Question

I have problems with using the OPTIONAL phrase in the SPARQL statements. When I’m quering dbpedia like this:

CONSTRUCT { ?guitarist rdfs:label ?name . ?guitarist rdfs:comment ?desc . ?guitarist dbpprop:placeOfBirth ?placebirth }
WHERE {
  ?guitarist dbpprop:wikiPageUsesTemplate <http://dbpedia.org/resource/Template:Infobox_musical_artist> . 
  ?guitarist rdfs:label ?name .
  ?guitarist rdfs:comment ?desc .
  ?guitarist dbpprop:placeOfBirth ?placebirth .
  FILTER ( lang(?name) = "en" && lang(?desc) = "en" )
} 

Roger Waters birthplace and other data are returned. But when I turn it to this, all Roger_Waters records are missing:

CONSTRUCT { ?guitarist rdfs:label ?name . ?guitarist rdfs:comment ?desc . ?guitarist dbpprop:placeOfBirth ?placebirth }
WHERE {
  ?guitarist dbpprop:wikiPageUsesTemplate <http://dbpedia.org/resource/Template:Infobox_musical_artist> . 
  ?guitarist rdfs:label ?name .
  ?guitarist rdfs:comment ?desc .
  OPTIONAL { ?guitarist dbpprop:placeOfBirth ?placebirth }
  FILTER ( lang(?name) = "en" && lang(?desc) = "en" )
}

What I’m doing wrong – I suppose the complete records not to be missing after applying OPTIONAL…

Answer 1

The DBpedia endpoint returns partial results for queries that it deems too expensive.

Doing a SELECT COUNT(*) WHERE … with these two queries, it looks like the first query should return 8k results and the second should return 60k results. Your CONSTRUCT should generate 2-3 triples per result, so the first query should yield 16-24k triples, and the second should yield 120-180k. When I run the queries, I get the curious number of exactly 10001 triples from either query. So clearly the results are truncated.

Try LIMIT and OFFSET (possibly with an ORDER BY).