Stewart Smith wrote:
On Sun, Sep 14, 2008 at 09:11:14AM -0400, Jim Starkey wrote:
Per index is reasonable, but the case where the declare collation for
the index is different from the effective collation of the connection.
They the collations are different, the index can be used for equality
retrievals (99% of all retrievals anyway) but not for range retrieval or
for optimizing a sort.
I think this is the way it has to work.
Which is starting to sound like a relatively sane way of doing things.
Of course, this means somebody will come up with multi-collation indexes
or something - just so we can make the sql syntax a little bit more
incomprehensible. :)
Oh, no. It just means that some poor slob will have to define otherwise
redundant indexes with different collations the optimizer will have to
be smart enough to pick the right one.
_______________________________________________
Mailing list: https://launchpad.net/~drizzle-discuss
Post to : [email protected]
Unsubscribe : https://launchpad.net/~drizzle-discuss
More help : https://help.launchpad.net/ListHelp
