On Sat, Mar 13, 2010 at 9:57 PM, Brian Moon <[email protected]> wrote: > Being primarily an application developer, I find it off putting that your > answer to this performance issue is to not let anyone ever do it. How do you > page through millions of rows then? We can't rely on auto_incremnt to always > get smaller when paging backwards. Not all data is date ordered. There is > not always a "where keyfield < 12345" available for these types of paging > scenarios. As a "good" developer, I don't do it because MySQL really, really > sucks at it. But, if there is a solution that can help large offsets, I > would think the ideals behind Drizzle would be to fix what MySQL screwed up, > not to disallow the feature. > > Shouldn't the first response to any problem be "Yes, that is a problem. How > can we fix the problem?"
The customer isn't always right. My response was that efficient support for this isn't possible and then I published a note to explain why. Do you have any comments on that note? http://www.facebook.com/note.php?note_id=206034210932 This is not a problem with MySQL. No general-purpose RDBMS has efficient support for queries like this. Pagination is O(N*N). Only the constant factor in the cost of processing this query is changed if Yuan's request is supported. It is still O(N*N). -- Mark Callaghan [email protected] _______________________________________________ Mailing list: https://launchpad.net/~drizzle-discuss Post to : [email protected] Unsubscribe : https://launchpad.net/~drizzle-discuss More help : https://help.launchpad.net/ListHelp
