Neo4j Spatial is a library facilitating the import, storage and querying of spatial data in the Neo4j open source graph database.
This projects manual is deployed as part of the local build as the Neo4j Spatial Manual.
This library began as a collaborative vision between Neo-Technology and Craig Taverner in early 2010. The bulk of the initial work was done by Davide Savazzi as part of his 2010 Google Summer of Code (GSoC) project with Craig as mentor, as a project within the OSGeo GSoC program. In 2011 and 2012 two further GSoC projects contributed, the last of which saw Davide return as mentor.
The original vision for the library was a comprehensive suite of GIS capabilities somewhat inspired by PostGIS, while remaining aligned with the graph-nature of the underlying Neo4j database. To achieve this lofty goal the JTS and GeoTools libraries were used, giving a large suite of capabilities very early on.
However, back in 2010 Neo4j was an embedded database with deployments that saw a low level of concurrent operation. However, for many years now, Neo4j has been primarily deployed in high concurrent server environments. Over the years there have been various efforts to make the library more appropriate for use in these environments:
- REST API (early Neo4j 1.x servers had no Cypher)
- IndexProvider mechanism (used during Neo4j 1.x and 2.x server, and removed for Neo4j 3.0)
- 0.23: Addition of Cypher procedures for Neo4j 3.0
- 0.24: Addition of a high performance bulk importer to the in-graph RTree index
- 0.25: Addition of GeoHash indexes for point layers
- 0.26: Support for native Neo4j point types
- 0.27: Major port to Neo4j 4.x which deprecated many of the Neo4j API's the library depended on
- 0.29: Port to Neo4j 5.13
However, despite all these improvements, the core of the library only exposes the rich capabilities of JTS and GeoTools if used in an embedded environment. The large effort required to port the library to Neo4j 4.0 resulted in many backwards incompatibilities and highlighted the need for a new approach to spatial libraries for Neo4j.
This library was originally written in 2010 when Neo4j was still releasing early 1.x versions. This means it made use of internal Java API's that were deprecated over the years, some as early as Neo4j 2.x. When Neo4j 4.0 was released, many deprecated APIs were entirely removed. And the transaction API was changed in a fundamental way. This has meant that the spatial library needed a major refactoring to work with Neo4j 4.x:
- The library previously depended on nested transactions. This allowed code that was called within a transaction (like procedures) to be the same as code that was not (Java API). The removal of this support required that all internal API's needed to include parameters for the current transaction, and only the specific surface designed for embedded use not have that.
- The library made use of Lucene based explicit indexes in many places.
The removal of support for explicit indexes required completely new solutions in several places:
- The
OSMImporterwill instead now use normal Neo4j schema indexes (introduced in 2.0). However, these can only be created in separate index transactions. Due to the new transaction model this requires stopping the import transaction, starting an index transaction, and then restarting the import transaction. All of this is incompatible with procedures, which already have an incoming, non-stoppable transaction. The solution to this second problem was to run the actual import in another thread. This has the additional benefit of retaining the original batch-processing capabilities. The negative consequence of this it that it requires modifying the security model of the procedure context. - The
ExplicitIndexBackedPointIndexhas been modified to instead use a schema index. This required similar tricks to those employed in theOSMImporterdescribed above.
- The
- Neo4j 4.0 runs only in Java 11, and until recently GeoTools did not support newer Java versions. It was therefor necessary to upgrade the GeoTools libraries to version 24.2. This in turn required a re-write of the Neo4jDataStore interface since the older API had long been deprecated, and was entirely unavailable in newer versions.
- Neo4j 4.1 was slightly stricter in regard to passing nodes as parameters, requiring the nodes
objects to have been created in the current transaction.
