For RDF programming, RDFLib and Python may not execute the fastest, but we try hard to make it the fastest and most convenient way to write!
This is a collection of hints and pointers for hassle free RDF-coding.
Use label() to quickly look up the RDFS label of something, or better use preferredLabel() to find a label using several different properties (i.e. either rdfs:label, skos:preferredLabel, dc:title, etc.).
Use value() and set() to work with functional properties, i.e. properties than can only occur once for a resource.
Python allows slicing arrays with a slice object, a triple of start, stop index and step-size:
>>> range(10)[2:9:3]
[2, 5, 8]
RDFLib graphs override __getitem__ and we pervert the slice triple to be a RDF triple instead. This lets slice syntax be a shortcut for triples(), subject_predicates(), contains(), and other Graph query-methods:
graph[:]
# same as
iter(graph)
graph[bob]
# same as
graph.predicate_objects(bob)
graph[bob : FOAF.knows]
# same as
graph.objects(bob, FOAF.knows)
graph[bob : FOAF.knows : bill]
# same as
(bob, FOAF.knows, bill) in graph
graph[:FOAF.knows]
# same as
graph.subject_objects(FOAF.knows)
...
See examples.slice for a complete example.
Note
Slicing is convenient for run-once scripts of playing around in the Python REPL. However, since slicing returns tuples of varying length depending on which parts of the slice are bound, you should be careful using it in more complicated programs. If you pass in variables, and they are None or False, you may suddenly get a generator of different length tuples back than you expect.
SPARQL property paths are possible using overridden operators on URIRefs. See examples.foafpaths and rdflib.paths.
For simple output, or simple serialisation, you often want a nice readable representation of a term. All terms have a .n3(namespace_manager = None) method, which will return a suitable N3 format:
>>> from rdflib import Graph, URIRef, Literal, BNode
>>> from rdflib.namespace import FOAF, NamespaceManager
>>> person = URIRef('http://xmlns.com/foaf/0.1/Person')
>>> person.n3()
u'<http://xmlns.com/foaf/0.1/Person>'
>>> g = Graph()
>>> g.bind("foaf", FOAF)
>>> person.n3(g.namespace_manager)
u'foaf:Person'
>>> l = Literal(2)
>>> l.n3()
u'"2"^^<http://www.w3.org/2001/XMLSchema#integer>'
>>> l.n3(g.namespace_manager)
u'"2"^^xsd:integer'
You can parse data from a string with the data param:
graph.parse(data = '<urn:a> <urn:p> <urn:b>.', format='n3')
RDFLib includes a handful of commandline tools, see rdflib.tools.