On 12/20/2011 08:45 AM, Peter Wendorff wrote:
Hi. I'm working on the look-and-listen-map (toolserver project lalm), a web map portal for the Blind and Visually Impaired. That is not a rendering application like the several mapnik stylesheets running on the toolserver, but I agreed with the Admins, that it might be worth in respect of capacities of the toolserver to use the mapnik style nevertheless.
Today I finally got database access and looked deeper into the schema. Most queries in my application rely on the osm_id present in all tables (and index, too), so that's not a problem. Some queries will later on require geometric queries, but here I can combine the postprocessed tables (planet_nodes, planet_line, planet_polygon...) for filtering and the non-processed ones for fetching the values - so that's okay, too.
My problem lies, where mapnik does ugly tricks for rendering: I have a node-id and want to get all ways this node is part of.
That would be possible, querying planet_ways, as planet_ways contains a field nodes, that is a integer-array; but as far as I know there is no performant way to query an array-field of a database on containing a particular element.
Currently my code is based on a different database schema, but I have a query with the following characteristics, that I need with the osm_mapnik-scheme:
IN: node_id (int) OUT: a list of way_id, where the corresponding way contains the node with id node_id.
The mapnik updater has to do that as well. As a moved node needs to recreate the geometry of the way that contains that node.
There is a GIN index on the planet_ways table that allows to do this.
The query osm2pgsql runs for this is "UPDATE %p_ways SET pending = true WHERE nodes && ARRAY[$1] AND NOT pending;"
You don't want the update or pending part, but the "WHERE nodes && ARRAY[osm_id]" should work for you.
A second question would be, if errors would occur, when I combine e.g. planet_ways with planet_line in a query, and a matching row in planet_ways is pending. Usually the intermediate tables are only accessed by the import tool, so I would do unconventional things here, but am I right? If the planet_ways-row is pending, the data of this table is not in sync with the data in planet_line.
Yes, that is correct. But that should only be for a fairly short period of time.
Kai
regards Peter
Maps-l mailing list Maps-l@lists.wikimedia.org https://lists.wikimedia.org/mailman/listinfo/maps-l