Many hyperlinks are disabled.
Use anonymous login
to enable hyperlinks.
Overview
| Comment: | This is part of the previous check-in that didn't get committed for some reason - probably because I failed to press the "Save" button on the text editor. |
|---|---|
| Downloads: | Tarball | ZIP archive |
| Timelines: | family | ancestors | descendants | both | trunk |
| Files: | files | file ages | folders |
| SHA3-256: |
51d32173ff1f610b5d75a50f20fa1e0e |
| User & Date: | drh 2019-05-08 23:53:50.194 |
Context
|
2019-05-08
| ||
| 23:55 | Ensure that the sqlite3BtreeLast() interface initializes the *pRes return to zero if the cursor is already sitting on the last entry of the table. Manuel Rigger found the obscure test case for which this makes a difference. (check-in: ebe4845cd0 user: drh tags: trunk) | |
| 23:53 | This is part of the previous check-in that didn't get committed for some reason - probably because I failed to press the "Save" button on the text editor. (check-in: 51d32173ff user: drh tags: trunk) | |
| 21:14 | Add another test case to cover a previously uncovered branch in the RBU module. (check-in: 3039298561 user: dan tags: trunk) | |
Changes
Changes to src/whereexpr.c.
| ︙ | ︙ | |||
258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 |
char *zNew = pPrefix->u.zToken;
zNew[cnt] = 0;
for(iFrom=iTo=0; iFrom<cnt; iFrom++){
if( zNew[iFrom]==wc[3] ) iFrom++;
zNew[iTo++] = zNew[iFrom];
}
zNew[iTo] = 0;
/* If the RHS begins with a digit or a +/- sign, then the LHS must be
** an ordinary column (not a virtual table column) with TEXT affinity.
** Otherwise the LHS might be numeric and "lhs >= rhs" would be false
** even though "lhs LIKE rhs" is true. But if the RHS does not start
** with a digit or +/-, then "lhs LIKE rhs" will always be false if
** the LHS is numeric and so the optimization still works.
**
** 2018-09-10 ticket c94369cae9b561b1f996d0054bfab11389f9d033
** The RHS pattern must not be '/%' because the termination condition
** will then become "x<'0'" and if the affinity is numeric, will then
** be converted into "x<0", which is incorrect.
*/
if( sqlite3Isdigit(zNew[0])
|| zNew[0]=='-'
|| zNew[0]=='+'
| > | | 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 |
char *zNew = pPrefix->u.zToken;
zNew[cnt] = 0;
for(iFrom=iTo=0; iFrom<cnt; iFrom++){
if( zNew[iFrom]==wc[3] ) iFrom++;
zNew[iTo++] = zNew[iFrom];
}
zNew[iTo] = 0;
assert( iTo>0 );
/* If the RHS begins with a digit or a +/- sign, then the LHS must be
** an ordinary column (not a virtual table column) with TEXT affinity.
** Otherwise the LHS might be numeric and "lhs >= rhs" would be false
** even though "lhs LIKE rhs" is true. But if the RHS does not start
** with a digit or +/-, then "lhs LIKE rhs" will always be false if
** the LHS is numeric and so the optimization still works.
**
** 2018-09-10 ticket c94369cae9b561b1f996d0054bfab11389f9d033
** The RHS pattern must not be '/%' because the termination condition
** will then become "x<'0'" and if the affinity is numeric, will then
** be converted into "x<0", which is incorrect.
*/
if( sqlite3Isdigit(zNew[0])
|| zNew[0]=='-'
|| zNew[0]=='+'
|| zNew[iTo-1]=='0'-1
){
if( pLeft->op!=TK_COLUMN
|| sqlite3ExprAffinity(pLeft)!=SQLITE_AFF_TEXT
|| IsVirtual(pLeft->y.pTab) /* Value might be numeric */
){
sqlite3ExprDelete(db, pPrefix);
sqlite3ValueFree(pVal);
|
| ︙ | ︙ |