[R-sig-Geo] agenda for R's Geo capabilities?
tkeitt at gmail.com
Fri Mar 16 19:54:30 CET 2007
On 3/16/07, elw at stderr.org <elw at stderr.org> wrote:
> > My OSGeo summer of code idea is to link the BGL to postgis. I wasn't
> > originally thinking about topology per se, but it would achieve much the
> > same thing.
> I've thought a bit about this project, too, as it butts up against some of
> my own work.
> My experience with graph libraries has not been altogether positive. ;)
> They are rather difficult to coerce into handling very large data sets.
Our approach is not to store the vertex/edge information in the graph
library, but rather to model it with iterator interfaces. Generic
programming methods make this possible. Essentially, GDAL becomes the
data storage mechanism and the graph is implicit. The limits in this
case are likely the algorithmic complexity rather than data storage
(the step of building a graph data structure is entirely avoided). Of
course there may be better alternatives that are custom built for the
Vertex/edge iterators could also be backed by postgis or any other
data source with sufficient structure to specify adjacency. The
enormous value from my perspective is that effort coding algorithms
becomes reusable in many contexts.
> I think it would be handy to have a layer to pump graph data through BGL
> into a postgis store. It would be even better if that postgis store
> looked something similar to the store used as a backend by the
> Redland/Raptor xml graph serialization bits. :-) Just a suggestion for a
> place to start.... I know the redland/raptor bits are extremely
> scale-friendly for attributed graph data.
I've never heard of it, but like I say, this is pretty far from my core areas.
Timothy H. Keitt, University of Texas at Austin
Contact info and schedule at http://www.keittlab.org/tkeitt/
Reprints at http://www.keittlab.org/tkeitt/papers/
ODF attachment? See http://www.openoffice.org/
More information about the R-sig-Geo