On Fri, Jan 21, 2011 at 4:44 AM, Dotan Cohen <dotanco...@gmail.com> wrote:

> Then I would have to check what values are available when inserting,
> and possibly normalise every so often. I'll think about that, and when
> I have enough data in the database I'll set up a test system to play
> with the possibility.

Yes, it's a much more complicated process, and one that I wouldn't enter
into without verifying there was a performance problem. The neat thing is
you could do this at any point as you can renumber the left and right values
at will--as long as you don't *reorder* them.

> I see, thanks. Good point about making sure that the problem exists
before trying to fix it, I've seen people optimise away where there is
> no bottleneck.

People who love to work on complicated problems often fall victim to
premature optimization because it's fun to do. I was guilty of this myself
earlier in my career, but as CPUs and memory and disk speeds increased over
the decades, it's harder to justify it. At least back then the optimization
that wasn't *necessary* still had a good payoff. Nowadays you can waste days
shaving 10ms off of an operation that takes 1s. Congratulations, you've just
poured $$ down the drain. But boy wasn't that fun! :)

Good luck with the implementation. I'd love to hear how it pans out once
you're done, and I'm sure the list would find it useful as well.


Reply via email to