On Thu, Apr 19, 2012 at 08:47:51AM -0400, Stephen Frost wrote:
> * Sandro Santilli (strk@keybit.net) wrote:
> > Actually a random sample would really be representative of the data
> > distribution. What the type analyzer gets is a sample and that sample
> > is what the estimator looks at to answer the question:
>
> That might work if all you have is point data, but lines, polygons, etc,
> you're typically going to want to see, just not at the same resolution..
> At least, when you're talking about 'zoom-out' tiles, which is what this
> was about up thread.
>
> > I'm looking for a way to fetch random samples these days so I confirm
> > the need for a quick way to fetch the same sample that "analyze"
> > command fetches but at SQL level.
>
> I'm all for supporting that and implementing this feature, I just don't
> think it's going to be all that useful for zoom-out tiles when complex
> geometries are involved.
Me neither. But for points it sounds very useful.
And we know it is useful for lines and polygons as well when it comes
to estimate overlaps... (since the estimator does a good job even for
lines and polygons)
I really hope Neil Conway work of 2007 could make it into PostgreSQL.
Look, the same work was a topic of an homework assignment at Berkley in
2005: http://inst.eecs.berkeley.edu/~cs186/fa05/hw/hw2/hw2.html
And the whole thing is in the SQL standard 2003
--strk;
,------o-. | __/ | Delivering high quality PostGIS 2.0 ! | / 2.0 | http://strk.keybit.net -
http://vizzuality.com`-o------'