,  1,  1,  1,
68 /* 5x */    1,  1,  1,  1,  1,  1,  1,  1,  0,  2,  2,  9, 28, 28, 28,  2,
69 /* 6x */    8,  1,  1,  1,  1,  1,  1,  1,  1,  1,  1,  1,  1,  1,  1,  1,
70 /* 7x */    1,  1,  1,  1,  1,  1,  1,  1,  0,  2,  2, 28, 10, 28, 25, 28,
71 /* 8x */    2,  2,  2,  2,  2,  2,  2,  2,  2,  2,  2,  2,  2,  2,  2,  2,
72 /* 9x */    2,  2,  2,  2,  2,  2,  2,  2,  2,  2,  2,  2,  2,  2,  2,  2,
73 /* Ax */    2,  2,  2,  2,  2,  2,  2,  2,  2,  2,  2,  2,  2,  2,  2,  2,
74 /* Bx */    2,  2,  2,  2,  2,  2,  2,  2,  2,  2,  2,  2,  2,  2,  2,  2,
75 /* Cx */    2,  2,  2,  2,  2,  2,  2,  2,  2,  2,  2,  2,  2,  2,  2,  2,
76 /* Dx */    2,  2,  2,  2,  2,  2,  2,  2,  2,  2,  2,  2,  2,  2,  2,  2,
77 /* Ex */    2,  2,  2,  2,  2,  2,  2,  2,  2,  2,  2,  2,  2,  2,  2,  2,
78 /* Fx */    2,  2,  2,  2,  2,  2,  2,  2,  2,  2,  2,  2,  2,  2,  2,  2
79 #endif
80 #ifdef SQLITE_EBCDIC
81 /*         x0  x1  x2  x3  x4  x5  x6  x7  x8  x9  xa  xb  xc  xd  xe  xf */
82 /* 0x */   29, 28, 28, 28, 28,  7, 28, 28, 28, 28, 28, 28,  7,  7, 28, 28,
83 /* 1x */   28, 28, 28, 28, 28,  7, 28, 28, 28, 28, 28, 28, 28, 28, 28, 28,
84 /* 2x */   28, 28, 28, 28, 28,  7, 28, 28, 28, 28, 28, 28, 28, 28, 28, 28,
85 /* 3x */   28, 28, 28, 28, 28, 28, 28, 28, 28, 28, 28, 28, 28, 28, 28, 28,
86 /* 4x */    7, 28, 28, 28, 28, 28, 28, 28, 28, 28, 28, 26, 12, 17, 20, 10,
87 /* 5x */   24, 28, 28, 28, 28, 28, 28, 28, 28, 28, 15,  4, 21, 18, 19, 28,
88 /* 6x */   11, 16, 28, 28, 28, 28, 28, 28, 28, 28, 28, 23, 22,  2, 13,  6,
89 /* 7x */   28, 28, 28, 28, 28, 28, 28, 28, 28,  8,  5,  5,  5,  8, 14,  8,
90 /* 8x */   28,  1,  1,  1,  1,  1,  1,  1,  1,  1, 28, 28, 28, 28, 28, 28,
91 /* 9x */   28,  1,  1,  1,  1,  1,  1,  1,  1,  1, 28, 28, 28, 28, 28, 28,
92 /* Ax */   28, 25,  1,  1,  1,  1,  1,  0,  2,  2, 28, 28, 28, 28, 28, 28,
93 /* Bx */   28, 28, 28, 28, 28, 28, 28, 28, 28, 28,  9, 28, 28, 28, 28, 28,
94 /* Cx */   28,  1,  1,  1,  1,  1,  1,  1,  1,  1, 28, 28, 28, 28, 28, 28,
95 /* Dx */   28,  1,  1,  1,  1,  1,  1,  1,  1,  1, 28, 28, 28, 28, 28, 28,
96 /* Ex */   28, 28,  1,  1,  1,  1,  1,  0,  2,  2, 28, 28, 28, 28, 28, 28,
97 /* Fx */    3,  3,  3,  3,  3,  3,  3,  3,  3,  3, 28, 28, 28, 28, 28, 28,
98 #endif
99 };
100 
101 /*
102 ** The charMap() macro maps alphabetic characters (only) into their
103 ** lower-case ASCII equivalent.  On ASCII machines, this is just
104 ** an upper-to-lower case map.  On EBCDIC machines we also need
105 ** to adjust the encoding.  The mapping is only valid for alphabetics
106 ** which are the only characters for which this feature is used.