To work around this we added
.byIdversions of thespatial.addNodeandspatial.removeNodeprocedures. We also changed thespatial.removeNodeprocedures to returnnodeIdinstead ofnode. - The change to Neo4j 4.2 was more subtle. Mostly only internal API's around the use of
Pathinstead ofFile. One change that could be noticed was theIndexManager.IndexAccessModeclass. In the 0.27.0 and 0.27.1 versions we usedOverridenAccessModeto take the users existing access mode and simply add on the rights to create tokens and indexes. In 0.27.2 we instead useRestrictedAccessModeto restrict the users access right to the built inAccessModel.Static.SCHEMAand then boost to enable index and token writes. The difference is subtle and should only be possible to notice in Enterprise Edition. - 0.28.0 tackles the ability to import multiple OSM files. The initial solution for Neo4j 4.x made use of schema indexes keyed by the label and property. However, that means that all OSM imports would share the same index. If they are completely disjointed data sets, this would not matter. But if you import overlapping OSM files or different versions of the same file, a mangled partial merger would result. 0.28.0 solves this by using different indexes, and keeping all imports completely separate. The more complex problems of importing newer versions, and stitching together overlapping areas, are not yet solved.
- Neo4j 4.3 has an issue with leaking RelationshipTraversalCursor, and we needed to do some workarounds to avoid this issue, usually by exhausting the iterator, which can have a higher performance cost in some cases. Version 0.28.1 includes this fix.
Consequences of the port to Neo4j 4.x:
- The large number of changes mean that the 0.27.x versions should be considered very alpha.
- Many API's have changed and client code might need to be adapted to take the changes into account.
- The new DataStore API is entirely untested in GeoServer, besides the existing unit and integration tests.
- The need to manage threads and create schema indexes results in the procedures requiring unrestricted access to internal API's of Neo4j.
This last point means that you need to set the following in your neo4j.conf file:
dbms.security.procedures.unrestricted=spatial.*
If you are concerned about the security implications of unrestricted access, my best advice is to review
the code and decide for yourself the level of risk you face. See, for example, the
method IndexAccessMode.withIndexCreate,
which adds index create capabilities to the security model.
This means that users without index creation privileges will be able to create the necessary spatial support indexes
described above.
This code was not written because we wanted to allow for that case, it was written because in the Neo4j security model,
procedures that can write data (mode=WRITE) are not allowed to create indexes.
So this security-fix was required even in the Community Edition of Neo4j.
The rest of the README might have information that is no longer accurate for the current version of this library. Please report any mistakes as issues, or consider raising a pull-request with an appropriate fix.
The key concepts of this library include:
- Allow the user to model geograph data in whatever way they wish, through providing an adapter (
extend
GeometryEncoder). Built-in encoders include:- WKT and WKB stored as properties of nodes
- Simple points as properties of nodes (two doubles, or a double[] or a native Neo4j
Point) - OpenStreetMap with complex geometries stored as sub-graphs to reflect the original topology of the OSM model
- Multiple CoordinationReferenceSystem support using GeoTools
- Support the concept of multiple geographic layers, each with its own CRS and Index
- Include an index capable of searching for complex geometries (in-graph RTree index)
- Support import and export in a number of known formats (e.g. Shapefile and OSM)
- Embed the library and Neo4j within GIS tools like uDig and GeoServer
Some key features include:
- Utilities for importing from ESRI Shapefile as well as Open Street Map files
- Support for all the common geometry types
- An RTree index for fast searches on geometries
- Support for topology operations during the search (contains, within, intersects, covers, disjoint, etc.)
- The possibility to enable spatial operations on any graph of data, regardless of the way the spatial data is stored, as long as an adapter is provided to map from the graph to the geometries.
- Ability to split a single layer or dataset into multiple sub-layers or views with pre-configured filters
- Server Plugin for Neo4j Server 2.x and 3.x
- REST API for creating layers and adding nodes or geometries to layers
- IndexProvider API (2.x only) for Cypher access using
START node=node:geom({query}) - Procedures (3.x only) for much more comprehensive access to spatial from Cypher
The current index is an RTree index, but it has been developed in an extensible way allowing for other indices to be added if necessary. The spatial queries implemented are:
- Contain
- Cover
- Covered By
- Cross
- Disjoint
- Intersect
- Intersect Window
- Overlap
- Touch
- Within
- Within Distance
The simplest way to build Neo4j Spatial is by using maven. Just clone the git repository and run
mvn installThis will download all dependencies, compiled the library, run the tests and install the artifact in your local
repository.
