/*
** Parameters z and n contain a pointer to and length of a buffer containing
** an fts3 query expression, respectively. This function attempts to parse the
** query expression and create a tree of Fts3Expr structures representing the
** parsed expression. If successful, *ppExpr is set to point to the head
** of the parsed expression tree and SQLITE_OK is returned. If an error
** occurs, either SQLITE_NOMEM (out-of-memory error) or SQLITE_ERROR (parse
** error) is returned and *ppExpr is set to 0.
**
** If parameter n is a negative number, then z is assumed to point to a
** nul-terminated string and the length is determined using strlen().
**
** The first parameter, pTokenizer, is passed the fts3 tokenizer module to
** use to normalize query tokens while parsing the expression. The azCol[]
** array, which is assumed to contain nCol entries, should contain the names
** of each column in the target fts3 table, in order from left to right.
** Column names must be nul-terminated strings.
**
** The iDefaultCol parameter should be passed the index of the table column
** that appears on the left-hand-side of the MATCH operator (the default
** column to match against for tokens for which a column name is not explicitly
** specified as part of the query string), or -1 if tokens may by default
** match any table column.
*/
int sqlite3Fts3ExprParse(
sqlite3_tokenizer *pTokenizer, /* Tokenizer module */
char **azCol, /* Array of column names for fts3 table */
int nCol, /* Number of entries in azCol[] */
int iDefaultCol, /* Default column to query */
const char *z, int n, /* Text of MATCH query */
Fts3Expr **ppExpr /* OUT: Parsed query structure */
){
int nParsed;
int rc;
ParseContext sParse;
sParse.pTokenizer = pTokenizer;
sParse.azCol = (const char **)azCol;
sParse.nCol = nCol;
sParse.iDefaultCol = iDefaultCol;
sParse.nNest = 0;
if( z==0 ){
*ppExpr = 0;
return SQLITE_OK;
}
if( n<0 ){
n = (int)strlen(z);
}
rc = fts3ExprParse(&sParse, z, n, ppExpr, &nParsed);
/* Check for mismatched parenthesis */
if( rc==SQLITE_OK && sParse.nNest ){
rc = SQLITE_ERROR;
sqlite3Fts3ExprFree(*ppExpr);
*ppExpr = 0;
}
return rc;
}
/*
** This function is similar to sqlite3Fts3ExprParse(), with the following
** differences:
**
** 1. It does not do expression rebalancing.
** 2. It does not check that the expression does not exceed the
** maximum allowable depth.
** 3. Even if it fails, *ppExpr may still be set to point to an
** expression tree. It should be deleted using sqlite3Fts3ExprFree()
** in this case.
*/
static int fts3ExprParseUnbalanced(
sqlite3_tokenizer *pTokenizer, /* Tokenizer module */
int iLangid, /* Language id for tokenizer */
char **azCol, /* Array of column names for fts3 table */
int bFts4, /* True to allow FTS4-only syntax */
int nCol, /* Number of entries in azCol[] */
int iDefaultCol, /* Default column to query */
const char *z, int n, /* Text of MATCH query */
Fts3Expr **ppExpr /* OUT: Parsed query structure */
){
int nParsed;
int rc;
ParseContext sParse;
memset(&sParse, 0, sizeof(ParseContext));
sParse.pTokenizer = pTokenizer;
sParse.iLangid = iLangid;
sParse.azCol = (const char **)azCol;
sParse.nCol = nCol;
sParse.iDefaultCol = iDefaultCol;
sParse.bFts4 = bFts4;
if( z==0 ){
*ppExpr = 0;
return SQLITE_OK;
}
if( n<0 ){
n = (int)strlen(z);
}
rc = fts3ExprParse(&sParse, z, n, ppExpr, &nParsed);
assert( rc==SQLITE_OK || *ppExpr==0 );
/* Check for mismatched parenthesis */
if( rc==SQLITE_OK && sParse.nNest ){
rc = SQLITE_ERROR;
}
return rc;
