It has never been easier to transform your RDF data into a property graph based on TinkerPop-Gremlin.

Apache TinkerPop is a graph computing framework for both graph databases (OLTP) and graph analytic systems (OLAP). Gremlin is the graph traversal language of TinkerPop.

The Resource Description Framework (RDF) is a standard model for data interchange on the Web originally designed as a metadata data model. It has come to be used as a general method for conceptual description or modeling of information that is implemented in web resources, using a variety of syntax notations and data serialization formats. This linking structure forms a directed labeled graph, where the edges represent the named link between two resources, represented by the graph nodes. This graph view is the easiest possible mental model for RDF and is often used in easy-to-understand visual explanations. Resources are denoted by IRIs. The general convention is to use http URIs.

RDF graphs are queries using SPARQL language. SPARQL 1.1 is a set of specifications that provide languages and protocols to query and manipulate RDF graph content on the Web or in an RDF store.

This library, rdf2gremlin, provides an easy way to load the RDF data-sets into a property graph in order to benefit from the features of traversal language, which are not available in SPARQL pattern matching language.

# switch to your local virtual environment 
. ./venv/activate

pip install rdf2gremlin
# ... enjoy ...

Note: The library tornado shall satisfy tornado>=4.4.1,<5.0 version restriction inherited from the python_gremlin library.

Gremlin-Python is designed to connect to a "server" that is hosting a TinkerPop-enabled graph system. That "server" could be Gremlin Server or a remote Gremlin provider that exposes protocols by which Gremlin-Python can connect. This requirement is inherited by the current library as well.

In order to use this library it is necessary to have a Gremlin service available locally or on a remote location. The easiest way is to run a TinkerProp server locally. This can be done by either: (a) downloading and running TinkerPop

wget https://archive.apache.org/dist/tinkerpop/3.4.3/apache-tinkerpop-gremlin-server-3.4.3-bin.zip
unzip apache-tinkerpop-gremlin-server-3.4.3-bin.zip
cd apache-tinkerpop-gremlin-server-3.4.3/
./bin/gremlin-server.sh start

# ... to stop the server ...

./bin/gremlin-server.sh stop

or (b) running it as a Docker container.

docker pull tinkerpop/gremlin-server
docker run --name gremlin-server -p 8182:8182 tinkerpop/gremlin-server

# ... to stop the server ...

docker stop gremlin-server

A typical connection to a server running on "localhost" that supports the Gremlin Server protocol using websockets from the Python shell looks like this:

from rdf2g import setup_graph

DEFAULT_LOCAL_CONNECTION_STRING = "ws://localhost:8182/gremlin"
g = setup_graph(DEFAULT_LOCAL_CONNECTION_STRING)

Once g has been created using a connection, it is then possible to start writing Gremlin traversals to query the remote graph.

Read an RDF graph.

import rdflib
import pathlib

OUTPUT_FILE_LAM_PROPERTIES = pathlib.Path("../resource/celex_project_properties_v2.ttl").resolve()

rdf_graph = rdflib.Graph()
rdf_graph.parse(str(OUTPUT_FILE_LAM_PROPERTIES), format="ttl")

Load the RDF graph into a property graph.

from rdf2g import load_rdf2g
load_rdf2g(g, rdf_graph)

The created property graph follows the following set of conventions.

• URIs and Blank nodes are transformed into property graph nodes.
• Predicates connecting an URI to another URI or a blank node are transformed into property graph edges. Edge labels correspond to qualified IRIs generated using the prefix definitions available in the RDF data-set.
• Node labels correspond to qualified IRIs generated using the prefix definitions available in the RDF data-set.
• RDF Litarals are transformed into values of the node properties, while the preceding predicates into keys of the node properties. In other words
• Predicates connecting an URI to a RDF Literal are transformed into {key:value} pairs and added as node properties.
• Nodes have a special property 'iri' that is equivalent to the absolute URI of the RDF resource.

Get a node referring to it either by label, iri or id

skos_concept_iri = rdflib.URIRef("http://www.w3.org/2004/02/skos/core#Concept")
v1 = rdf2g.get_node(g, skos_concept_iri)

skos_concept_label = "skos:Concept"
v2 = rdf2g.get_node(g, skos_concept_label)

hypothetical_node_id = 880
v3 = rdf2g.get_node(g, hypothetical_node_id)

print (v1 == v2 == v3) # should be true

Get nodes by their supposed rdf:type. This concept is inherited, of course, from RDF world.

skos_concept_label = "skos:Concept"

list_of_concept = rdf2g.get_nodes_of_type(g, skos_concept_label)

# print the list of concepts in the graph
print (list_of_concept)

It is possible to traverse the property graph and then generate the traversal tree from it. This is especially useful when the graph serves as structured document content say JSON or XML serialisation.

To do that first, get two levels deep traversal tree and the edges between them for all the nodes in the graph that have iri == known_iri. Further please see the Gremlin reference documentation at Apache TinkerPop for more information on usage.

known_iri = 'http://publications.europa.eu/resources/authority/celex/md_CODE' 
s = g.V().has('iri', known_iri).outE().inV().tree().next()

Altenatively use the function rdf2g.generate_traversal_tree

node = rdf2g.get_node(self.g, known_iri)
s = rdf2g.generate_traversal_tree(self.g, node)

Then expand and simplify that tree. First, simplify the dict structure to simple Python types, removing the Gremlin objects. Second, expand by providing the properties for each visited node, while the edges are considered as special properties leading to a another node dictionary.

from pprint import pprint
result = rdf2g.expand_tree(g, s)
pprint (result)

The traversal tree nodes contain, in addition to original RDF content, two special properties @id and @label which correspond to the standard Gremlin id and label properties. The @ sign is used to distinguish the original RDF from the Gremlin features. Property graph edges, are reduced to keys in the final dict and for this reason they have no additional descriptions just like in the original RDF graph.

You are more than welcome to help expand and mature this project.

When contributing to this repository, please first discuss the change you wish to make via issue, email, or any other method with the owners of this repository before making a change.

Please note we have a code of conduct, please follow it in all your interactions with the project.

This project is Licensed under the GPL v3 License - see LICENSE file