[postgis-devel] GiST JoinSel Bogus?

[Sandro Santilli]

On Tue, Nov 20, 2012 at 09:31:16AM -0800, Paul Ramsey wrote:

> constant) but not yet the ideal. The current simple selective code is
> close to the ideal, the only thing left to improve is this nagging
> ticket:
> 
> http://trac.osgeo.org/postgis/ticket/1828
> 
> Which, incidentally, despite some great discussion on the ticket, we
> don't really have an answer for. The problem isn't the selectivity
> *calculation*, which is fine, it's the interface with PostgreSQL and
> ensuring a useful answer for the <col> && <func> case. I wish there
> was some way we could return "I dunno" on that case.

The only well tested and close to ideal is really <col> && <constant>.
<col> && <func> is that other ticket.
Let's focus on <col1> && <col2> in this thread (is already complex enough).

--strk;