}
/*
** The output variable *ppExpr is populated with an allocated Fts3Expr
** structure, or set to 0 if the end of the input buffer is reached.
**
** Returns an SQLite error code. SQLITE_OK if everything works, SQLITE_NOMEM
** if a malloc failure occurs, or SQLITE_ERROR if a parse error is encountered.
** If SQLITE_ERROR is returned, pContext is populated with an error message.
*/
static int getNextNode(
ParseContext *pParse, /* fts3 query parse context */
const char *z, int n, /* Input string */
Fts3Expr **ppExpr, /* OUT: expression */
int *pnConsumed /* OUT: Number of bytes consumed */
){
static const struct Fts3Keyword {
char *z; /* Keyword text */
unsigned char n; /* Length of the keyword */
unsigned char parenOnly; /* Only valid in paren mode */
unsigned char eType; /* Keyword code */
} aKeyword[] = {
{ "OR" , 2, 0, FTSQUERY_OR },
{ "AND", 3, 1, FTSQUERY_AND },
{ "NOT", 3, 1, FTSQUERY_NOT },
{ "NEAR", 4, 0, FTSQUERY_NEAR }
};
int ii;
int iCol;
int iColLen;
int rc;
Fts3Expr *pRet = 0;
const char *zInput = z;
int nInput = n;
/* Skip over any whitespace before checking for a keyword, an open or
** close bracket, or a quoted string.
*/
while( nInput>0 && fts3isspace(*zInput) ){
nInput--;
zInput++;
}
if( nInput==0 ){
return SQLITE_DONE;
}
/* See if we are dealing with a keyword. */
for(ii=0; ii<(int)(sizeof(aKeyword)/sizeof(struct Fts3Keyword)); ii++){
const struct Fts3Keyword *pKey = &aKeyword[ii];
if( (pKey->parenOnly & ~sqlite3_fts3_enable_parentheses)!=0 ){
continue;
}
if( nInput>=pKey->n && 0==memcmp(zInput, pKey->z, pKey->n) ){
int nNear = SQLITE_FTS3_DEFAULT_NEAR_PARAM;
int nKey = pKey->n;
char cNext;
/* If this is a "NEAR" keyword, check for an explicit nearness. */
if( pKey->eType==FTSQUERY_NEAR ){
assert( nKey==4 );
if( zInput[4]=='/' && zInput[5]>='0' && zInput[5]<='9' ){
nNear = 0;
for(nKey=5; zInput[nKey]>='0' && zInput[nKey]<='9'; nKey++){
nNear = nNear * 10 + (zInput[nKey] - '0');
}
}
}
/* At this point this is probably a keyword. But for that to be true,
** the next byte must contain either whitespace, an open or close
** parenthesis, a quote character, or EOF.
*/
cNext = zInput[nKey];
if( fts3isspace(cNext)
|| cNext=='"' || cNext=='(' || cNext==')' || cNext==0
){
pRet = (Fts3Expr *)fts3MallocZero(sizeof(Fts3Expr));
if( !pRet ){
return SQLITE_NOMEM;
}
pRet->eType = pKey->eType;
pRet->nNear = nNear;
*ppExpr = pRet;
*pnConsumed = (int)((zInput - z) + nKey);
return SQLITE_OK;
}
/* Turns out that wasn't a keyword after all. This happens if the
** user has supplied a token such as "ORacle". Continue.
*/
}
}
/* Check for an open bracket. */
if( sqlite3_fts3_enable_parentheses ){
if( *zInput=='(' ){
int nConsumed;
pParse->nNest++;
rc = fts3ExprParse(pParse, &zInput[1], nInput-1, ppExpr, &nConsumed);
if( rc==SQLITE_OK && !*ppExpr ){
rc = SQLITE_DONE;
}
*pnConsumed = (int)((zInput - z) + 1 + nConsumed);
return rc;
}
/* Check for a close bracket. */
if( *zInput==')' ){
pParse->nNest--;
*pnConsumed = (int)((zInput - z) + 1);
return SQLITE_DONE;
}
}
/* See if we are dealing with a quoted phrase. If this is the case, then
** search for the closing quote and pass the whole string to getNextString()
** for processing. This is easy to do, as fts3 has no syntax for escaping
** a quote character embedded in a string.
*/
if( *zInput=='"' ){
for(ii=1; ii<nInput && zInput[ii]!='"'; ii++);
*pnConsumed = (int)((zInput - z) + ii + 1);
if( ii==nInput ){
return SQLITE_ERROR;
}
return getNextString(pParse, &zInput[1], ii-1, ppExpr);
}
/* If control flows to this point, this must be a regular token, or
** the end of the input. Read a regular token using the sqlite3_tokenizer
** interface. Before doing so, figure out if there is an explicit
** column specifier for the token.
**
** TODO: Strangely, it is not possible to associate a column specifier
** with a quoted phrase, only with a single token. Not sure if this was
** an implementation artifact or an intentional decision when fts3 was
** first implemented. Whichever it was, this module duplicates the
** limitation.
*/
iCol = pParse->iDefaultCol;
iColLen = 0;
for(ii=0; ii<pParse->nCol; ii++){
const char *zStr = pParse->azCol[ii];
int nStr = (int)strlen(zStr);
if( nInput>nStr && zInput[nStr]==':'
&& sqlite3_strnicmp(zStr, zInput, nStr)==0
){
iCol = ii;
iColLen = (int)((zInput - z) + nStr + 1);
break;
}
}
rc = getNextToken(pParse, iCol, &z[iColLen], n-iColLen, ppExpr, pnConsumed);
*pnConsumed += iColLen;
return rc;
}
