On Oct 27, 2013, at 5:22 AM, Sandro Santilli wrote:
> On Sat, Oct 26, 2013 at 06:27:41PM -0500, William Kyngesburye wrote:
>> Is there any way to spatially index topogeometry?
>
> Nope, due to the fact that TopoGeometry spatial extent is not
> in the value itself but derived using a volatile function from
> external (referenced) data.
>
> This is a known problem:
>
http://trac.osgeo.org/postgis/ticket/2124
>
http://trac.osgeo.org/postgis/ticket/853
>
> One idea would be to make the && operator be
> implemented by an SQL-defined function to be
> fully inlined and thus giving the planner enough
> hints to use the index on the underlying primitive
> tables, but feasibility of that is uncertain.
>
> Another idea would be to introduce a trigger-based
> mechanism to update some form of bbox cache into
> each TopoGeometry affected by any change into any
> primitive table, and then use that cache in select
> querys. Note that making it transparent (using &&
> operator) would require a change in the TopoGeometry
> structure, to add a bounding box field.
>
> --strk;
So for now, working with a large topology data set is not practical. Some processing should not be a problem, if it operates on the primitives, which are indexed.