inconsistant Dataset internal data

[ticapix]

Hello,

I'm manipulating Dataset and I ended up in a weird situation where I have an Dataset instance, with triples in it, in a graph, but:

• I can't get get the graph. It's always empty
• If I serialise the dataset, it's empty.
• If I loop over the dataset quads, the data is there.

I couldn't reproduce the dataset with a simple code (it comes from a large code base), but I was able to dump it with pickle and reproduce the issue in the small script below.

import pickle

from rdflib import Dataset, URIRef

with open("/tmp/ds2.png", "rb") as fd:
    ds2: Dataset = pickle.load(fd)

graph = URIRef("http://example.com/cde/#graph")


for i, (s, p, o, g) in enumerate(ds2):
    assert g == graph  # ok
    print(i, s, p, o, g)

assert len(ds2) == 10, len(ds2)  # ok
assert len(ds2.graph(graph)) == 10, len(ds2.graph(graph))  # fails
assert ds2.serialize(format="trig") != "\n"  # fails

output

0 http://example.com/cde/#guid http://example.com/cde/#ontology_version 0.0.0 http://example.com/cde/#graph
1 http://example.com/cde/#guid http://example.com/cde/#created_on 2025-11-26T12:16:21.555303+00:00 http://example.com/cde/#graph
2 http://example.com/cde/#guid http://www.w3.org/1999/02/22-rdf-syntax-ns#type http://example.com/cde/#Claim http://example.com/cde/#graph
3 http://example.com/cde/#guid http://example.com/cde/#b hello http://example.com/cde/#graph
4 http://example.com/cde/#guid http://example.com/cde/#visibility_scope confidential http://example.com/cde/#graph
5 http://example.com/cde/#guid http://www.w3.org/1999/02/22-rdf-syntax-ns#type http://example.com/cde/#Person http://example.com/cde/#graph
6 http://example.com/cde/#guid http://example.com/cde/#a 11 http://example.com/cde/#graph
7 http://example.com/cde/#guid http://example.com/cde/#last_update_on 2025-11-26T12:16:21.555309+00:00 http://example.com/cde/#graph
8 http://example.com/cde/#guid http://example.com/cde/#graph http://example.com/cde/#graph http://example.com/cde/#graph
9 http://example.com/cde/#guid http://example.com/cde/#created_by http://example.com/cde/#pytest http://example.com/cde/#graph
Traceback (most recent call last):
  File "/home/pierre/workspace/.../back-api/./test02.py", line 39, in <module>
    assert len(ds2.graph(graph)) == 10, len(ds2.graph(graph))  # fails
           ^^^^^^^^^^^^^^^^^^^^^^^^^^^
AssertionError: 0

Would anyone have an idea ?

I'm using rdflib-7.4.0

[edmondchuc]

I believe I've seen this issue before - if it's the same as what I think it is, it should be fixed in #3259 and released in 7.3.0.

You mentioned you're running RDFLib 7.4.0, but is it possible that the pickled Dataset was originally created using an earlier version? If so, this comment illustrates how the issue can be reproduced: #698 (comment).

If you can confirm that your Dataset was created with code older than 7.4.0, then you have two options:

1. Upgrade the code that created the pickled Dataset to 7.3.0 or later, or
2. Add graphs to the Dataset correctly, following the example described in Dataset.graph should not allow adding random graphs to the store #698 (comment)

[ticapix]

Thanks a lot for the pointers!

It was created with 7.4.0.
Now that I get the good and bad way to do it, I'll double check our code base to see how we do it.