Ann Harrison [2012-01-02 18:26] :

> Obviously if there's more than one secondary index that could be used, use it
> the same way. So you start with all the records numbers in New York City,
> drop out all the ones that cost less than $500,000 or more than $1,500,000,
> drop out all the ones that have fewer than 3 or more than 6 bedrooms, etc.
>

interesting
if it could help us to have more efficient range queries, without adding
some R-Tree index

for you information I have a test case
(http://firebird.svn.sourceforge.net/viewvc/firebird/qa/trunk/benchmark/packinginterval.py)
for packing interval
this bench have been written after reading this article (sorry in
French, but understandable I guess) Inspired by article on the web (Fred
Bouard)
http://blog.developpez.com/sqlpro/p9821/langage-sql-norme/agregation-d-intervalles-en-sql-1/

I have some others queries about interval that can be use as bench on my box