[JENA-874] Subqueries not functioning properly when Filter clauses are present

• Type: Bug
• Status: Closed
• Priority: Major
• Resolution: Fixed
• Affects Version/s: Jena 2.13.0
• Fix Version/s: Jena 2.13.0
• Component/s: ARQ
• Assignee: Andy Seaborne
• Reporter: Stephen Allen
• Created: 30/Jan/15 01:28
• Updated: 13/Mar/15 19:39
• Resolved: 31/Jan/15 18:03

Description

Queries with Subqueries are not functioning properly when a filter clause exists. This is a regression, as it worked properly in 2.11.2 (I haven't checked 2.12.1).

Example:

data.ttl

<http://r1> a <http://type> .
<http://r1> a <http://type2> .
<http://r2> a <http://type> .
<http://r2> a <http://type2> .
<http://r3> a <http://type> .
<http://r3> a <http://type2> .

Bind Query

select *
where {
  graph <http://graph1> {
    {
      select distinct (?_id as ?id)
      where {
        values ?_type {
          <http://type2>
        }
        ?_id a ?_type .
        ?_id a <http://type> .
        filter ( ?_id = <http://r1> || ?_id = <http://r2> )
      }
    }
    ?id a <http://type> . bind (<http://type2> as ?type) . ?id a ?type .
  }
}
order by ?id

Bind Algebra

(base <http://example/base/>
  (order (?id)
    (graph <http://graph1>
      (join
        (extend ((?type <http://type2>))
          (join
            (distinct
              (project (?id)
                (extend ((?id ?_id))
                  (filter (|| (= ?_id <http://r1>) (= ?_id <http://r2>))
                    (join
                      (table (vars ?_type)
                        (row [?_type <http://type2>])
                      )
                      (bgp
                        (triple ?_id <http://www.w3.org/1999/02/22-rdf-syntax-ns#type> ?_type)
                        (triple ?_id <http://www.w3.org/1999/02/22-rdf-syntax-ns#type> <http://type>)
                      ))))))
            (bgp (triple ?id <http://www.w3.org/1999/02/22-rdf-syntax-ns#type> <http://type>))))
        (bgp (triple ?id <http://www.w3.org/1999/02/22-rdf-syntax-ns#type> ?type))))))

Filter Query

select *
where {
  graph <http://graph1> {
    {
      select distinct (?_id as ?id)
      where {
        values ?_type {
          <http://type2>
        }
        ?_id a ?_type .
        ?_id a <http://type> .
        filter ( ?_id = <http://r1> || ?_id = <http://r2> )
      }
    }
    ?id a <http://type> . ?id a ?type . filter (?type != <http://type> ) .
  }
}
order by ?id

Filter Algebra

(base <http://example/base/>
  (order (?id)
    (graph <http://graph1>
      (filter (!= ?type <http://type>)
        (join
          (distinct
            (project (?id)
              (extend ((?id ?_id))
                (filter (|| (= ?_id <http://r1>) (= ?_id <http://r2>))
                  (join
                    (table (vars ?_type)
                      (row [?_type <http://type2>])
                    )
                    (bgp
                      (triple ?_id <http://www.w3.org/1999/02/22-rdf-syntax-ns#type> ?_type)
                      (triple ?_id <http://www.w3.org/1999/02/22-rdf-syntax-ns#type> <http://type>)
                    ))))))
          (bgp
            (triple ?id <http://www.w3.org/1999/02/22-rdf-syntax-ns#type> <http://type>)
            (triple ?id <http://www.w3.org/1999/02/22-rdf-syntax-ns#type> ?type)
          ))))))

For both queries, the results should look like:

--------------------------------
| id          | type           |
================================
| <http://r1> | <http://type2> |
| <http://r2> | <http://type2> |
--------------------------------

However, only the Bind query actually works correctly. The Filter query returns no rows.