Clarify confusion around type of context element in ConjunctiveGraphs and context aware stores

gromgull, 2011-08-20T07:58:07.000Z

In the ConjunctiveGraph, the context field is assumed to be a graph with the identifier set to the URI of the context, i.e. this is what happens if you create context like this:

g=ConjunctiveGraph()
uri1=URIRef("http://example.org/mygraph1")
uri2=URIRef("http://example.org/mygraph2")

bob = URIRef(u'urn:bob')
likes = URIRef(u'urn:likes')
pizza = URIRef(u'urn:pizza')

g.get_context(uri1).add((bob, likes, pizza))
g.get_context(uri2).add((bob, likes, pizza))

Now g.contexts() returns a generator over some graphs.

Now, for code working on the store level, i.e. serializers and parsers, there should perhaps not be any graph objects?

I came across this when looking at the nquad parser, here:

https://github.com/RDFLib/rdflib/blob/master/rdflib/plugins/parsers/nquads.py#L106

This adds context as simply an URI ref.

I added a "work-around" to make the conjunctivegraph.contexts generator work "correctly" here:

https://github.com/RDFLib/rdflib/blob/master/rdflib/graph.py#L1075

Is this ok?

[gromgull]

I've semi-fixed this in the last two commits.

If you work with the ConjunctiveGraph class directly:

g = ConjunctiveGraph()
g.get_context('urn:1').add( ( s, p, o ) )

You end up with a Graph object in the store, with identifier set to the URIRef "urn:1".
I dont know saving the whole graph+store in the store is the best way to do it, saving just the identifier seems much cleaner.
AND - I can now mix and match stores in one ConjunctiveGraph - I am not even sure what this could mean:

>>> g1=ConjunctiveGraph()
>>> g2=ConjunctiveGraph()
>>> g1.get_context('urn:a').add( (s, p, o) )
>>> g2.addN( (x,y,z, g1.get_context('urn:a') ) )
>>> g2.store != g2.get_context('urn:a').store 
True

Hmm ...

In [43]: list(g1.contexts())[0]
Out[43]: <Graph identifier=urn:a (<class 'rdflib.graph.Graph'>)>

In [44]: list(list(g1.contexts())[0])
Out[44]: 
[(rdflib.term.BNode('Nae2ff5ac0056427e852347e0a58ff925'),
  rdflib.term.BNode('N7c8cbd571a51409e8a92c2870a30eddd'),
  rdflib.term.BNode('Nbc486ecd9a5b46d5913dafdc4458fc3f'))]

In [45]: list(g1.get_context('urn:a'))
Out[45]: 
[(rdflib.term.URIRef(u'a'),
  rdflib.term.URIRef(u'b'),
  rdflib.term.URIRef(u'c'))]

This is no good :)

[gromgull]

Something else, whatever we decide - the store.add method should check the type of the context parameter to make sure it is either Graph/Identifier to make sure we avoid very subtle and odd bugs.