Incorrect JSON-LD serialization of lists starting with int/double zero literal

[svenschneider]

If a list that starts with a zero literal ("0"^^xs:integer or "0"^^xs:double) is serialized to JSON-LD the result seems to be incorrect. The following example demonstrates this for two lists: [1, 0] and [0, 1].

import rdflib
from rdflib.namespace import RDF

EX = rdflib.Namespace("http://example.org/")

zero = rdflib.Literal(0.0)
one = rdflib.Literal(1.0)

g = rdflib.Graph()

bn1 = rdflib.BNode()
c1 = rdflib.collection.Collection(g, bn1, [one, zero])
g.add((EX.s, EX.p1, bn1))

bn2 = rdflib.BNode()
c2 = rdflib.collection.Collection(g, bn2, [zero, one])
g.add((EX.s, EX.p2, bn2))

print(g.serialize(format="json-ld", indent=4))

Running this example on the master branch actually produces the following output:

[
    {
        "@id": "http://example.org/s",
        "http://example.org/p1": [
            {
                "@list": [
                    {
                        "@value": 1.0
                    },
                    {
                        "@value": 0.0
                    }
                ]
            }
        ],
        "http://example.org/p2": [
            {
                "@id": "_:N0952ac52472941a59065bc0d92837b0e"
            }
        ]
    },
    {
        "@id": "_:N0952ac52472941a59065bc0d92837b0e",
        "http://www.w3.org/1999/02/22-rdf-syntax-ns#first": [
            {
                "@value": 0.0
            }
        ],
        "http://www.w3.org/1999/02/22-rdf-syntax-ns#rest": [
            {
                "@list": [
                    {
                        "@value": 1.0
                    }
                ]
            }
        ]
    }
]

The result I would expect here however is:

[
    {
        "@id": "http://example.org/s",
        "http://example.org/p1": [
            {
                "@list": [
                    {
                        "@value": 1.0
                    },
                    {
                        "@value": 0.0
                    }
                ]
            }
        ],
        "http://example.org/p2": [
            {
                "@list": [
                    {
                        "@value": 0.0
                    },
                    {
                        "@value": 1.0
                    }
                ]
            }
        ]
    }
]

The problem seems to be caused by the second condition on this line. Changing

if l_ != RDF.nil and not graph.value(l_, RDF.first):

to

if l_ != RDF.nil and RDF.first not in graph.predicates(l_):

solves this issue for me. But I'm not sure if that breaks anything else. If desired I can create a pull request with that change.

Please let me know if I can help with any further information or input.

Thanks and regards
Sven

[svenschneider]

To add to my previous point there seems to be one more problematic case on this line. Similarly to the previous suggestion that line supposedly should be changed from

containers = [LIST, None] if graph.value(o, RDF.first) else [None]

to

containers = [LIST, None] if RDF.first in graph.predicates(o) else [None]

[devonsparks]

I just stumbled on this same problem in a different context using rdflib==7.0.0 and get the same results @svenschneider does using the original example. What additional regression testing is recommended before the proposed change could be merged?