Trent Nelson wrote on Wed, Mar 14, 2012 at 15:05:20 -0700:
> On 3/14/12 3:40 PM, "Daniel Shahaf" <danielsh_at_elego.de> wrote:
>
> >Philip: I recalled last year's discussions about implied indexes, but
> >between Trent's reported observations on IRC and a back-of-the-envelope
> >test with sqlite3(1) I was led to believe that an implied index does not
> >get created for in this case (due to the TEXT column, as my comment
> >says).
> >
> >I'm more than happy to revert this on trunk (if it hasn't been already)
> >assuming it's indeed superfluous.
> >
> >Trent -- have you looked into things from your end yet? Can you confirm
> >or deny the hypothesis that the explicit INDEX was necessary in your
> >environment?
>
> Try as I might, I can't reproduce the original behavior that set us off
> down this superfluous index route. +1 to revert; I was wrong.
>
Done.
Thanks to Philip for spotting the bug.
Daniel
> Trent.
>
Received on 2012-03-14 23:18:11 CET