database/sqlite
1
0
Fork 0
mirror of https://github.com/sqlite/sqlite.git synced 2025-11-16 23:02:26 +03:00
Code Activity

Refactor the sqlite3_normalized_sql() implementation. This is a

Browse Source 
work-in-progress. There are still issues.

FossilOrigin-Name: a4c890b0af9786295e6df05022009d8946550adb873535c610be805c2b7a4083
This commit is contained in:
drh
2018-12-10 16:00:57 +00:00
parent fb34559eb1
commit 643d855da9
7 changed files with 148 additions and 294 deletions
Download Patch File Download Diff File Expand all files Collapse all files

198
src/prepare.c
View File

@@ -709,204 +709,6 @@ static int sqlite3LockAndPrepare(
return rc;
}
#ifdef SQLITE_ENABLE_NORMALIZE
/*
** Attempt to estimate the final output buffer size needed for the fully
** normalized version of the specified SQL string. This should take into
** account any potential expansion that could occur (e.g. via IN clauses
** being expanded, etc). This size returned is the total number of bytes
** including the NUL terminator.
*/
static int estimateNormalizedSize(
const char *zSql, /* The original SQL string */
int nSql /* Length of original SQL string */
){
int nOut = nSql + 4;
const char *z = zSql;
while( nOut<nSql*5 ){
while( z[0]!=0 && z[0]!='I' && z[0]!='i' ){ z++; }
if( z[0]==0 ) break;
z++;
if( z[0]!='N' && z[0]!='n' ) break;
z++;
while( sqlite3Isspace(z[0]) ){ z++; }
if( z[0]!='(' ) break;
z++;
nOut += 5; /* ?,?,? */
}
return nOut;
}
/*
** Copy the current token into the output buffer while dealing with quoted
** identifiers. By default, all letters will be converted into lowercase.
** If the bUpper flag is set, uppercase will be used. The piOut argument
** will be used to update the target index into the output string.
*/
static void copyNormalizedToken(
const char *zSql, /* The original SQL string */
int iIn, /* Current index into the original SQL string */
int nToken, /* Number of bytes in the current token */
int tokenFlags, /* Flags returned by the tokenizer */
char *zOut, /* The output string */
int *piOut /* Pointer to target index into the output string */
){
int bQuoted = tokenFlags & SQLITE_TOKEN_QUOTED;
int bKeyword = tokenFlags & SQLITE_TOKEN_KEYWORD;
int j = *piOut, k = 0;
for(; k<nToken; k++){
if( bQuoted ){
if( k==0 && iIn>0 ){
zOut[j++] = '"';
continue;
}else if( k==nToken-1 ){
zOut[j++] = '"';
continue;
}
}
if( bKeyword ){
zOut[j++] = sqlite3Toupper(zSql[iIn+k]);
}else{
zOut[j++] = sqlite3Tolower(zSql[iIn+k]);
}
}
*piOut = j;
}
/*
** Compute a normalization of the SQL given by zSql[0..nSql-1]. Return
** the normalization in space obtained from sqlite3DbMalloc(). Or return
** NULL if anything goes wrong or if zSql is NULL.
*/
char *sqlite3Normalize(
Vdbe *pVdbe, /* VM being reprepared */
const char *zSql, /* The original SQL string */
int nSql /* Size of the input string in bytes */
){
sqlite3 *db; /* Database handle. */
char *z; /* The output string */
int nZ; /* Size of the output string in bytes */
int i; /* Next character to read from zSql[] */
int j; /* Next character to fill in on z[] */
int tokenType = 0; /* Type of the next token */
int prevTokenType = 0; /* Type of the previous token, except spaces */
int n; /* Size of the next token */
int nParen = 0; /* Nesting level of parenthesis */
int iStartIN = 0; /* Start of RHS of IN operator in z[] */
int nParenAtIN = 0; /* Value of nParent at start of RHS of IN operator */
db = sqlite3VdbeDb(pVdbe);
assert( db!=0 );
if( zSql==0 ) return 0;
nZ = estimateNormalizedSize(zSql, nSql);
z = sqlite3DbMallocRawNN(db, nZ);
if( z==0 ) goto normalizeError;
for(i=j=0; i<nSql && zSql[i]; i+=n){
int flags = 0;
if( tokenType!=TK_SPACE ) prevTokenType = tokenType;
n = sqlite3GetTokenNormalized((unsigned char*)zSql+i, &tokenType, &flags);
switch( tokenType ){
case TK_SPACE: {
break;
}
case TK_ILLEGAL: {
goto normalizeError;
}
case TK_STRING:
case TK_INTEGER:
case TK_FLOAT:
case TK_VARIABLE:
case TK_BLOB: {
z[j++] = '?';
break;
}
case TK_LP:
case TK_RP: {
if( tokenType==TK_LP ){
nParen++;
if( prevTokenType==TK_IN ){
iStartIN = j;
nParenAtIN = nParen;
}
}else{
if( iStartIN>0 && nParen==nParenAtIN ){
assert( iStartIN+6<nZ );
memcpy(z+iStartIN+1, "?,?,?", 5);
j = iStartIN+6;
assert( nZ-1-j>=0 );
assert( nZ-1-j<nZ );
memset(z+j, 0, nZ-1-j);
iStartIN = 0;
}
nParen--;
}
assert( nParen>=0 );
/* Fall through */
}
case TK_MINUS:
case TK_SEMI:
case TK_PLUS:
case TK_STAR:
case TK_SLASH:
case TK_REM:
case TK_EQ:
case TK_LE:
case TK_NE:
case TK_LSHIFT:
case TK_LT:
case TK_RSHIFT:
case TK_GT:
case TK_GE:
case TK_BITOR:
case TK_CONCAT:
case TK_COMMA:
case TK_BITAND:
case TK_BITNOT:
case TK_DOT:
case TK_IN:
case TK_IS:
case TK_NOT:
case TK_NULL:
case TK_ID: {
if( tokenType==TK_NULL ){
if( prevTokenType==TK_IS || prevTokenType==TK_NOT ){
/* NULL is a keyword in this case, not a literal value */
}else{
/* Here the NULL is a literal value */
z[j++] = '?';
break;
}
}
if( j>0 && sqlite3IsIdChar(z[j-1]) && sqlite3IsIdChar(zSql[i]) ){
z[j++] = ' ';
}
if( tokenType==TK_ID ){
if( zSql[i]=='"'
&& sqlite3VdbeUsesDoubleQuotedString(db,pVdbe,zSql+i,n)
){
z[j++] = '?';
break;
}
if( nParen==nParenAtIN ) iStartIN = 0;
}
copyNormalizedToken(zSql, i, n, flags, z, &j);
break;
}
}
}
assert( j<nZ && "one" );
while( j>0 && z[j-1]==' ' ){ j--; }
if( j>0 && z[j-1]!=';' ){ z[j++] = ';'; }
z[j] = 0;
assert( j<nZ && "two" );
return z;
normalizeError:
sqlite3DbFree(db, z);
return 0;
}
#endif /* SQLITE_ENABLE_NORMALIZE */
/*
** Rerun the compilation of a statement after a schema change.