W3C home > Mailing lists > Public > public-sparql-dev@w3.org > October to December 2014

Re: SparqlScore : bugs or tests 1.1 obsoletes ?

From: Andy Seaborne <andy@apache.org>
Date: Wed, 08 Oct 2014 09:07:24 +0100
Message-ID: <5434F0BC.6090806@apache.org>
To: public-sparql-dev@w3.org
On 06/10/14 14:27, Karima Rafes wrote:
> Hi
> The project TFT/SparqlScore will be presented during the next Semantic
> Web Challenge.
> http://challenge.semanticweb.org/2014/submissions/swc2014_submission_4.pdf
> http://sparqlscore.com/
> We can now add your endpoint in the system. If you want test your
> software, you can install an instance of your endpoint (empty without
> security) and send me the http address. Each day, you will see the
> result of tests of your software. You can remove your software of the
> list when you want.
> About 95% of tests passed with Fuseki but there are some tests obsoletes.
> Somebody can check and update the tests of sparql 1.1 ?
> After I will see and fix my last bugs.
> Thanks
> Best regards
> Karima Rafes
> karima.rafes@gmail.com

Fuseki should pass all of the SPARQL 1.1 Query Language and SPARQL 1.1 
Update.  You should be running the server in strict mode.

sq14 - limit by resource

In addition, in the version you are running,

"STRDT() TypeErrors"
"STRLANG() TypeErrors"

should pass because the published results are RDF 1.0 and the published 
server is running with RDF 1.0 semantics.

Looking sq14, the W3C published results have 11 triples but the expected 
results in sparqlscore have 14.

For example: there is the triple in the expected results:
      [s] => <http://example.org/ns#a>
      [p] => <http://www.w3.org/1999/02/22-rdf-syntax-ns#type>
      [o] => "Alan"

Where does this come from?

Apache Marmotta: (I sampled a few answers)

BNODE() --
Just because the labelling is a different style does not automatically 
mean the test fails.  You need blank node isomorphism testing.

constructwhere02  :
Duplicates in a graph syntax are fine - the resulting graph after 
parsing the syntax is a set of triples so Marmotta's way of doing it works.

This might apply to other constructwhere tests.


This non-W3C test:
"10 Substract date" is at odds with XSD.

The difference of two xsd:dateTime is an xsd:duration, not an integer.

as you say "SPARQLScore is an attempt to evaluate the conformance of 
triplestores to the W3C standards. "

	Hope that helps
Received on Wednesday, 8 October 2014 08:07:54 UTC

This archive was generated by hypermail 2.3.1 : Tuesday, 6 January 2015 20:15:52 UTC