Re: Finding matching ranges

In article <87iti6of5u.fsf_at_altavista.net>, Martin Drautzburg says...
>
>
>If you index the value (l+h)/2 then all matching ranges (plus some
>others) will certainly satisfy
>
> x/2 < (l+h)/2 < (M+x)/2
>
>This is a simple range scan of (l+h)/2.

This inequality certainly depends upon your earlier assumption

>If you assume a minimum value of 0 and a maximum value M then the RM
>problem can be written as
>
> where 0 < l < x
> and x < h < M
>
>(with l,h being the range boundaries in the table and x the value to
>match).

Intervals matching on the other hand must be invariant against linear transformations in general and horizontal shifts in particular. This makes your approach look suspicious. Received on Sun Jul 22 2001 - 01:30:26 CEST