>>>>> "Joshua" == Joshua Chamas <joshua@stripped> writes:
Joshua> Monty & MySQL hacking team,
Joshua> Hate to answer my own question, but it looks like this might
Joshua> have been recently implemented in 3.23.9, with this note:
Joshua> Changed optimizer to prefer a range key instead of a ref key when the
> range key can
Joshua> uses more columns than the ref key (which only can use columns with =).
Joshua> example, the following type of queries should now be faster: SELECT *
Joshua> key_part_1=const and key_part_2 > const2
Joshua> Is this right, will this cover my case, where it seems that
Joshua> 'const' keys, like PRIMARY or UNIQUE seem to override other
Joshua> indexes with more & better keys ?
Yes; If h.next_hit is part of your key, the above new optimization may help.
Joshua> Note that I have not tried 3.23.9 yet because there is
Joshua> no WinNT binary version available yet for download, and
Joshua> I am now in pre-production testing on NT, so no more Unix
Joshua> source building for me :(
I hope to have a 3.23.9 ready for testing before tomorrow mid day.
(If not, it has to wait one week before I am back from Tel Aviv).
Anyway, do you think you at some point in the future could try the
queries on Linux? I would really like to know if the file caching is
as good on NT as on Linux; According to all my tests it seams that
Linux is 30 % faster when it comes to MySQL, but it would be nice if
someone else also could verify this...