Re: SERVICE tests available

  Agreed. I commited the change to the SVN.

cheers,

Carlos

On 21/06/2011 13:12, Lee Feigenbaum wrote:
> On 6/21/2011 11:18 AM, Gregory Williams wrote:
>> On Jun 21, 2011, at 11:07 AM, Carlos Buil Aranda wrote:
>>
>>> Hi all,
>>>
>>> I updated the manifest.ttl in the service tests using Greg's
>>> proposal for action 482.
>>
>> I have a question about test :service6. The manifest has this:
>>
>> :service6 rdf:type mf:QueryEvaluationTest ; mf:name    "SERVICE test
>> 6" ; dawgt:approval dawgt:NotClassified ; mf:requires
>> mf:BasicFederation ; mf:action [ qt:query<service06.rq>  ;
>> qt:data<data06.ttl>  ; qt:serviceData [
>> qt:endpoint<http://example1.org/sparql>  ;
>> qt:data<data06endpoint1.ttl> ] ; qt:serviceData [
>> qt:endpoint<http://invalid.endpoint.org/sparql>  ; qt:data     "" ] ]
>> ; mf:result<service06.srx>  .
>>
>> I'm not sure how I'm meant to interpret the string literal value for
>> qt:data on the "invalid" endpoint. Since this test seems to be
>> testing the SERVICE SILENT operation (for endpoints that don't exist
>> or don't respond), I'd prefer some more explicit way to indicate
>> that<http://invalid.endpoint.org/sparql>  is an endpoint mentioned in
>> the query but that it is intended for that endpoint to not actually
>> exist (for the purposes of the test).
>>
>> As it is now, I'd think the two reasonable interpretations of {
>> :service6 qt:data "" } would be to either throw an error (expecting
>> an IRI for the data but finding a literal), or ignoring the literal
>> and setting up a mock endpoint at that address with no data in it.
>> Neither of these will result in the expected behavior.
>>
>> I think the best way forward would either be to come up with some new
>> syntax for indicating an endpoint that is used in the query but that
>> is not meant to participate in query evaluation, or two simply remove
>> that serviceData block and let the test harness notice that the query
>> mentions an endpoint that wasn't described in the manifest.
>
> Agreed. I prefer the 2nd option -- just leave out that endpoint from 
> the manifest.
>
> Lee
>
>>
>> Thoughts?
>>
>> thanks, .greg
>>
>>
>>
>

Received on Wednesday, 22 June 2011 14:19:32 UTC