Clarify comparison of literals in SPARQL test suite #59
Conversation
|
Yes, https://www.w3.org/TR/rdf-concepts/#section-graph-equality defines same-term isomorphism; only blank nodes are remapped. There is a level above basic RDF, called D-entailment, that considers "same value". Tests should be marked if they are value-sensitive but usually is in the test as FILTERS -- Test should be marked up if they are value sensitive but may be some got missed. PS as changes get made, can we keep text to a length that's reviewable in github please? |
There was a problem hiding this comment.
I actually think that no change is necessary. As @afs says, if there are specific tests should be marked if they use value comparison, but this doesn't typically happen with CONSTRUCT, which is where graph isomorphism comes in, but with SELECT or ASK where the FILTER determines value equality. For example with expr-equals/query-eq2-1.rq.
PREFIX xsd: <http://www.w3.org/2001/XMLSchema#>
PREFIX : <http://example.org/things#>
SELECT ?v1 ?v2
WHERE
{ ?x1 :p ?v1 .
?x2 :p ?v2 .
FILTER ( ?v1 = ?v2 ) .
}My feeling is that we should close the PR with no merge as being unnecessary. SPARQL 1.2 should be updated based on RDF 1.1 principles, and add any necessary clarification.
|
(In spite of saying I approved, my comment was that this should be closed without change). |
|
I like the implementation advice in
As a pedant, I think that approach of testing graph isomorphism works for comparing result set with rows of (?s, ?p, ?o), i.e. graphs. I propose changing that first sentence to
|
|
The goal of this PR was to document the suggestion from @cygri and @afs:
The consensus from the comments above seem to be that instead of comparing literals in all tests like this, only tests that require this comparison should be marked with a flag. I'm happy with either approach, so I'm fine with closing this PR (and creating a new one) if everyone agrees. |
Closes #58