sparql — consolidated language guide — DedSec Project

SPARQL

RDF stores subject–predicate–object triples. SPARQL matches graph patterns; a missing triple is not automatically a false statement. Use a local SPARQL 1.1 editor or triple store with a disposable dataset.

Reproducible two-file example

Save and import friends.ttl:

@prefix ex: <https://example.org/> .
ex:ada ex:knows ex:grace, ex:lin .
ex:grace ex:knows ex:lin .
ex:lin ex:knows ex:ada .

Execute friends.rq against the same dataset:

PREFIX ex: <https://example.org/>
SELECT ?person (COUNT(?friend) AS ?connections)
WHERE {
  ?person ex:knows ?friend .
  FILTER(?person != ?friend)
}
GROUP BY ?person
ORDER BY DESC(?connections) ?person

Expected rows are Ada=2, Grace=1 and Lin=1. The GROUP BY defines which values are counted together. The secondary ORDER BY expression makes ties deterministic.

Boundary cases

Duplicate Ada→Lin in Turtle: RDF graph set semantics means the count stays 2. Add Lin→Lin: the FILTER excludes it. Remove GROUP BY while selecting ?person to observe a query validation error.

Exercise: add HAVING (COUNT(?friend) >= 2); then use OPTIONAL for labels and explain unbound variables.

Troubleshooting: for zero rows, start with SELECT * WHERE { ?s ?p ?o } LIMIT 5. Check the selected graph and prefix IRIs before rewriting the query. Never update a public endpoint without permission.

Reference: https://www.w3.org/TR/sparql11-query/

SPARQL: applied exercises

Make the data explicit

The starter asks for up to five subjects matching any triple. That depends entirely on the graph supplied by the RDF store; an empty graph returns no bindings. Avoid interpreting the query as a database-independent demo. In an isolated local RDF dataset save this people.ttl:

@prefix ex: <https://example.org/people/> .
ex:ada ex:age 21 .
ex:ben ex:age 17 .
ex:cy ex:age 21 .

Then run the following query with a SPARQL 1.1 engine, loading people.ttl as its default graph:

PREFIX ex: <https://example.org/people/>
SELECT ?person WHERE {
  ?person ex:age ?age .
  FILTER (?age >= 18)
}
ORDER BY ?person

The expected bindings, in this order, are ex:ada and ex:cy; implementations may display full IRIs instead of prefixed names. This is a predicted fixture outcome, not a recorded engine run.

Understand graph matching

The triple pattern binds ?person and ?age. FILTER accepts only numeric ages meeting the threshold. A resource lacking ex:age does not match the mandatory pattern; it is not treated as age zero. Add ex:dan ex:label "Dan". without age and confirm it is absent. Use an OPTIONAL block only when you intentionally want a row even without an age, and decide how to handle an unbound variable.

Find a real failure

Deliberately remove the PREFIX ex: declaration and capture the parser's message about an undeclared prefix. This tests query syntax, unlike changing the threshold to 30, which should be a valid query returning an empty result set. Distinguish parsing failure from a successful query with no matches.

Extension and evidence

Insert a second ex:age value for a person and explain why a SELECT result can repeat that person unless DISTINCT is specified. Add SELECT DISTINCT and compare cardinality. Avoid sending synthetic personal names to a public endpoint; the local fixture is sufficient. Record graph loading method, SPARQL engine version and exact output for each mutation. Standard: https://www.w3.org/TR/sparql11-query/