Clarify `VALUES` with repeated variable names

[Tpt]

It seems to me that the SPARQL spec is not stating what should/must be the behavior of SPARQL implementations when VALUES clause contain repeated variable.

For example, in VALUES (?a ?a) { (1 2) } should the behavior be:

1. a syntax error
2. an empty solution multiset

And for VALUES (?a ?a) { (1 1) } should the behavior be:

1. a syntax error
2. a single solution ?a -> 1

Quick survey:

• Jena and Blazegraph raise a syntax error (option 1)
• Oxigraph, QLever and Comunica parse the first example as ?a -> 2

Proposal: extend the grammar note 11 to add and variables must not be repeated in the variable list.

[afs]

It might be nice to also add a sentence to this effect in 10.2.1 VALUES syntax, after the third example.

[TallTed]

Do you have a full query (or several) that can be run on other SPARQL engines?

[Tpt]

@TallTed Sure SELECT * WHERE { VALUES (?a ?a) { (1 2) } }

[TallTed]

FWIW, current Virtuoso appears to deliver the empty solution multiset, so your proposed change would make us non-compliant, if I understand it correctly.

[Tpt]

so your proposed change would make us non-compliant

Yes! My proposed change is to make it a parse error

[afs]

While Jena does raise an exception, it's not a proper parse error; it's an internal consistency check.

The proposal to make it a syntax error is the right answer.