107 **
108 ** Used by keywordhash.h
109 */
110 #ifdef SQLITE_ASCII
111 # define charMap(X) sqlite3UpperToLower[(unsigned char)X]
112 #endif
113 #ifdef SQLITE_EBCDIC
114 # define charMap(X) ebcdicToAscii[(unsigned char)X]
115 const unsigned char ebcdicToAscii[] = {
116 /* 0   1   2   3   4   5   6   7   8   9   A   B   C   D   E   F */
117    0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  /* 0x */
118    0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  /* 1x */
119    0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  /* 2x */
120    0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  /* 3x */
121    0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  /* 4x */
122    0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  /* 5x */
123    0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0, 95,  0,  0,  /* 6x */
124    0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  /* 7x */
125    0, 97, 98, 99,100,101,102,103,104,105,  0,  0,  0,  0,  0,  0,  /* 8x */
126    0,106,107,108,109,110,111,112,113,114,  0,  0,  0,  0,  0,  0,  /* 9x */
127    0,  0,115,116,117,118,119,120,121,122,  0,  0,  0,  0,  0,  0,  /* Ax */
128    0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  /* Bx */
129    0, 97, 98, 99,100,101,102,103,104,105,  0,  0,  0,  0,  0,  0,  /* Cx */
130    0,106,107,108,109,110,111,112,113,114,  0,  0,  0,  0,  0,  0,  /* Dx */
131    0,  0,115,116,117,118,119,120,121,122,  0,  0,  0,  0,  0,  0,  /* Ex */
132    0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  /* Fx */
133 };
134 #endif
135 
136 /*
137 ** The sqlite3KeywordCode function looks up an identifier to determine if
138 ** it is a keyword.  If it is a keyword, the token code of that keyword is
139 ** returned.  If the input is not a keyword, TK_ID is returned.
140 **
141 ** The implementation of this routine was generated by a program,
142 ** mkkeywordhash.c, located in the tool subdirectory of the distribution.
143 ** The output of the mkkeywordhash.c program is written into a file
144 ** named keywordhash.h and then included into this source file by
145 ** the #include below.
146 */
147 #include "keywordhash.h"
148 
149 
150 /*
151 ** If X is a character that can be used in an identifier then
152 ** IdChar(X) will be true.  Otherwise it is false.
153 **
154 ** For ASCII, any character with the high-order bit set is
155 ** allowed in an identifier.  For 7-bit characters,
156 ** sqlite3IsIdChar[X] must be 1.
157 **
158 ** For EBCDIC, the rules are more complex but have the same
159 ** end result.
160 **
161 ** Ticket #1066.  the SQL standard does not allow '$' in the
162 ** middle of identifiers.  But many SQL implementations do.
163 ** SQLite will allow '$' in identifiers for compatibility.
164 ** But the feature is undocumented.
165 */
166 #ifdef SQLITE_ASCII
167 #define IdChar(C)  ((sqlite3CtypeMap[(unsigned char)C]&0x46)!=0)
168 #endif
169 #ifdef SQLITE_EBCDIC
170 const char sqlite3IsEbcdicIdChar[] = {
171 /* x0 x1 x2 x3 x4 x5 x6 x7 x8 x9 xA xB xC xD xE xF */
172     0, 0, 1, 1, 1, 1, 1, 1, 1, 1, 0, 0, 0, 0, 0, 0,  /* 4x */
173     0, 1, 1, 1, 1, 1, 1, 1, 1, 1, 0, 1, 0, 0, 0, 0,  /* 5x */
174     0, 0, 1, 1, 1, 1, 1, 1, 1, 1, 0, 0, 0, 1, 0, 0,  /* 6x */
175     0, 1, 1, 1, 1, 1, 1, 1, 1, 0, 0, 0, 0, 0, 0, 0,  /* 7x */
176     0, 1, 1, 1, 1, 1, 1, 1, 1, 1, 0, 0, 1, 1, 1, 0,  /* 8x */
177     0, 1, 1, 1, 1, 1, 1, 1, 1, 1, 0, 0, 1, 0, 1, 0,  /* 9x */
178     1, 0, 1, 1, 1, 1, 1, 1, 1, 1, 1, 0, 1, 1, 1, 0,  /* Ax */
179     0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,  /* Bx */
180     0, 1, 1, 1, 1, 1, 1, 1, 1, 1, 0, 1, 1, 1, 1, 1,  /* Cx */
181     0, 1, 1, 1, 1, 1, 1, 1, 1, 1, 0, 1, 1, 1, 1, 1,  /* Dx */
182     0, 0, 1, 1, 1, 1, 1, 1, 1, 1, 0, 1, 1, 1, 1, 1,  /* Ex */
183     1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 0, 1, 1, 1, 1, 0,  /* Fx */
184 };
185 #define IdChar(C)  (((c=C)>=0x42 && sqlite3IsEbcdicIdChar[c-0x40]))
186 #endif
187 
188 /* Make the IdChar function accessible from ctime.c and alter.c */
189 int sqlite3IsIdChar(u8 c){ return IdChar(c); }
190 
191 #ifndef SQLITE_OMIT_WINDOWFUNC
192 /*
193 ** Return the id of the next token in string (*pz). Before returning, set
194 ** (*pz) to point to the byte following the parsed token.
