How do we want index_read and the like to behave? Currently they're implemented to varying degrees by each engine... and i'm not entirely sure it's even correct when it is. So I've been working on moving the maintenance of the statistics to the upper layer.
I've so far gone through the counts for update, write, and delete. These are pretty simple as there are existing wrapper functions in Cursor. The next are the scans and lookups. This crates the added thing to solve of what about engines that implement rnd_pos differently? InnoDB does it by using the clustered index, embedded_innodb does it relatively similar to InnoDB, but other engines will use file position or rowid. So should we count these differently? Just as index reads? Whichever way we go, we're going to have some new wrapper functions in Cursor. -- Stewart Smith _______________________________________________ Mailing list: https://launchpad.net/~drizzle-discuss Post to : [email protected] Unsubscribe : https://launchpad.net/~drizzle-discuss More help : https://help.launchpad.net/ListHelp
