Imports OpenStreetMap data from a binary PBF file format into PostgreSQL in a way which allows to do a topological routing.
Program was created because osm2pgrouting (C++) was choking on a full country import. For a single Masovian voivodeship: osm2pgrouting uses 10 minutes and 14GB of RAM. This project needs 2 minutes and 1GB of RAM - with a similar end result (same length of migrated ways, less unnecessary splits).
This later turned out to be easy to extend and allowed us to easily modify the import to add a way splitting.
Two tables are used:
Ways table:
Column | Type
--------|---------------------------
id | bigint # Generated ID, as OSM ways are split at intersections
id_osm | bigint # Original (duplicated) osm_id
type | integer # Type based on highway tag
source | bigint # source node_id of intersection
target | bigint # target node_id
lon1 | double precision # start/target coords
lat1 | double precision
lon2 | double precision
lat2 | double precision
name | text # Way name - if any
length | double precision # Length in meters
geom | geometry(LineString,4326) # Geometry
Nodes table, contains only nodes on intersections of the ways.
Column | Type
--------|----------------------
id | bigint # OSM ID of a node
lon | double precision
lat | double precision
geom | geometry(Point,4326)
If using the way splitter the new, artificial, nodes are given IDs equal to the original node ID * 10000 + internal counter.
Address exporter exports address data from OSM PBF file using pyosmium library.
Example call:
./topo_import.py --address-import --pbf ~/poland-latest.osm.pbf --cache-mem
once exported, the IPython embed console will show and can store the data with:
extractor.save_to_csv("output.csv")
With --cache-mem will require around 12GB of free RAM to extract from Poland.
Without this option uses a node-cache in file that can take around 30GB of
space.
This program is licensed under GNU GPLv3 license.
The PBF file parsing is based on a GPLv3 parsepbf project, found originally at: http://pbf.raggedred.net/parsepbf.tar.bz2 Original readme is located in osmpbf/readme.txt.
Library was heavily rewritten to use a callback approach during reading to ease the RAM usage and refresh the code a bit.