Dataset: NQuads serialization only makes sense for context-aware stores!

[sdasda7777]

Hi, I'd like to serialize JSON-LD to N-Quads, however I keep getting NQuads serialization only makes sense for context-aware stores! exception. From the Dataset definition and this issue it seems that it however should work, correct? Here is my code (formatting might be bit off):

from rdflib import Dataset
        
    try:
        data = """{
  "@context": "http://schema.org/",
  "@type": "Person",
  "name": "Jane Doe",
  "jobTitle": "Professor",
  "telephone": "(425) 123-4567",
  "url": "http://www.janedoe.com"
}"""
         ds = Dataset()
         ds = ds.parse(data=data, format="json-ld")
         print(ds.serialize(format="nquads"))
    except Exception as e:
         print(e)

[aucampia]

@sdasda7777 this is a known bug (IMO at least) recorded here #1892.

I will close this as a duplicate, however, I think that NQuads for a non-context-aware store should be identical to NTriples [ref].

[sdasda7777]

@sdasda7777 this is a known bug (IMO at least) recorded here #1892.

I see, I was hoping for some kind of solution, but since it's open for that long I guess I can forget about that.

I think that NQuads for a non-context-aware store should be identical to NTriples.

For this example, yes. However JSON-LD may actually have context information, and I want it to work in that case as well.

[sdasda7777]

Section number 8 of the standard has an example that should translate into multiple graphs. You can see it really does when you try converting it to N-Quads on the JSON-LD playground, which is exactly what I want to be able to do.

[aucampia]

For this example, yes. However JSON-LD may actually have context information, and I want it to work in that case as well.

But then you should use a context-aware store like rdflib.graph.Dataset, otherwise it will be lost anyway.

I see, I was hoping for some kind of solution, but since it's open for that long I guess I can forget about that.

The reason this bug is not fixed, even though it is simple to fix, is that it can easily be worked around by users, even if we fix it, it will be no different from just using NTriples, and this is because non-context aware stores cannot contain named graphs, and if there are no graph names NQuads is equivalent to NTriples.

We will be grateful for PRs that fix #1892.

[sdasda7777]

But then you should use a context-aware store like rdflib.graph.Dataset, otherwise it will be lost anyway.

But I am using a Dataset, am I not?

The reason this bug is not fixed, even though it is simple to fix, is that it can easily be worked around by users, even if we fix it, it will be no different from just using NTriples, and this is because non-context aware stores cannot contain named graphs, and if there are no graph names NQuads is equivalent to NTriples.

Okay, so I should be able to do this conversion just fine if I use context-aware store? Which stores are those?

[aucampia]

But I am using a Dataset, am I not?

I overlooked that, you are indeed using Dataset. In that case, this is a somewhat separate bug from #1892.

Okay, so I should be able to do this conversion just fine if I use context-aware store? Which stores are those?

Dataset should be context aware, but I guess for some reason the serializers disagree. rdflib.graph.ConjunctiveGraph is also context aware, and I suspect that should work, could you try using that instead of rdflib.graph.Dataset ?

[sdasda7777]

Dataset should be context aware, but I guess for some reason the serializers disagree. rdflib.graph.ConjunctiveGraph is also context aware, and I suspect that should work, could you try using that instead of rdflib.graph.Dataset ?

Yeah, it is the same output, NQuads serialization only makes sense for context-aware stores! as well.

I am using Python 3.10.4 and rdflib 6.3.2, if that helps in any way.

[aucampia]

Okay, then I don't really have an immediate answer. I will look at it when I have time. It does seem like a bug, and a pull requests fixing this will be welcome.

[aucampia]

Oh wait, I think the problem is this:

ds = ds.parse(data=data, format="json-ld")

Don't reassign ds, I think what is being returned there is not a Dataset. This in itself is probably a bit messed up, but I think if you just don't reassign ds, and use the created ds, it should work.

parse mutates the object, so it is also a bit redundant in this case.

[sdasda7777]

Oh, you're right! Thank you so much!!!

[namedgraph]

@aucampia how about making nquads the default serialization for Dataset? Currently ds.serialize() returns nothing because I'm assuming it defaults to ntriples? Should I create a separate issue for this?

[aucampia]

@namedgraph there is actually a PR that will fix this - though it will use trig instead of nquads:

• fix: ConjunctiveGraph.serialize to format as TriG by default (#1674) #2373

But this needs some other things fixed before we can merge it.

[aucampia]

The reason the output is empty is because of this bug:

• Formats that don't support named graphs serialize Datasets and ConjunctiveGraphs with non-default graphs without raising any errors #2393