At the moment there are a few key issues with the current rdflib.Dataset, particularly in its handling of parsing Default Graph triples. I am very interested in working on a Pull Request to solve these issues since it will improve Named Graph support for all projects making use of rdflib.

Key issues with rdflib.Dataset:

• Parsed default graph triples are put under a randomized Blank Node named graph.
• Serializing a rdflib.Dataset to application/n-quads or application/trig will always place default graph triples under a Blank Node named graph.
• There is no way of accessing only the default graph in a rdflib.Dataset.
• The default format for Dataset().parse(data=...) is text/turtle when it really should be application/trig.

By adding a new term rdflib.term.DefaultGraph, we can handle the serialization of these triples independent from Named Graph triples. Additionally, we should be able to access the default graph by specifying the following:

from rdflib import Dataset from rdflib.term import DefaultGraph ds = Dataset() trig_string = ( '<dg:s> <dg:p> <dg:o>. ' 'GRAPH <ng:g> { <ng:s> <ng:p> <ng:o>. }' ) ds.parse(data=trig_string, format="trig") default_graph = ds.graph(DefaultGraph) for triple in default_graph: print(triple)

(rdflib.term.URIRef('dg:s'), rdflib.term.URIRef('dg:p'), rdflib.term.URIRef('dg:o'))