Invalid warnings for predicate-object list when one of the objects is a blank node containing a predicate path

[dwvisser]

Here's the simplest query I find that reproduces the issue:

PREFIX ex: <http://example.org/>

SELECT * WHERE {
  ?a ex:foo ?b ;
     ex:bar [ ex:baz/ex:buz ?c ].
}

The issue seems to be occur whenever a graph pattern contains:

1. a predicate-object list, where
2. the second (or later) object is a blank node, and
3. the blank node contains a predicate path.

This bug is also filed at https://openrdf.atlassian.net/browse/SES-2255

[LaurensRietveld]

Are you sure this is valid SPARQL 1.1? If I look at the spec, I see that the BlankNodePropertyList ([99]) lists verbs (i.e. IRI/var) as possible property values, but not property paths

[dwvisser]

I think that TriplesSameSubjectPath allows this to be valid SPARQL, using

• VarOrTerm PropertyListPathNotEmpty
• PropertyListPathNotEmpty --> VerbSimple ObjectListPath
• ObjectListPath -> ObjectPath
• ObjectPath -> GraphNodePath
• GraphNodePath -> TriplesNodePath
• TriplesNodePath -> BlankNodePropertyListPath

which leads into property paths.

[dwvisser]

Wait. I see your point. The construction I gave in my the second bullet of my last comment is only valid for the first predicate-object pair, not the subsequent ones, which is precisely what is manifesting here.

[abrokenjester]

Hm. I think this may be an error in the SPARQL grammar. It certainly seems arbitrary to allow property paths in a blank node object for the first object after a subject/predicate pairing, but not for any subsequent ones. It doesn't look like this was done by design, to me. Agreed though that YASQE implements what the current spec prescribes.

[LaurensRietveld]

If you have an EBNF for queries that sesame supports, it should be pretty straightforward to generate a JS grammar based on that. (see https://github.com/YASGUI/YASQE/tree/gh-pages/lib/grammar)

[dwvisser]

Thank you. I have tested and Sesame does successfully run the query, despite it not being valid according to the official SPARQL 1.1 grammar. I have posted the result on the Sesame issue tracker.