nquads parser creates one context per triple in the default graph

[pchampin]

In nquads, the 4th component (graph id) of each line can be omitted, meaning that the triple belongs to the default graph.

Currently, the nquads parser puts every such triple in a different context, identified by a fresh BNode, which is clearly wrong.

E.g.:

from rdflib import ConjunctiveGraph, Dataset
from rdflib.parser import StringInputSource

ds = ConjunctiveGraph()

src = StringInputSource("""
<http://example.org/s1> <http://example.org/p1> <http://example.org/o1> .
<http://example.org/s2> <http://example.org/p2> <http://example.org/o2> .
<http://example.org/s3> <http://example.org/p3> <http://example.org/o3> <http://example.org/g3> .
""")

ds.parse(src, format="nquads")

assert len(ds) == 3, len(g)
assert len(list(ds.contexts())) == 2, len(list(ds.contexts()))

[niklasl]

Closely related to #436. We should coordinate solutions.

[pchampin]

You are right, those issues related.
The reason I opened a new one is that, regardless of the chosen solution for handling the default graph,
it is clear that those triples should not end up in different contexts.

So those are two different problems, fixing one requires fixing the other. Indeed, we should coordinate.

[niklasl]

Yes, I agree, this issue for the nquads parser is distinct.

(I'll also look into coordinating a solution for RDFLib/rdflib-jsonld#34 with this. I wrote down #436 when I originally hit that. But attempting to resolve all parsing and serializing of datasets at once might have been too much.)

[joernhees]

hmm, one question wrt. the version in which this will be fixed, e.g. by merging #536 :

as we use http://semver.org, is this a bug fix only (so 4.2.2) or an actual backwards incompatible change (5.x.x)?

[pchampin]

The current behaviour of the nquads parser is wrong, to a point that no one
could reasonably rely on it.
So I would go for a bug fix.

On Thu, Oct 29, 2015 at 3:43 PM, Jörn Hees notifications@github.com wrote:

hmm, one question wrt. the version in which this will be fixed, e.g. by
merging #536 #536 :

as we use http://semver.org, is this a bug fix only (so 4.2.2) or an
actual backwards incompatible change (5.x.x)?

—
Reply to this email directly or view it on GitHub
#535 (comment).

[joernhees]

fixed in #536