The spatial plugin will also be created in the target directory, and can be copied to your local server using
instructions on the spatial server plugin below.
The primary type that defines a collection of geometries is the Layer. A layer contains an index for querying. In addition, a Layer can be an EditableLayer if it is possible to add and modify geometries in the layer. The next most important interface is the GeometryEncoder.
The DefaultLayer is the standard layer, making use of the WKBGeometryEncoder for storing all geometry types as byte[] properties of one node per geometry instance.
The OSMLayer is a special layer supporting Open Street Map and storing the OSM model as a single fully connected graph. The set of Geometries provided by this layer includes Points, LineStrings and Polygons, and as such cannot be exported to Shapefile format, since that format only allows a single Geometry per layer. However, OMSLayer extends DynamicLayer, which allow it to provide any number of sub-layers, each with a specific geometry type and in addition based on an OSM tag filter. For example, you can have a layer providing all cycle paths as LineStrings, or a layer providing all lakes as Polygons. Underneath these are all still backed by the same fully connected graph, but exposed dynamically as apparently separate geometry layers.
Spatial data is divided in Layers and indexed by a RTree.
GraphDatabaseService database = new GraphDatabaseFactory().newEmbeddedDatabase(storeDir);
try{
ShapefileImporter importer = new ShapefileImporter(database);
importer.importFile("roads.shp","layer_roads");
} finally {
database.shutdown();
}If using the server, the same can be achieved with spatial procedures (3.x only):
CALL spatial.addWKTLayer('layer_roads', 'geometry')
CALL spatial.importShapefileToLayer('layer_roads', 'roads.shp')This is more complex because the current OSMImporter class runs in two phases, the first requiring a batch-inserter on the database. There is ongoing work to allow for a non-batch-inserter on the entire process, and possibly when you have read this that will already be available. Refer to the unit tests in classes TestDynamicLayers and TestOSMImport for the latest code for importing OSM data. At the time of writing the following worked:
OSMImporter importer = new OSMImporter("sweden");
Map<String, String> config = new HashMap<String, String>();
config.put("neostore.nodestore.db.mapped_memory", "90M" );
config.put("dump_configuration", "true");
config.put("use_memory_mapped_buffers", "true");
BatchInserter batchInserter = BatchInserters.inserter(new File(dir), config);
importer.importFile(batchInserter, "sweden.osm", false);
batchInserter.shutdown();
GraphDatabaseService db = new GraphDatabaseFactory().newEmbeddedDatabase(dir);
importer.reIndex(db, 10000);
db.shutdown();GraphDatabaseService database = new GraphDatabaseFactory().newEmbeddedDatabase(storeDir);
try {
SpatialDatabaseService spatialService = new SpatialDatabaseService(database);
Layer layer = spatialService.getLayer("layer_roads");
SpatialIndexReader spatialIndex = layer.getIndex();
Search searchQuery = new SearchIntersectWindow(new Envelope(xmin, xmax, ymin, ymax));
spatialIndex.executeSearch(searchQuery);
List<SpatialDatabaseRecord> results = searchQuery.getResults();
} finally {
database.shutdown();
}If using the server, the same can be achieved with spatial procedures (3.x only):
CALL spatial.bbox('layer_roads', {lon: 15.0, lat: 60.0}, {lon: 15.3, lat: 61.0})Or using a polygon:
WITH 'POLYGON((15.3 60.2, 15.3 60.4, 15.7 60.4, 15.7 60.2, 15.3 60.2))' AS polygon
CALL spatial.intersects('layer_roads', polygon) YIELD node
RETURN node.name AS nameFor further Java examples, refer to the test code in the LayersTest and the TestSpatial classes.
For further Procedures examples, refer to the code in the SpatialProceduresTest class.