195 */
196 static int getToken(const unsigned char **pz){
197   const unsigned char *z = *pz;
198   int t;                          /* Token type to return */
199   do {
200     z += sqlite3GetToken(z, &t);
201   }while( t==TK_SPACE );
202   if( t==TK_ID
203    || t==TK_STRING
204    || t==TK_JOIN_KW
205    || t==TK_WINDOW
206    || t==TK_OVER
207    || sqlite3ParserFallback(t)==TK_ID
208   ){
209     t = TK_ID;
210   }
211   *pz = z;
212   return t;
213 }
214 
215 /*
216 ** The following three functions are called immediately after the tokenizer
217 ** reads the keywords WINDOW, OVER and FILTER, respectively, to determine
218 ** whether the token should be treated as a keyword or an SQL identifier.
219 ** This cannot be handled by the usual lemon %fallback method, due to
220 ** the ambiguity in some constructions. e.g.
221 **
222 **   SELECT sum(x) OVER ...
223 **
224 ** In the above, "OVER" might be a keyword, or it might be an alias for the
225 ** sum(x) expression. If a "%fallback ID OVER" directive were added to
226 ** grammar, then SQLite would always treat "OVER" as an alias, making it
227 ** impossible to call a window-function without a FILTER clause.
228 **
229 ** WINDOW is treated as a keyword if:
230 **
231 **   * the following token is an identifier, or a keyword that can fallback
232 **     to being an identifier, and
233 **   * the token after than one is TK_AS.
234 **
235 ** OVER is a keyword if:
236 **
237 **   * the previous token was TK_RP, and
238 **   * the next token is either TK_LP or an identifier.
239 **
240 ** FILTER is a keyword if:
241 **
242 **   * the previous token was TK_RP, and
243 **   * the next token is TK_LP.
244 */
245 static int analyzeWindowKeyword(const unsigned char *z){
246   int t;
247   t = getToken(&z);
248   if( t!=TK_ID ) return TK_ID;
249   t = getToken(&z);
250   if( t!=TK_AS ) return TK_ID;
251   return TK_WINDOW;
252 }
253 static int analyzeOverKeyword(const unsigned char *z, int lastToken){
254   if( lastToken==TK_RP ){
255     int t = getToken(&z);
256     if( t==TK_LP || t==TK_ID ) return TK_OVER;
257   }
258   return TK_ID;
259 }
260 static int analyzeFilterKeyword(const unsigned char *z, int lastToken){
261   if( lastToken==TK_RP && getToken(&z)==TK_LP ){
262     return TK_FILTER;
263   }
264   return TK_ID;
265 }
266 #endif /* SQLITE_OMIT_WINDOWFUNC */
267 
268 /*
269 ** Return the length (in bytes) of the token that begins at z[0].
270 ** Store the token type in *tokenType before returning.
