`ExprTripleTermSubject` grammar rule allows to write always invalid expression

[Tpt]

The current SPARQL grammar states:

[137]  ExprTripleTerm  ::=  '<<(' ExprTripleTermSubject Verb ExprTripleTermObject ')>>'
[138]  ExprTripleTermSubject  ::=  iri | RDFLiteral | NumericLiteral | BooleanLiteral | Var | ExprTripleTerm

And the Triple( function for which ExprTripleTerm is a syntactic sugar is defined using: If the 3-tuple (subj, pred, obj) is an RDF triple (that is, subj is an IRI or blank node; pred is an IRI; and obj is an IRI, triple term, blank node or literal) the function returns a triple term with these three elements. Otherwise, the function raises an error..

It means that expressions like <<( "foo" ex:p "bar" )>> will always return an error at runtime. I tend to think it might be nicer to make them invalid syntax and change the grammar into:

[137]  ExprTripleTerm  ::=  '<<(' ExprTripleTermSubject Verb ExprTripleTermObject ')>>'
[138]  ExprTripleTermSubject  ::=  VarOrIri

[afs]

[138]  ExprTripleTermSubject  ::=  VarOrIri

means it is not possible to a syntax time condition.

Maybe the text TRIPLE should be changed reflect symmetric RDF.

[Tpt]

means it is not possible to a syntax time condition.

Sorry for the naive question. What are you referring to with "time condition"?

Maybe the text TRIPLE should be changed reflect symmetric RDF.

This is definitely the other option.

[afs]

What are you referring to with "time condition"?

It can't only be enforced with a static (parse-time) syntax check though a syntax change would help users.
Because of variables, it is still possible to pass a literals etc to the triple term creator, just like in the graph a pattern.

Maybe the text TRIPLE should be changed reflect symmetric RDF.

This is definitely the other option.

since we seem to be "all in" on symmetric RDF, I tend towards changing TRIPLE. CONSTRUCT will still reject non-RDF triples.

To get github to link issues : #282 (use in VALUES).

[TallTed]

since we seem to be "all in" on symmetric RDF

Are we? And if so, what does that mean? I think this would be a bigger change from RDF 1.1 than we are chartered for, and even if we decide to make this change anyway, it certainly will have a larger downstream impact than seems to have been considered.

My immediate inclination is to make the originally suggested change, i.e., from —

[138]  ExprTripleTermSubject  ::=  iri | RDFLiteral | NumericLiteral | BooleanLiteral | Var | ExprTripleTerm

— to —

[138]  ExprTripleTermSubject  ::=  VarOrIri

— which I think keeps TripleTerms parallel to Triples.