Imports OpenStreetMap data into a Neo4j graph database. Produces a simple schema that includes nodes, ways, and all properties/tags contained in the source OSM data. Geometries are stored as Points in Neo4j to support geospatial queries.
Designed to read OSM data in XML format. Uses SAX event-driven XML parsing to accomidate large XML files.
Uses the scenic-routing/javasdk to help with Neo4j Graph database interactions.
OSM Nodes are imported as graph nodes with the label of INTERSECTION. Their point geometry is stored as a Point in the geom property, and in WKT format in the geom_wkt property.
OSM Ways are imported as relationships between the nodes (or intersections). Ways are labeled as CONNECTS, with their LineString geometry being stored as an array of Points in the geom property, and a WKT string stored in the way property.
Only ways with a highway tag are imported (must match Xpath: /osm/way[tag/@k = 'highway']). More information about possible highway values is avilable here.
All other properties and tags in the source OSM data are flattened and attached as properties on the graph nodes/relationships that are created during import.
Built with the Java 11 JDK and Neo4j Server Community (version: 4.4.1). Code should be fairly portable, as no advanced features of Java or Neo4j are used.
This is a maven project and can be built using:
mvn clean install
Make sure the Neo4j library version in the pom.xml file matches the Neo4j Server version. Available versions in Maven Central can be found here.
The importer is executed via java CLI and accepts 3 paramters in this order:
- osmFile - required - path to the source OSM XML file
- graphDb - required - filesystem path to the Neo4j GraphDB (web/Bolt API not supported)
- action - optional - allows execution of specific actions listed below (only needed in advanced scenarios)
To perform the default import:
java -jar target/osm2graph-neo4j-0.1.0-SNAPSHOT.jar /development/workspace/SanFrancisco.osm /development/workspace/neo4j/graph.db
To perform a specific action:
java -jar target/osm2graph-neo4j-0.1.0-SNAPSHOT.jar /development/workspace/SanFrancisco.osm /development/workspace/neo4j/graph.db loadnodes
Available actions (all executed as part of defeault action):
- loadnodes - Loads OSM nodes into GraphDB
- loadways - Loads OSM ways into GraphDB
- createindexes - Creates an GraphDB index of Nodes and Relationships on the
osm_idproperty (Nodes only) andgeomproperty; used to speed up node lookup during way import, and support geospatial queries - resetgraphdb - Clears the GraphDB of all nodes, relationships, and indexes.
Get street by name:
MATCH (a)-[r{name: 'Marlene-Dietrich-Straße'}]-(b) RETURN a, r, b
Get street by type:
MATCH (a)-[r]-(b) WHERE r.highway = 'secondary' OR r.highway = 'residential' RETURN a, r, b
Get intersection by bounding box:
WITH
point({x: -89.36, y: 43.07, crs: "WGS-84"}) AS lowerLeft,
point({x: -89.06, y: 43.57, crs: "WGS-84"}) AS upperRight
MATCH (n)
WHERE point.withinBBox(n.geom, lowerLeft, upperRight)
RETURN (n)
Get street by bounding box:
WITH
point({x: -89.36, y: 43.07, crs: "WGS-84"}) AS lowerLeft,
point({x: -89.06, y: 43.57, crs: "WGS-84"}) AS upperRight
MATCH (a)-[r]-(b)
WHERE
point.withinBBox(r.geom[0], lowerLeft, upperRight) OR
point.withinBBox(r.geom[1], lowerLeft, upperRight)
RETURN a, r, b
An index is created on the osm_id property of Nodes (used during import). A Point Index is created on the geom property of both Nodes and Relationships to support geospatial queries.
To view the available indexes, run the following Cypher query:
CALL db.indexes();
City exports of OSM data are made avaiable via BBBike.
Exports by country/region are available from Geofabrik.
JUnit tests can be executed with the following command:
mvn test
Here is a blog entry I wrote about creating this loader, which goes into further detail on a few of the design decisions and methodologies used.