Re: CSV/TSV results test cases and suggested adaption of JSON results test cases

On 10/08/11 17:18, Gregory Williams wrote:
>
> On Aug 10, 2011, at 4:07 AM, Axel Polleres wrote:
>
>>
>> On 9 Aug 2011, at 17:23, Gregory Williams wrote:
>>
>>> I'm mostly happy with the 01 and 02 tests. csv03 still has a c14n
>>> problem ("1"^^xsd:decimal). Can we change that like with did with
>>> "5.5"?
>>
>> Well, I deliberately isolated that in a separate test case
>> collecting more "corner cases" than in 01 and 02 , but I still
>> think we shouldn't throw it away, should we? It is a valid test
>> case in the end, isn't it?
>
> I'm not sure if it's "valid". If I've understood what's going on, it
> seems to me that it's testing something that's entirely outside the
> scope of SPARQL, and so definitely shouldn't be a required test.
> Marking it with an mf:requires is one possibility, but I'm not sure
> we've had a negative requirement before. ":csvtsvXX mf:requires
> :NoC14N" seems really strange to me.
>
> .greg
>
>

The issue of canonicalization as a part of value-based processing is a
specific issue to the csv-tsv results.  The same can be true across many
of the tests, not just results formats (e.g. use STR).

In this case there is an additional special issue - the canonical form
of decimals which are integer valued changes between XSD 1.0 to XSD 1.1.

The test case is only valid for XSD 1.1 and D-entailment (taking 
canonicalization is an implementation of partial D-entailment).

Maybe keeping the tests focused on result formatted specifics is easier 
for implementers reporting SPARQL 1.1 conformance.

	Andy

Received on Wednesday, 10 August 2011 16:43:22 UTC