|
|
Posted by drawnai on 04/24/06 11:10
David Portas wrote:
> In your example you try to force a particular execution plan onto your
> code using an INDEX hint. But there is absolutely no reason why SQL
> Server should always be required to implement a hint. Indexes are
> intended as an optimization tool - they are not supposed to affect
> logical behaviour of code.
I _am_ using it as an optimisation tool. I'm getting the results I
want, by forcing the
plan I want. If at some point, index hints stop working (like that's
going to happen,) I'll find some other way of breaking relational
theory.
What would be nice, (if you're listening Mr Gates,) is for you to
modify this mechanic so,
1. You can do it with selects as well.
2. The where clause checks @variable conditions on every row, instead
of just at the beginning.
You really are taking this too seriously. The code works, and it's a
hundred times faster than the next nearest Oracle technique, and the
code has to be checked every now and then anyway, but so what? The cost
savings to market that this achieves more than outweighs the later
potential corrections.
If run forever reliability was the issue, I'm sure we can both agree
that we wouldn't
be using a product or suite of products written in c++ anyway.
Navigation:
[Reply to this message]
|