Awesome

Oxrdflib

Oxrdflib provides an rdflib store based on pyoxigraph. This store is named "Oxigraph".

It also exposes pyoxigraph parsers and serializers as rdflib parser and serializer plugins.

Oxigraph store can be used as drop-in replacement of the rdflib default one. It support context but not formulas. Transaction support is not implemented yet.

SPARQL query and update evaluation is done by pyoxigraph instead of rdflib if the Oxigraph store is used. SPARQL update evaluation on Graph and ConjunctiveGraph is still done using rdflib because of a limitation in rdflib context management.

Oxrdflib is available on Pypi and installable with:

pip install oxrdflib

The oxrdflib store is automatically registered as an rdflib store plugin by setuptools.

Warning: Oxigraph is not stable yet and its storage format might change in the future. To migrate to future version you might have to dump and load the store content. However, Oxigraph should be in a good enough shape to power most of use cases if you are not afraid of down time and data loss.

API

Store

To create a rdflib graph using the Oxigraph store use

rdflib.Graph(store="Oxigraph")

instead of the usual

rdflib.Graph()

Similarly, to get a dataset, use

rdflib.Dataset(store="Oxigraph")

instead of the usual

rdflib.Dataset()

If you want to get the store data persisted on disk, use the open method on the Graph or Dataset object with the directory where data should be persisted. For example:

graph = rdflib.Graph(store="Oxigraph", identifier="http://example.com") # without identifier, some blank node will be used
graph.open("test_dir")

The store is closed with the close() method or automatically when Python garbage collector collects the store object.

If the open method is not called Oxigraph will automatically use a ramdisk on Linux and a temporary file in the other operating systems.

To do anything else, use the usual rdflib python API.

It is also possible to directly inject a pyoxigraph Store object directly into an Oxrdflib store:

graph = rdflib.Graph(store=oxrdflib.OxigraphStore(store=pyoxigraph.Store("test_dir")))

This might be handy to e.g. open the database as read-only:

graph = rdflib.Graph(store=oxrdflib.OxigraphStore(store=pyoxigraph.Store.read_only("test_dir")))

Parsers and serializers

To use Oxigraph parser, prefix the format identifiers with ox-. For example, to load data using the Oxigraph NTriples parser:

graph.parse(data, format="ox-nt")

and to serialize to Turtle:

graph.serialize(format="ox-ttl")

The following formats are supported:

• ox-ntriples (ox-nt)
• ox-nquads (ox-nq)
• ox-turtle (ox-ttl)
• ox-trig
• ox-xml

Note that Oxigraph parser and serializers are not 1:1 compatible with the rdflib ones and some minor differences exist.

An optimization has also been setup to skip Python entirely if the Oxigraph parsers and serializers are used with the Oxigraph store.

Differences with rdflib default store

• relative IRIs are not supported by Oxigraph.
• Just like the SPARQLStore, Oxigraph joins the initBindings parameter of the query method after the query has been evaluated, instead of injecting them at the beginning of the query.
• IRI prefixes set using the Graph bind method are not persisted on disk but kept in memory. They should be added again each time the store is opened.

Migration guide

From 0.2 to 0.3

• The 0.2 stores named "OxSled" and "OxMemory" have been merged into the "Oxigraph" store.
• The on-disk storage system provided by "OxSled" has been dropped and replaced by a new storage system based on RocksDB. To migrate you need to first dump your data in RDF using oxrdflib 0.2 and the serialize method, then upgrade to oxrdflib 0.3, and finally reload the data using the parse method.

Development

To run the test do first pip install -e . to register the stores in rdflib plugin registry. Then, cd tests && python -m unittest should run the tests.

The code is automatically formatted using black. A pre-commit configuration is provided. Run pip install pre-commit && pre-commit install to install pre-commit as a git pre-commit hook in your clone.