Inconsistent Labeling for Blank Nodes Across Writers

[aliariff]

Version

4.9.0

What happened?

There's an observed inconsistency in how blank nodes are labelled across different writer classes. The RowSetWriterCSV class labels blank nodes using the NodeToLabelMap. However, the RowSetWriterJSON and RowSetWriterXML writers decide on the labelling strategy based on the provided context:

boolean outputGraphBNodeLabels = (context != null) && context.isTrue(ARQ.outputGraphBNodeLabels);
labels = outputGraphBNodeLabels
    ? SyntaxLabels.createNodeToLabelAsGiven()
    : SyntaxLabels.createNodeToLabel();

For consistency across output formats, it's suggested to handle blank node labelling the same way across all writers. Specifically, the CSV writer should also decide the labelling strategy based on context, as in the JSON and XML writers.

Relevant output and stacktrace

No response

Are you interested in making a pull request?

Yes

[afs]

Hi @aliariff - could you show a concrete example and say why it matters to you?

Especially for the CSV format, which is lossy anyway, it uses the _:bN form for brevity of display.

TSV preserves more of the RDF-ness like strings vs URIs.

Blank nodes in the results format are scoped to the result document - they don't represent the original blank node in the data. (That's a feature of blank nodes.)

[aliariff]

Hi @afs,

Our platform allows users to export data in various formats, including JSON, XML, and CSV. When using ARQ.outputGraphBNodeLabels = true, the blank node labels in JSON and XML outputs appear without the _:bN form. However, in the CSV format, they're represented in the _:bN form.

Imagine a user referencing a blank node labelled node1 in a JSON file. If they decide to cross-reference or re-analyze the data in CSV, they would expect to see node1 but instead, they'd encounter something like _:b0, which could lead to confusion.

Though the _:bN form in CSV provides brevity, the inconsistency across formats can be misleading. Consistent representations across all export formats reduce potential misunderstandings for our users.

[afs]

That is relying on a Jena-specific feature.

By the standards, blank nodes in the results format are scoped to the result document, not across documents. See, for example, the prettified Turtle output.

Bug label removed - this is a feature request.

If you want to put in a PR for an opt-in (using ARQ.outputGraphBNodeLabels) form of labels for CSV, then please do so.