Blank nodes in `CONSTRUCT` output

[namedgraph]

Blank nodes in `CONSTRUCT` output

[afs]

Blank nodes in the template are unique across the solution mappings used while instantiating the template.

The query CONSTRUCT WHERE { ?S ?P ?O } has a template of { ?S ?P ?O } which has no blank nodes.

Blank nodes from the WHERE clause are not affected - they are just data.

[rubensworks]

Ok, then our implementation in Comunica is incorrect.
(and I'm actually happy about this, as it will allow us to significantly simplify things)

Thanks for the input @afs!

[afs]

And anyone can make a PR to w3c/sparql-query for improvements of the spec text 😄

[github-actions]

Thanks for reporting!

[jeswr]

This is a dupe of #1160

[rubensworks]

The query is equivalent to the following:

CONSTRUCT
{
    ?s ?p ?o
}
WHERE
{
    ?s ?p ?o
}

The WHERE clause will be executed first, and will produce one query solution for each triple inside the dataset.
For each query solution, the query template will be instantiated.
According to what the SPARQL spec says, blank nodes will be scoped per query solution:

The blank node labels are scoped to the template for each solution.

So based on my understanding of the spec, this behaviour is correct.

The example in https://www.w3.org/TR/sparql11-query/#templatesWithBNodes aligns with this.

[namedgraph]

I'm sorry but I'll trust Jena on this.

Note that there are no bnodes in my query (which is what that section is about), only in my data.

[rubensworks]

Note that there are no bnodes in my query (which is what that section is about), only in my data.

This part of the spec seems to confirm your interpretation:

Blank node labels are scoped to a result set (see "SPARQL Query Results XML Format" and "SPARQL 1.1 Query Results JSON Format") or, for the CONSTRUCT query form, the result graph. Use of the same label within a result set indicates the same blank node.

Tbh, I would love to avoid this blank node weirdness, as it has given us plenty of headaches already.

@afs, what is your interpretation of https://www.w3.org/TR/sparql11-query/#templatesWithBNodes ? Did we implement it incorrectly in Comunica?

[rubensworks]

Steps that are needed to resolve this issue:

• Modify BindingsToQuadsIterator, so that the template is localized before the bindings are applied.
• Remove the localizeBlankNodes context, CLI option, and all usages, as it won't be needed anymore.