IMPORTANT: Examples in this readme were originally tested with GeoServer 2.1.1. However, regular testing of new releases of Neo4j Spatial against GeoServer is not done, and so we welcome feedback on which versions are known to work, and which ones do not, and perhaps some hints as to the errors or problems encountered.
Each release of Neo4j Spatial builds against a specific version of GeoTools and should then be used in the version of GeoServer that corresponds to that. The list of releases below starting at Neo4j 2.0.8 were built with GeoTools 9.0 for GeoServer 2.3.2, but most release for Neo4j 3.x were ported to GeoTools 14.4 for GeoServer 2.8.4.
For the port to Neo4j 4.0 we needed to upgrade GeoTools to 24.x to avoid bugs with older GeoTools in Java 11. This also required a complete re-write of the Neo4jDataStore and related classes. This has not been tested at all in any GeoTools enabled application, but could perhaps work with GeoServer 2.18.
mvn clean install- unzip the
target/xxxx-server-plugin.zipand the Neo4j libraries from your Neo4j download under$NEO4J_HOME/libinto$GEOSERVER_HOME/webapps/geoserver/WEB-INF/lib - restart geoserver
- configure a new workspace
- configure a new datasource neo4j in your workspace. Point the "The directory path of the Neo4j database:" parameter to the relative (form the GeoServer working dir) or absolute path to a Neo4j Spatial database with layers ( see Neo4j Spatial)
- in Layers, do "Add new resource" and choose your Neo4j datastore to see the existing Neo4j Spatial layers and add them.
- check out the geoserver source
svn co https://svn.codehaus.org/geoserver/trunk geoserver-trunk- build the source
cd geoserver-trunk
mvn clean install- check that you can run the web app as of The GeoServer Maven build guide
cd src/web/app
mvn jetty:run- in
$GEOSERVER_SOURCE/src/web/app/pom.xmlhttps://svn.codehaus.org/geoserver/trunk/src/web/app/pom.xml, add the following lines under the profiles section:
<profile>
<id>neo4j</id>
<dependencies>
<dependency>
<groupId>org.neo4j</groupId>
<artifactId>neo4j-spatial</artifactId>
<version>5.20.0</version>
</dependency>
</dependencies>
</profile>The version specified on the version line can be changed to match the version you wish to work with (based on the version of Neo4j itself you are using). To see which versions are available see the list at Neo4j Spatial Releases.
- start the GeoServer webapp again with the added neo4j profile
cd $GEOSERVER_SRC/src/web/app
mvn jetty:run -Pneo4j- find Neo4j installed as a datasource under http://localhost:8080
For more info head over to Neo4j Wiki on uDig (This wiki is currently dead, but there appears to be a working mirror in Japan at http://oss.infoscience.co.jp/neo4j/wiki.neo4j.org/content/Neo4j_Spatial_in_uDig.html).
The Neo4j Spatial Plugin is available for inclusion in the server version of Neo4j 2.x, Neo4j 3.x and Neo4j 4.x.
- Using GeoTools 9.0 (for GeoServer 2.3.2):
- Using GeoTools 14.4 (for GeoServer 2.8.4):
- Using GeoTools 24.2 (for GeoServer 2.18.x):
For versions up to 0.15-neo4j-2.3.4:
#install the plugin
unzip neo4j-spatial-XXXX-server-plugin.zip -d $NEO4J_HOME/plugins
#start the server
$NEO4J_HOME/bin/neo4j start
#list REST API (edit to correct password)
curl -u neo4j:neo4j http://localhost:7474/db/data/For versions for neo4j 3.0 and later:
#install the plugin
cp neo4j-spatial-XXXX-server-plugin.jar $NEO4J_HOME/plugins/
#start the server
$NEO4J_HOME/bin/neo4j start
#list REST API (edit to correct password)
curl -u neo4j:neo4j http://localhost:7474/db/data/
#list spatial procedures (edit to correct password)
curl -u neo4j:neo4j -H "Content-Type: application/json" -X POST -d '{"query":"CALL spatial.procedures"}' http://localhost:7474/db/data/cypherThe server plugin provides access to the internal spatial capabilities using three APIs:
- A REST API for creating layers and adding nodes or geometries to layers.
