Graph parse method overrides prefix bindings

[edmondchuc]

If I run the following code:

from rdflib import Graph, Namespace

data = """
PREFIX dcterms: <http://purl.org/dc/terms/>
PREFIX ns: <https://example.com/>
PREFIX rdfs: <http://www.w3.org/2000/01/rdf-schema#>
PREFIX schema: <https://schema.org/>
PREFIX skos: <http://www.w3.org/2004/02/skos/core#>

ns:08429fce-4d70-4be4-9c64-ffc80f554ea7
    a skos:Concept .
"""

EX = Namespace("https://example.com/")

graph = Graph()

graph.bind("ex", EX)
graph.parse(data=data, format="turtle")

graph.print(format="turtle")

It will print:

@prefix ns: <https://example.com/> .
@prefix skos: <http://www.w3.org/2004/02/skos/core#> .

ns:08429fce-4d70-4be4-9c64-ffc80f554ea7 a skos:Concept ;
    skos:definition "definition" ;
    skos:prefLabel "label" .

I would have expected the bind to persist through the life of the graph object and print the following result:

@prefix ex: <https://example.com/> .
@prefix skos: <http://www.w3.org/2004/02/skos/core#> .

ex:08429fce-4d70-4be4-9c64-ffc80f554ea7 a skos:Concept ;
    skos:definition "definition" ;
    skos:prefLabel "label" .

If I swap the two lines from:

graph.bind("ex", EX)
graph.parse(data=data, format="turtle")

to:

graph.parse(data=data, format="turtle")
graph.bind("ex", EX)

it then prints what I expect.

Is this the expected behaviour where calling the parse() method overwrites prefix bindings in the graph's namespace manager?

[ghost]

Is this the expected behaviour where calling the parse() method overwrites prefix bindings in the graph's namespace manager?

It is according to my discoveries when working through testing the override/replace interactions and I included an observation on the matter:

rdflib/test/test_graph/test_namespace_rebinding.py

Line 113 in 05dced2

# Yes, namespace bindings in parsed data replace existing

which is intended to find its way into the documentation.

I have a sense that the domain modelling (of the source as a serialized RDF Graph) is slightly more faithfully represented in the traditional RDFLib invocation idiom:

g = Graph().parse(data=source, format=format)

There doesn't seem to be any elegant and straightforward way of handling prefix-namespace bindings in a format-independent manner. The turtle parser doesn't use either override or replace:

rdflib/rdflib/plugins/parsers/notation3.py

Line 1946 in 05dced2

graph.bind(prefix, namespace)

Also consider:

def test_parse_namespace():
    data = """
    PREFIX dcterms: <http://purl.org/dc/terms/>
    PREFIX ns: <https://example.com/>
    PREFIX rdfs: <http://www.w3.org/2000/01/rdf-schema#>
    PREFIX schema: <https://schema.org/>
    PREFIX skos: <http://www.w3.org/2004/02/skos/core#>

    ns:08429fce-4d70-4be4-9c64-ffc80f554ea7
        a skos:Concept .
    """

    EX = Namespace("https://example.com/")

    graph = Graph()

    graph.parse(data=data, format="turtle")
    graph.bind("ex", EX)

    assert graph.serialize(format="turtle") == (
        "@prefix ex: <https://example.com/> .\n"
        "@prefix skos: <http://www.w3.org/2004/02/skos/core#> .\n"
        "\n"
        "ex:08429fce-4d70-4be4-9c64-ffc80f554ea7 a skos:Concept .\n"
        "\n"
    )

    graph2 = Graph()
    graph2 += graph  # Namespace bindings in graph not preserved

    assert graph2.serialize(format="turtle") == (
        "\n"
        "<https://example.com/08429fce-4d70-4be4-9c64-ffc80f554ea7> a "
        "<http://www.w3.org/2004/02/skos/core#Concept> .\n"
        "\n"
    )

    graph2 = Graph()
    graph2.bind("xe", EX)

    graph2 += graph  # Namespace bindings in graph2 preserved

    assert graph2.serialize(format="turtle") == (
        "@prefix xe: <https://example.com/> .\n"
        "\n"
        "xe:08429fce-4d70-4be4-9c64-ffc80f554ea7 a "
        "<http://www.w3.org/2004/02/skos/core#Concept> .\n"
        "\n"
    )

[aucampia]

I think your expectaion is reasonable @edmondchuc - but changing this will be a breaking change, so should be targeted for 7.x, see #2108 for some options.