271 */
272 int sqlite3GetToken(const unsigned char *z, int *tokenType){
273   int i, c;
274   switch( aiClass[*z] ){  /* Switch on the character-class of the first byte
275                           ** of the token. See the comment on the CC_ defines
276                           ** above. */
277     case CC_SPACE: {
278       testcase( z[0]==' ' );
279       testcase( z[0]=='\t' );
280       testcase( z[0]=='\n' );
281       testcase( z[0]=='\f' );
282       testcase( z[0]=='\r' );
283       for(i=1; sqlite3Isspace(z[i]); i++){}
284       *tokenType = TK_SPACE;
285       return i;
286     }
287     case CC_MINUS: {
288       if( z[1]=='-' ){
289         for(i=2; (c=z[i])!=0 && c!='\n'; i++){}
290         *tokenType = TK_SPACE;   /* IMP: R-22934-25134 */
291         return i;
292       }
293       *tokenType = TK_MINUS;
294       return 1;
295     }
296     case CC_LP: {
297       *tokenType = TK_LP;
298       return 1;
299     }
300     case CC_RP: {
301       *tokenType = TK_RP;
302       return 1;
303     }
304     case CC_SEMI: {
305       *tokenType = TK_SEMI;
306       return 1;
307     }
308     case CC_PLUS: {
309       *tokenType = TK_PLUS;
310       return 1;
311     }
312     case CC_STAR: {
313       *tokenType = TK_STAR;
314       return 1;
315     }
316     case CC_SLASH: {
317       if( z[1]!='*' || z[2]==0 ){
318         *tokenType = TK_SLASH;
319         return 1;
320       }
321       for(i=3, c=z[2]; (c!='*' || z[i]!='/') && (c=z[i])!=0; i++){}
322       if( c ) i++;
323       *tokenType = TK_SPACE;   /* IMP: R-22934-25134 */
324       return i;
325     }
326     case CC_PERCENT: {
327       *tokenType = TK_REM;
328       return 1;
329     }
330     case CC_EQ: {
331       *tokenType = TK_EQ;
332       return 1 + (z[1]=='=');
333     }
334     case CC_LT: {
335       if( (c=z[1])=='=' ){
336         *tokenType = TK_LE;
337         return 2;
338       }else if( c=='>' ){
339         *tokenType = TK_NE;
340         return 2;
341       }else if( c=='<' ){
342         *tokenType = TK_LSHIFT;
343         return 2;
344       }else{
345         *tokenType = TK_LT;
346         return 1;
347       }
348     }
349     case CC_GT: {
350       if( (c=z[1])=='=' ){
351         *tokenType = TK_GE;
352         return 2;
353       }else if( c=='>' ){
354         *tokenType = TK_RSHIFT;
355         return 2;
356       }else{
357         *tokenType = TK_GT;
358         return 1;
359       }
360     }
361     case CC_BANG: {
362       if( z[1]!='=' ){
363         *tokenType = TK_ILLEGAL;
364         return 1;
365       }else{
366         *tokenType = TK_NE;
367         return 2;
368       }
369     }
370     case CC_PIPE: {
371       if( z[1]!='|' ){
372         *tokenType = TK_BITOR;
373         return 1;
374       }else{
375         *tokenType = TK_CONCAT;
376         return 2;
377       }
378     }
379     case CC_COMMA: {
380       *tokenType = TK_COMMA;
381       return 1;
382     }
383     case CC_AND: {
384       *tokenType = TK_BITAND;
385       return 1;
386     }
387     case CC_TILDA: {
388       *tokenType = TK_BITNOT;
389       return 1;
390     }
391     case CC_QUOTE: {
392       int delim = z[0];
393       testcase( delim=='`' );
394       testcase( delim=='\'' );
395       testcase( delim=='"' );
396       for(i=1; (c=z[i])!=0; i++){
397         if( c==delim ){
398           if( z[i+1]==delim ){
399             i++;
400           }else{
401             break;
402           }
403         }
404       }
405       if( c=='\'' ){
406         *tokenType = TK_STRING;
407         return i+1;
408       }else if( c!=0 ){
409         *tokenType = TK_ID;
410         return i+1;
411       }else{
412         *tokenType = TK_ILLEGAL;
413         return i;
414       }
415     }
416     case CC_DOT: {
417 #ifndef SQLITE_OMIT_FLOATING_POINT
418       if( !sqlite3Isdigit(z[1]) )
419 #endif
420       {
421         *tokenType = TK_DOT;
422         return 1;
423       }
424       /* If the next character is a digit, this is a floating point
425       ** number that begins with ".".  Fall thru into the next case */
426       /* no break */ deliberate_fall_through
427     }
