Type issues with SPARQL `graph.query()`

[multimeric]

Consider the following example:

from rdflib import Graph, URIRef, Literal
from typing import reveal_type

graph = Graph()
graph.add((URIRef("http://example.org/subject"), URIRef("http://schema.org/name"), Literal("foo")))
for result in graph.query("""
    SELECT ?name ?description
    WHERE {
        ?file_name <http://schema.org/name> ?name .
        OPTIONAL {
            ?file_name <http://schema.org/description> ?description .
        }
    }
"""):
    reveal_type(result)
    reveal_type(result.name)
    reveal_type(result.description)

Firstly, the type of result.description is incorrect.
The static type according to the Pyright type checker is:

information: Type of "result.description" is "Unknown | Identifier"

But at runtime we get:

Runtime type is 'NoneType'

So, because of OPTIONAL{}, we need to type result variables as Identifier | None.

Secondly, as noted in #2283, result is an annoying type union. Pyright says Type of "result" is "Tuple[Node, Node, Node] | bool | ResultRow". The reason why this should not be typed as a union is because the syntax of the SPARQL query defines the result type, meaning that I know statically which result type it should be, but I have no way to express this. Downstream this union causes issues because:

error: Cannot access attribute "description" for class "_TripleType"
error: Cannot access attribute "description" for class "bool"

[nicholascar]

If this issue is addressed, I would also like to see some improvement in the alignment of SPARQL query result objects with RDFLib Graph objects so we don't have to do hacks like this: https://github.com/Kurrawong/kurra/blob/main/kurra/sparql.py#L38-L42

[ajnelson-nist]

Issue 2283 and PR 2320 seem like they would address this.

I would like to get back to finishing 2320. My spirit is willing, but my calendar might be weak.