- For usage information see Neo4j Spatial Manual REST
- Note that this API provides only limited access to Spatial, with no access the GeoPipes or import utilities
- This API was entirely removed when support for Neo4j 4.0 was added (version 0.27)
- An IndexProvider API (2.x only) for Cypher access using START node=node:geom({query})
- It is only possible to add nodes and query for nodes, and the resulting graph structure is not compatible with any other spatial API (not compatible with Java API, REST or Procedures), so if you use this approach, do not blend it with the other approaches.
- There is some brief documentation at Finding geometries within distance using cypher
- This API was removed for 3.0 releases, and so is only available for Neo4j 2.x
- Procedures for much more comprehensive access to spatial from Cypher
- Documentation is not yet available, but you can list the available procedures within Neo4j using the
query
CALL spatial.procedures - This API uses the Procedures capabilities released in Neo4j 3.0, and is therefor not available for Neo4j 2.x
- Documentation is not yet available, but you can list the available procedures within Neo4j using the
query
During the Neo4j 3.x releases, support for spatial procedures changed a little, through the addition of various new capabilities. They were very quickly much more capable than the older REST API, making them by far the best option for accessing Neo4j remotely or through Cypher.
The Java API (the original API for Neo4j Spatial) still remains, however, the most feature rich, and therefor we recommend that if you need to access Neo4j server remotely, and want deeper access to Spatial functions, consider writing your own Procedures. The Neo4j 3.0 documentation provides some good information on how to do this, and you can also refer to the Neo4j Spatial procedures source code for examples.
git clone https://github.com/neo4j-contrib/spatial/spatial.git
cd spatial
mvn clean packageAdd your GitHub credentials in your ~/.m2/settings.xml
<settings>
<servers>
<server>
<id>github</id>
<username>[email protected]</username>
<password>secret</password>
</server>
</servers>
</settings>To build and deploy:
git clone https://github.com/neo4j/spatial.git
cd spatial
mvn clean install site -Pneo-docs-build Add the following repositories and dependency to your project's pom.xml:
<dependency>
<groupId>org.neo4j</groupId>
<artifactId>neo4j-spatial</artifactId>
<version>5.20.0</version>
</dependency>The version specified on the last version line can be changed to match the version you wish to work with (based on the version of Neo4j itself you are using). To see which versions are available see the list at Neo4j Spatial Releases.
Some of the classes in Neo4j-Spatial include main() methods and can be run on the command-line. For example there are command-line options for importing SHP and OSM data. See the main methods in the OSMImporter and ShapefileImporter classes. Here we will describe how to set up the dependencies for running the command-line, using the OSMImporter and the sample OSM file two-street.osm. We will show two ways to run this on the command line, one with the java command itself, and the other using the 'exec:java' target in maven. In both cases we use maven to set up the dependencies.
git clone git://github.com/neo4j-contrib/spatial.git
cd spatial
mvn clean compilemvn dependency:copy-dependencies
java -cp target/classes:target/dependency/* org.neo4j.gis.spatial.osm.OSMImporter osm-db two-street.osm Note: On windows remember to separate the classpath with ';' instead of ':'.
The first command above only needs to be run once, to get a copy of all required JAR files into the directory target/dependency. Once this is done, all further java commands with the -cp specifying that directory will load all dependencies. It is likely that the specific command being run does not require all dependencies copied, since it will only be using parts of the Neo4j-Spatial library, but working out exactly which dependencies are required can take a little time, so the above approach is most certainly the easiest way to do this.
mvn exec:java -Dexec.mainClass=org.neo4j.gis.spatial.osm.OSMImporter -Dexec.args="osm-db two-street.osm"Note that the OSMImporter cannot re-import the same data multiple times, so you need to delete the database between runs if you are planning to do that.
React
Typescript
Django
Laravel
Facebook
Microsoft
Google
Alibaba
D3
Tencent