> >After the last select, you'll have the "0" values first, then the
> >negative ones, then the positive ones...
>
> That's as expected. The zero isn't zero, it's a very small positive
> number, e.g. 0.00000000000001. When you negate it, it becomes a
> very small negative number, e.g. -0.00000000000001 - so it's smaller
> than all of the other negative numbers.

Huh? This part goes against my Norwegian brain as well, not just
Bjørges (although I admit that checking for = 0 is stupid with floats)
. Are you saying that a sorted list should create the following order:

-0.00000000000001
-2.5
-1.2
2.3
76.3

whereas if it was a positive 0 the order would be

-2.5
-1.2
0.00000000000001
2.3
76.3

Or am I missing something?

Set