428     case CC_DIGIT: {
429       testcase( z[0]=='0' );  testcase( z[0]=='1' );  testcase( z[0]=='2' );
430       testcase( z[0]=='3' );  testcase( z[0]=='4' );  testcase( z[0]=='5' );
431       testcase( z[0]=='6' );  testcase( z[0]=='7' );  testcase( z[0]=='8' );
432       testcase( z[0]=='9' );
433       *tokenType = TK_INTEGER;
434 #ifndef SQLITE_OMIT_HEX_INTEGER
435       if( z[0]=='0' && (z[1]=='x' || z[1]=='X') && sqlite3Isxdigit(z[2]) ){
436         for(i=3; sqlite3Isxdigit(z[i]); i++){}
437         return i;
438       }
439 #endif
440       for(i=0; sqlite3Isdigit(z[i]); i++){}
441 #ifndef SQLITE_OMIT_FLOATING_POINT
442       if( z[i]=='.' ){
443         i++;
444         while( sqlite3Isdigit(z[i]) ){ i++; }
445         *tokenType = TK_FLOAT;
446       }
447       if( (z[i]=='e' || z[i]=='E') &&
448            ( sqlite3Isdigit(z[i+1])
449             || ((z[i+1]=='+' || z[i+1]=='-') && sqlite3Isdigit(z[i+2]))
450            )
451       ){
452         i += 2;
453         while( sqlite3Isdigit(z[i]) ){ i++; }
454         *tokenType = TK_FLOAT;
455       }
456 #endif
457       while( IdChar(z[i]) ){
458         *tokenType = TK_ILLEGAL;
459         i++;
460       }
461       return i;
462     }
463     case CC_QUOTE2: {
464       for(i=1, c=z[0]; c!=']' && (c=z[i])!=0; i++){}
465       *tokenType = c==']' ? TK_ID : TK_ILLEGAL;
466       return i;
467     }
468     case CC_VARNUM: {
469       *tokenType = TK_VARIABLE;
470       for(i=1; sqlite3Isdigit(z[i]); i++){}
471       return i;
472     }
473     case CC_DOLLAR:
474     case CC_VARALPHA: {
475       int n = 0;
476       testcase( z[0]=='$' );  testcase( z[0]=='@' );
477       testcase( z[0]==':' );  testcase( z[0]=='#' );
478       *tokenType = TK_VARIABLE;
479       for(i=1; (c=z[i])!=0; i++){
480         if( IdChar(c) ){
481           n++;
482 #ifndef SQLITE_OMIT_TCL_VARIABLE
483         }else if( c=='(' && n>0 ){
484           do{
485             i++;
486           }while( (c=z[i])!=0 && !sqlite3Isspace(c) && c!=')' );
487           if( c==')' ){
488             i++;
489           }else{
490             *tokenType = TK_ILLEGAL;
491           }
492           break;
493         }else if( c==':' && z[i+1]==':' ){
494           i++;
495 #endif
496         }else{
497           break;
498         }
499       }
500       if( n==0 ) *tokenType = TK_ILLEGAL;
501       return i;
502     }
503     case CC_KYWD0: {
504       for(i=1; aiClass[z[i]]<=CC_KYWD; i++){}
505       if( IdChar(z[i]) ){
506         /* This token started out using characters that can appear in keywords,
507         ** but z[i] is a character not allowed within keywords, so this must
508         ** be an identifier instead */
509         i++;
510         break;
511       }
512       *tokenType = TK_ID;
513       return keywordCode((char*)z, i, tokenType);
514     }
515     case CC_X: {
516 #ifndef SQLITE_OMIT_BLOB_LITERAL
517       testcase( z[0]=='x' ); testcase( z[0]=='X' );
518       if( z[1]=='\'' ){
519         *tokenType = TK_BLOB;
520         for(i=2; sqlite3Isxdigit(z[i]); i++){}
521         if( z[i]!='\'' || i%2 ){
522           *tokenType = TK_ILLEGAL;
523           while( z[i] && z[i]!='\'' ){ i++; }
524         }
525         if( z[i] ) i++;
526         return i;
527       }
528 #endif
529       /* If it is not a BLOB literal, then it must be an ID, since no
530       ** SQL keywords start with the letter 'x'.  Fall through */
531       /* no break */ deliberate_fall_through
532     }
533     case CC_KYWD:
534     case CC_ID: {
535       i = 1;
536       break;
537     }
538     case CC_NUL: {
539       *tokenType = TK_ILLEGAL;
540       return 0;
541     }
542     default: {
543       *tokenType = TK_ILLEGAL;
544       return 1;
545     }
546   }
547   while( IdChar(z[i]) ){ i++; }
548   *tokenType = TK_ID;
549   return i;
550 }
551 
552 /*
553 ** Run the parser on the given SQL string.  The parser structure is
554 ** passed in.  An SQLITE_ status code is returned.  If an error occurs
555 ** then an and attempt is made to write an error message into
556 ** memory obtained from sqlite3_malloc() and to make *pzErrMsg point to that
557 ** error message.
