[postgis-users] Histogram2d formation

David Blasby dblasby at refractions.net
Thu Oct 10 10:23:05 PDT 2002

The results for the "bigger" bounding boxes are in.

150,000 boxes were made with side between 10k and 500km. 95,985 of these boxes
actually hit features (avg of 5324/box).

The average error for the 150,000 boxes was 257, and the avg error not including the
zero-error estimates was 392.  The largest error was 31,555 (actual was 66,162).

%99 of errors were <6,955  95% were < 1,355 and 90% were < 691

I have the new histogram estimator working on my system - unfortunately the postgresql
index-cost-computer is incorrectly estimating the cost of  the GiST index vs
sequential scanning.

I'm looking into it now, but its pretty deep mojo!


More information about the postgis-users mailing list