Enable rdf validate compare graph names with blank nodes

[adanilenka]

So currently, rdf validate command with --compare-to-rdf-file cannot properly validate quads where the name of the graph is a blank node in certain cases.

Example input:

<http://example.org/resource/A> <http://example.org/property/p> <http://example.org/resource/B> .
<http://example.org/resource/A> <http://example.org/property/p> <http://example.org/resource/C> <http://example.org/graph/G>.
<http://example.org/resource/A> <http://example.org/property/p> <http://example.org/resource/D> _:bn1.

Encoding and decoding these n-quads with:

jelly-cli rdf to-jelly ./in_001.nq --to ./out.jelly --opt.generalized-statements=false --opt.rdf-star=false --opt.max-name-table-size=8 --opt.max-prefix-table-size=0 --opt.max-datatype-table-size=4

jelly-cli rdf from-jelly ./out.jelly --to ./deserialized_001.nq

Gives the output of:

<http://example.org/resource/A> <http://example.org/property/p> <http://example.org/resource/B> .
<http://example.org/resource/A> <http://example.org/property/p> <http://example.org/resource/C> <http://example.org/graph/G> .
<http://example.org/resource/A> <http://example.org/property/p> <http://example.org/resource/D> _:B66a3e1ec00fdd8c0072f62662a2f5b2a .

So the only difference really is the ID of the blank node. For triples, rdf validate handles BNs, but running
jelly-cli rdf validate ./out.jelly --compare-to-rdf-file ./in_001.nq

Gives the following error:

What is interesting is that adding --compare-ordered=true removes the error, and validation goes okay.

Also, the ID of the dataset in the error changes each run.

[Ostrzyciel]

This is a limitation of Jena, as it only has a graph isomorphism algorithm, and not a dataset one, at least as far as I know. This means that the blank node resolution algorithm can't take graph names into account.

The ordered algorithm works on statement level, so it's immune to that limitation. If possible, you should use that instead.

I will look into this later, but no promises. RDF isomorphism in general is a big pain.