558 */
559 int sqlite3RunParser(Parse *pParse, const char *zSql, char **pzErrMsg){
560   int nErr = 0;                   /* Number of errors encountered */
561   void *pEngine;                  /* The LEMON-generated LALR(1) parser */
562   int n = 0;                      /* Length of the next token token */
563   int tokenType;                  /* type of the next token */
564   int lastTokenParsed = -1;       /* type of the previous token */
565   sqlite3 *db = pParse->db;       /* The database connection */
566   int mxSqlLen;                   /* Max length of an SQL string */
567 #ifdef sqlite3Parser_ENGINEALWAYSONSTACK
568   yyParser sEngine;    /* Space to hold the Lemon-generated Parser object */
569 #endif
570   VVA_ONLY( u8 startedWithOom = db->mallocFailed );
571 
572   assert( zSql!=0 );
573   mxSqlLen = db->aLimit[SQLITE_LIMIT_SQL_LENGTH];
574   if( db->nVdbeActive==0 ){
575     AtomicStore(&db->u1.isInterrupted, 0);
576   }
577   pParse->rc = SQLITE_OK;
578   pParse->zTail = zSql;
579   assert( pzErrMsg!=0 );
580 #ifdef SQLITE_DEBUG
581   if( db->flags & SQLITE_ParserTrace ){
582     printf("parser: [[[%s]]]\n", zSql);
583     sqlite3ParserTrace(stdout, "parser: ");
584   }else{
585     sqlite3ParserTrace(0, 0);
586   }
587 #endif
588 #ifdef sqlite3Parser_ENGINEALWAYSONSTACK
589   pEngine = &sEngine;
590   sqlite3ParserInit(pEngine, pParse);
591 #else
592   pEngine = sqlite3ParserAlloc(sqlite3Malloc, pParse);
593   if( pEngine==0 ){
594     sqlite3OomFault(db);
595     return SQLITE_NOMEM_BKPT;
596   }
597 #endif
598   assert( pParse->pNewTable==0 );
599   assert( pParse->pNewTrigger==0 );
600   assert( pParse->nVar==0 );
601   assert( pParse->pVList==0 );
602   pParse->pParentParse = db->pParse;
603   db->pParse = pParse;
604   while( 1 ){
605     n = sqlite3GetToken((u8*)zSql, &tokenType);
606     mxSqlLen -= n;
607     if( mxSqlLen<0 ){
608       pParse->rc = SQLITE_TOOBIG;
609       break;
610     }
611 #ifndef SQLITE_OMIT_WINDOWFUNC
612     if( tokenType>=TK_WINDOW ){
613       assert( tokenType==TK_SPACE || tokenType==TK_OVER || tokenType==TK_FILTER
614            || tokenType==TK_ILLEGAL || tokenType==TK_WINDOW
615       );
616 #else
617     if( tokenType>=TK_SPACE ){
618       assert( tokenType==TK_SPACE || tokenType==TK_ILLEGAL );
619 #endif /* SQLITE_OMIT_WINDOWFUNC */
620       if( AtomicLoad(&db->u1.isInterrupted) ){
621         pParse->rc = SQLITE_INTERRUPT;
622         break;
623       }
624       if( tokenType==TK_SPACE ){
625         zSql += n;
626         continue;
627       }
628       if( zSql[0]==0 ){
629         /* Upon reaching the end of input, call the parser two more times
630         ** with tokens TK_SEMI and 0, in that order. */
631         if( lastTokenParsed==TK_SEMI ){
632           tokenType = 0;
633         }else if( lastTokenParsed==0 ){
634           break;
635         }else{
636           tokenType = TK_SEMI;
637         }
638         n = 0;
639 #ifndef SQLITE_OMIT_WINDOWFUNC
640       }else if( tokenType==TK_WINDOW ){
641         assert( n==6 );
642         tokenType = analyzeWindowKeyword((const u8*)&zSql[6]);
643       }else if( tokenType==TK_OVER ){
644         assert( n==4 );
645         tokenType = analyzeOverKeyword((const u8*)&zSql[4], lastTokenParsed);
