SPARQL DESCRIBE behavior with respect to the default graph

The SPARQL DESCRIBE query form lets you retrieve information about resources without knowing the structure of the data and without having to compose a query. How this information is assembled is left up to the SPARQL implementation. Neptune provides several query hints that invoke different modes and algorithms for DESCRIBE to use.

In Neptune's implementation, regardless of the mode, DESCRIBE only uses data present in the SPARQL default graph. This is consistent with the way SPARQL treats datasets (see Specifying RDF Datasets in the SPARQL specification).

In Neptune, the default graph contains all unique triples in the union of all named graphs in the database, unless particular named graphs are specified using FROM and/or FROM NAMED clauses. All RDF data in Neptune is stored in a named graph. If a triple is inserted without a named-graph context, Neptune stores it in a named graph designated http://aws.amazon.com/neptune/vocab/v01/DefaultNamedGraph.

When one or more named graphs are specified using the FROM clause, the default graph is the union of all unique triples in those named graphs. If there is no FROM clause and there are one or more FROM NAMED clauses, then the default graph is empty.

SPARQL DESCRIBE examples

Consider the following data:

PREFIX ex: <https://example.com/>

GRAPH ex:g1 {
    ex:s ex:p1 "a" .
    ex:s ex:p2 "c" .
}

GRAPH ex:g2 {
    ex:s ex:p3 "b" .
    ex:s ex:p2 "c" .
}

ex:s ex:p3 "d" .

For this query:

PREFIX ex: <https://example.com/>
DESCRIBE ?s
FROM ex:g1
FROM NAMED ex:g2
WHERE {
  GRAPH ex:g2 { ?s ?p "b" . }
}

Neptune would return:

ex:s ex:p1 "a" .
ex:s ex:p2 "c" .

Here, the graph pattern GRAPH ex:g2 { ?s ?p "b" } is evaluated first, resulting in bindings for ?s, and then the DESCRIBE part is evaluated over the default graph, which is now just ex:g1.

However, for this query:

PREFIX ex: <https://example.com/>
DESCRIBE ?s 
FROM NAMED ex:g1 
WHERE { 
  GRAPH ex:g1 { ?s ?p "a" . } 
}

Neptune would return nothing, because when a FROM NAMED clause is present without any FROM clause, the default graph is empty.

In the following query, DESCRIBE is used with no FROM or FROM NAMED clause present:

PREFIX ex: <https://example.com/>
DESCRIBE ?s 
WHERE { 
  GRAPH ex:g1 { ?s ?p "a" . } 
}

In this situation, the default graph is composed of all the unique triples in the union of all the named graphs in the database (formally, the RDF merge), so Neptune would return:

ex:s ex:p1 "a" . 
ex:s ex:p2 "c" . 
ex:s ex:p3 "b" .
ex:s ex:p3 "d" .