Formats that don't support named graphs serialize Datasets and ConjunctiveGraphs with non-default graphs without raising any errors #2393
Description
If RDFLib is used to serialize a Dataset or ConjunctiveGraph that contains non-default graphs [ref] as a format that does not support named graphs (i.e. N-Triples or Turtle) no error is raised, and the output is wrong.
Given this data in test/data/variants/diverse_quads.nq (equivalent to this trig document):
rdflib/test/data/variants/diverse_quads.nq
Lines 1 to 10 in ddcc4eb
Using rdfpipe to convert it to turtle gives:
$ pipx run --spec git+https://github.com/RDFLib/rdflib.git@main rdfpipe -o turtle test/data/variants/diverse_quads.nq; echo status=$? ⚠️ rdfpipe is already on your PATH and installed at /home/iwana/.local/bin/rdfpipe. Downloading and running anyway. @prefix ns1: <http://example.com/> . @prefix ns2: <urn:example:> . @prefix ns3: <example:> . @prefix xsd: <http://www.w3.org/2001/XMLSchema#> . ns1:subject ns1:predicate "typeless", "日本語の表記体系"@jpx . ns2:subject ns3:predicate ns3:subject ; ns2:predicate ns2:object . ns3:subject ns3:predicate ns3:object, 12 ; ns1:predicate ns1:object, "XSD string" . status=0And using rdfpipe to convert it to ntriples gives:
$ pipx run --spec git+https://github.com/RDFLib/rdflib.git@main rdfpipe -o ntriples test/data/variants/diverse_quads.nq; echo status=$? ⚠️ rdfpipe is already on your PATH and installed at /home/iwana/.local/bin/rdfpipe. Downloading and running anyway. /home/iwana/.local/pipx/.cache/bf79805433c3d94/lib64/python3.11/site-packages/rdflib/plugins/serializers/nt.py:40: UserWarning: NTSerializer always uses UTF-8 encoding. Given encoding was: None warnings.warn( <http://example.com/subject> <http://example.com/predicate> "typeless" . <example:subject> <http://example.com/predicate> <http://example.com/object> . <urn:example:subject> <example:predicate> <example:subject> . <example:subject> <example:predicate> "12"^^<http://www.w3.org/2001/XMLSchema#integer> . <example:subject> <example:predicate> <example:object> . <http://example.com/subject> <http://example.com/predicate> "日本語の表記体系"@jpx . <urn:example:subject> <urn:example:predicate> <urn:example:object> . <example:subject> <http://example.com/predicate> "XSD string" . status=0In both cases, the output is wrong, not just incomplete.
I would say the right behaviour is that an error is raised when a ConjunctiveGraph or Dataset with non-default graphs are serialized using a format that does not support named graphs.
I'm making this issue to get some feedback, but I will make a PR to fix it shortly.
There is a similar issue with riot from Jena [ref], but riots behaviour is not quite as bad