[postgis-devel] possible st_estimated_extent improvement: use root of R-tree

[David Fuhry]

It seems to me that the accuracy of st_estimated_extent could be 
improved by looking at the root entry of the column's spatial index, if 
one exists.  In that case, the result would be nearly exact, save only 
double => float roundoff error.

Except in the case of deleted outlier entries without a VACUUM or 
REINDEX, of course.  But that case would be no worse than using 
statistics, as is currently done.

In any case, it should be a cheap lookup.

Thoughts?

-Dave