646       }else if( tokenType==TK_FILTER ){
647         assert( n==6 );
648         tokenType = analyzeFilterKeyword((const u8*)&zSql[6], lastTokenParsed);
649 #endif /* SQLITE_OMIT_WINDOWFUNC */
650       }else{
651         sqlite3ErrorMsg(pParse, "unrecognized token: \"%.*s\"", n, zSql);
652         break;
653       }
654     }
655     pParse->sLastToken.z = zSql;
656     pParse->sLastToken.n = n;
657     sqlite3Parser(pEngine, tokenType, pParse->sLastToken);
658     lastTokenParsed = tokenType;
659     zSql += n;
660     assert( db->mallocFailed==0 || pParse->rc!=SQLITE_OK || startedWithOom );
661     if( pParse->rc!=SQLITE_OK ) break;
662   }
663   assert( nErr==0 );
664 #ifdef YYTRACKMAXSTACKDEPTH
665   sqlite3_mutex_enter(sqlite3MallocMutex());
666   sqlite3StatusHighwater(SQLITE_STATUS_PARSER_STACK,
667       sqlite3ParserStackPeak(pEngine)
668   );
669   sqlite3_mutex_leave(sqlite3MallocMutex());
670 #endif /* YYDEBUG */
671 #ifdef sqlite3Parser_ENGINEALWAYSONSTACK
672   sqlite3ParserFinalize(pEngine);
673 #else
674   sqlite3ParserFree(pEngine, sqlite3_free);
675 #endif
676   if( db->mallocFailed ){
677     pParse->rc = SQLITE_NOMEM_BKPT;
678   }
679   if( pParse->rc!=SQLITE_OK && pParse->rc!=SQLITE_DONE && pParse->zErrMsg==0 ){
680     pParse->zErrMsg = sqlite3MPrintf(db, "%s", sqlite3ErrStr(pParse->rc));
681   }
682   assert( pzErrMsg!=0 );
683   if( pParse->zErrMsg ){
684     *pzErrMsg = pParse->zErrMsg;
685     sqlite3_log(pParse->rc, "%s in \"%s\"",
686                 *pzErrMsg, pParse->zTail);
687     pParse->zErrMsg = 0;
688     nErr++;
689   }
690   pParse->zTail = zSql;
691   if( pParse->pVdbe && pParse->nErr>0 && pParse->nested==0 ){
692     sqlite3VdbeDelete(pParse->pVdbe);
693     pParse->pVdbe = 0;
694   }
695 #ifndef SQLITE_OMIT_SHARED_CACHE
696   if( pParse->nested==0 ){
697     sqlite3DbFree(db, pParse->aTableLock);
698     pParse->aTableLock = 0;
699     pParse->nTableLock = 0;
700   }
701 #endif
702 #ifndef SQLITE_OMIT_VIRTUALTABLE
703   sqlite3_free(pParse->apVtabLock);
704 #endif
705 
706   if( !IN_SPECIAL_PARSE ){
707     /* If the pParse->declareVtab flag is set, do not delete any table
708     ** structure built up in pParse->pNewTable. The calling code (see vtab.c)
709     ** will take responsibility for freeing the Table structure.
710     */
711     sqlite3DeleteTable(db, pParse->pNewTable);
712   }
713   if( !IN_RENAME_OBJECT ){
714     sqlite3DeleteTrigger(db, pParse->pNewTrigger);
715   }
716   sqlite3DbFree(db, pParse->pVList);
717   db->pParse = pParse->pParentParse;
718   pParse->pParentParse = 0;
719   assert( nErr==0 || pParse->rc!=SQLITE_OK );
720   return nErr;
721 }
722 
723 
724 #ifdef SQLITE_ENABLE_NORMALIZE
725 /*
726 ** Insert a single space character into pStr if the current string
727 ** ends with an identifier
728 */
729 static void addSpaceSeparator(sqlite3_str *pStr){
730   if( pStr->nChar && sqlite3IsIdChar(pStr->zText[pStr->nChar-1]) ){
731     sqlite3_str_append(pStr, " ", 1);
732   }
733 }
734 
735 /*
736 ** Compute a normalization of the SQL given by zSql[0..nSql-1].  Return
737 ** the normalization in space obtained from sqlite3DbMalloc().  Or return
738 ** NULL if anything goes wrong or if zSql is NULL.
