SPARQL using subquery with limit

Question

I am developing a java application that uses ARQ to execute SPARQL queries using a Fuseki endpoint over TDB.

The application needs a query that returns the place of

Answer 1

You can use limits with subqueries

You can use limits in subqueries. Here's an example:

select ?x ?y where {
  values ?x { 1 2 3 4 }
  {
    select ?y where {
      values ?y { 5 6 7 8 }
    }
    limit 2
  }
}
limit 5

---------
| x | y |
=========
| 1 | 5 |
| 1 | 6 |
| 2 | 5 |
| 2 | 6 |
| 3 | 5 |
---------

As you can see, you get two values from the subquery (5 and 6), and these are combined with the bindings from the outer query, from which we get five rows in total (because of the limit).

Subqueries are evaluated innermost first

However, keep in mind that subqueries are evaluated from the innermost first, to the outermost. That means that in your query,

select ?person_id ?place_of_birth ?other_person_id
where {
    ?person_id fb:type.object.type fb:people.person .
    ?person_id fb:people.person.place_of_birth ?place_of_birth_id .
    ?place_of_birth_id fb:type.object.name ?place_of_birth .
    {
       select  ?other_person_id
       where {
       ?place_of_birth_id fb:location.location.people_born_here ?other_person_id .
       }
       LIMIT 1
     }
     FILTER (langMatches(lang(?place_of_birth),"en"))
}
LIMIT 3

you are finding one match for

?place_of_birth_id fb:location.location.people_born_here ?other_person_id .

and passing the ?other_person_id binding out into the outer query. The rest of the outer query doesn't use ?other_person_id, though, so it doesn't really have any effect on the results.

What to do instead

If you need only one other person

The application needs a query that returns the place of birth of each person and other person that was born in the same place.

Conceptually, you could look at this as picking a person, finding their place of birth, and sampling one more person from the people born in that place. You can actually write the query like that, too:

select ?person_id ?place_of_birth (sample(?other_person_idx) as ?other_person_id)
where {
    ?person_id fb:type.object.type fb:people.person .
    ?person_id fb:people.person.place_of_birth ?place_of_birth_id .
    ?place_of_birth_id fb:type.object.name ?place_of_birth .
    FILTER (langMatches(lang(?place_of_birth),"en"))
    ?place_of_birth_id fb:location.location.people_born_here ?other_person_idx .
    filter ( ?other_person_idx != ?person_id )
}
group by ?person_id ?place_of_birth

If you need more than one

This is a much trickier problem if you need more than one "other result" for each result. That's the problem in Nested queries in sparql with limits. There's an approach in How to limit SPARQL solution group size? that can be used for this.