-
Olav Sandstaa authored
This worklog adds an interface that storage engines can use for providing records per key estimates using float values. By using float, the records per key estimates used by the optimizer will be more accurate and the rounding issues the current integer based rec_per_key values have can be avoided. This should result in more accurate records per key estimates and improved query plans. The interface consists of the following: -the KEY struct is extended with a new array for storing records per key values using float. This corresponds to the existing rec_per_key array. -new member functions to the KEY struct for setting and getting records per key values. Note that with this worklog, only the interface is introduced. It is not yet used by the optimizer.
Olav Sandstaa authoredThis worklog adds an interface that storage engines can use for providing records per key estimates using float values. By using float, the records per key estimates used by the optimizer will be more accurate and the rounding issues the current integer based rec_per_key values have can be avoided. This should result in more accurate records per key estimates and improved query plans. The interface consists of the following: -the KEY struct is extended with a new array for storing records per key values using float. This corresponds to the existing rec_per_key array. -new member functions to the KEY struct for setting and getting records per key values. Note that with this worklog, only the interface is introduced. It is not yet used by the optimizer.
Loading