739 */
740 char *sqlite3Normalize(
741   Vdbe *pVdbe,       /* VM being reprepared */
742   const char *zSql   /* The original SQL string */
743 ){
744   sqlite3 *db;       /* The database connection */
745   int i;             /* Next unread byte of zSql[] */
746   int n;             /* length of current token */
747   int tokenType;     /* type of current token */
748   int prevType = 0;  /* Previous non-whitespace token */
749   int nParen;        /* Number of nested levels of parentheses */
750   int iStartIN;      /* Start of RHS of IN operator in z[] */
751   int nParenAtIN;    /* Value of nParent at start of RHS of IN operator */
752   u32 j;             /* Bytes of normalized SQL generated so far */
753   sqlite3_str *pStr; /* The normalized SQL string under construction */
754 
755   db = sqlite3VdbeDb(pVdbe);
756   tokenType = -1;
757   nParen = iStartIN = nParenAtIN = 0;
758   pStr = sqlite3_str_new(db);
759   assert( pStr!=0 );  /* sqlite3_str_new() never returns NULL */
760   for(i=0; zSql[i] && pStr->accError==0; i+=n){
761     if( tokenType!=TK_SPACE ){
762       prevType = tokenType;
763     }
764     n = sqlite3GetToken((unsigned char*)zSql+i, &tokenType);
765     if( NEVER(n<=0) ) break;
766     switch( tokenType ){
767       case TK_SPACE: {
768         break;
769       }
770       case TK_NULL: {
771         if( prevType==TK_IS || prevType==TK_NOT ){
772           sqlite3_str_append(pStr, " NULL", 5);
773           break;
774         }
775         /* Fall through */
776       }
777       case TK_STRING:
778       case TK_INTEGER:
779       case TK_FLOAT:
780       case TK_VARIABLE:
781       case TK_BLOB: {
782         sqlite3_str_append(pStr, "?", 1);
783         break;
784       }
785       case TK_LP: {
786         nParen++;
787         if( prevType==TK_IN ){
788           iStartIN = pStr->nChar;
789           nParenAtIN = nParen;
790         }
791         sqlite3_str_append(pStr, "(", 1);
792         break;
793       }
794       case TK_RP: {
795         if( iStartIN>0 && nParen==nParenAtIN ){
796           assert( pStr->nChar>=(u32)iStartIN );
797           pStr->nChar = iStartIN+1;
798           sqlite3_str_append(pStr, "?,?,?", 5);
799           iStartIN = 0;
800         }
801         nParen--;
802         sqlite3_str_append(pStr, ")", 1);
803         break;
804       }
805       case TK_ID: {
806         iStartIN = 0;
807         j = pStr->nChar;
808         if( sqlite3Isquote(zSql[i]) ){
809           char *zId = sqlite3DbStrNDup(db, zSql+i, n);
810           int nId;
811           int eType = 0;
812           if( zId==0 ) break;
813           sqlite3Dequote(zId);
814           if( zSql[i]=='"' && sqlite3VdbeUsesDoubleQuotedString(pVdbe, zId) ){
815             sqlite3_str_append(pStr, "?", 1);
816             sqlite3DbFree(db, zId);
817             break;
818           }
819           nId = sqlite3Strlen30(zId);
820           if( sqlite3GetToken((u8*)zId, &eType)==nId && eType==TK_ID ){
821             addSpaceSeparator(pStr);
822             sqlite3_str_append(pStr, zId, nId);
823           }else{
824             sqlite3_str_appendf(pStr, "\"%w\"", zId);
825           }
826           sqlite3DbFree(db, zId);
827         }else{
828           addSpaceSeparator(pStr);
829           sqlite3_str_append(pStr, zSql+i, n);
830         }
831         while( j<pStr->nChar ){
832           pStr->zText[j] = sqlite3Tolower(pStr->zText[j]);
833           j++;
834         }
835         break;
836       }
837       case TK_SELECT: {
838         iStartIN = 0;
839         /* fall through */
840       }
841       default: {
842         if( sqlite3IsIdChar(zSql[i]) ) addSpaceSeparator(pStr);
843         j = pStr->nChar;
844         sqlite3_str_append(pStr, zSql+i, n);
845         while( j<pStr->nChar ){
846           pStr->zText[j] = sqlite3Toupper(pStr->zText[j]);
847           j++;
848         }
849         break;
850       }
851     }
852   }
853   if( tokenType!=TK_SEMI ) sqlite3_str_append(pStr, ";", 1);
854   return sqlite3_str_finish(pStr);
855 }
856 #endif /* SQLITE_ENABLE_NORMALIZE */
857