Brian Aker wrote: > Hi! > > On Sep 16, 2008, at 10:15 PM, Jay Pipes wrote: > >> sort (ORDER BY) or compare (WHERE) values in a column based on a > > What if you do not have an index but you still want to call ORDERY BY? > To me the collation is a property of the column... and possibly we need > one on the index as well (in the case of INDEX(A,B).
I answered that in my original email (which you snipped out). There is no reason to store a collation for a column, since collation is for sorting and searching. If a user is sorting or searching on a field that is not contained in an index, a filesort is needed anyway (just like it is in MySQL today). In this case, the filesort is performed using the qsort compare function pointer in the COLLATION_INFO pointer of the utf8 charset. In the case of a field that does have an index on it, the filesort would not need to be performed if the requested collation is what the index is stored as. If it isn't, same as above. Pull index keys into a temporary area and issue a sort with the appropriate collation's qsort comparison function. The solution I am proposing means that multiple indexes could be constructed on the same field, but having different collations. -jay _______________________________________________ Mailing list: https://launchpad.net/~drizzle-discuss Post to : [email protected] Unsubscribe : https://launchpad.net/~drizzle-discuss More help : https://help.launchpad.net/ListHelp
