Hi Jourquin:

To do this I make a preprocessing of OSM geometries using the ST_SimplifyPreserveTopology Postgis fuinction https://postgis.net/docs/ST_SimplifyPreserveTopology.html

Then you have a new geometry dataset where you can aply your routing algorithms.

Regards

Thanks. I'll give it a try.

If I'm right, ST_SimplifyPreserveTopology simplifies geometries. This is not exactly what I'm looking for. In the illustration given in my first post, each visible node connects (at least) two different polylines. I'm looking for a method that merges a series of successive polylines into a single one. The rule to merge two polylines is basically the fact that they must be connected by a same node and that the latest is not connected to any other polyline. Additional rules, based on the attributes (merge two highways, but not a higway and a secondary road for instance) must be possible. To be simple, a road that is currently stored as 10 successive polylines should be merged into a single polyline...

Thank you for your answer Marcus, but I'm not sure to understand.

Let's take an example: Link 1 goes from node A to node B. Link 2 connects nodes B and C. C is not an intersection. Thus, node B must be deleted, and nodes A and C should be connected by a single geomegry (polyline) representing the geometries of 1 + 2. Indeed, I also need to represent the "merged" link in a GIS.

From what I have understood from your answer, I have to delete node B from the "nodes" table. But what about "update the source/target nodes of the edges" in the example given here, as I have to create either a new "polyline" edge between A and C or "add" the geometry of edge 2 to edge 1 (or 1 to 2) ?

Regards

Hi Marcus,

Thank you for your answer. You are indeed right as long as the routing is done with pgRouting. I realize that my question was not clear enough (now edited). In fact, I'm looking for "simplified" OSM based networks to use in other GIS or routing softwares. In other words, the simplified networks will have to be exported as ESRI shapefiles for instance. That is why I need to keep the geometries...

It could also that my approach (i.e. the creation of a routable network with pgRouting) is not an adequate first step...

Thanks, but this seems to be an algorithm able to "connect" edges which endpoints are not exactly located at the same position. I wrote something similar 20+ years ago when the only usable networks were available from DCW (Digital Chart of the World). The DCW was not routable, and I wrote a series of utilities to convert them.

I think I'll write my own routine to achieve what I want using OSM. Shouldn't be too difficult, but an attention will have to be paid to performance issues when large networks must be processed...

I hope we'll have a couple of rainy days this Summer :-)