Re: [R2RML Test Cases]

Hi Eric

On 27/11/2010 14:51, Eric Prud'hommeaux wrote:
> * Alex Miller<alexdmiller@yahoo.com>  [2010-11-26 18:39-0800]
>>> From: Eric Prud'hommeaux<eric@w3.org>
>>> To: Boris Villazón Terrazas<bvillazon@fi.upm.es>
>>> Cc: Michael Hausenblas<michael.hausenblas@deri.org>; RDB2RDF WG
>>> <public-rdb2rdf-wg@w3.org>
>>> Sent: Fri, November 26, 2010 6:58:52 PM
>>> Subject: Re: [R2RML Test Cases]
>>>
>>> * Boris Villazón Terrazas<bvillazon@fi.upm.es>  [2010-11-24 13:19+0100]
>>>> Dear all,
>>>>
>>>> Thanks for the feedback about the test cases.
>>>> I'll update the test cases including all the feedback we got from
>>>> Juan, ericP, Souri, etc.
>>>>
>>>> During the telco, Eric created a new test case, you can check it at:
>>>> http://www.w3.org/2001/sw/rdb2rdf/wiki/R2RML_Test_Cases#R2RMLTC0005b
>>>>
>>>> Eric wants to discuss about this particular test case.
>>> I added a discriminating query 'cause it's easier to compare the
>>> results of selected SPARQL and SQL queries than to do graph compare.
>>>
>>> Actually, this brings up an interesting question about re-using the
>>> SPARQL test suite. How many of our implementors will offer a SPARQL
>>> interface and already exploit the SPARQL suite? Do we want to try to
>>> make their lives easier by using mostly the same test suite?
>> I think reusing SPARQL from existing test suites would be extremely
>> helpful for implementors.
> The SPARQL test suite takes as input an RDF database (a set of
> default and named graphs) and a query and produces one of:
>    result set (SELECT, ASK)
>    graph (CONSTRUCT)
>    RDF database (INSERT, DELETE)
>
> We can expect that most SPARQL implementations can:
>    compare result sets
>    compare graphs (that's actually a bit difficult)
>    compare RDF databases (by comparing names and contents of graphs)
>
> The SPARQL tests are defined in manifest files with include
>    name
>    description
>    input
>    output
>    required extensions
>    approval status
You are referring to this, right?
http://www.w3.org/2001/sw/DataAccess/tests/r2


> For direct mapping, we'd like a manifest to associate an RDB defined
> by some SQL and a result graph. For r2rml, a an RDB, and R2RML config,
> and a result RDF database. This seems like pretty good synergy with
> the SPARQL tests.

Just to clarify,
- Direct Mapping Test Cases will include:
     RDB defined by some SQL
     Result Grahp

- Following Michael's suggestiion R2RML Test Cases include:
    Id
    Title
    Purpose
    Specification Reference
    Review status
    Input: RDB defined by some SQL
    Expected result: RDF graph
    R2RML Mapping

The plan is to describe the test cases using the Test Metadata 
vocabulary  [1] in RDFa.

So, now you are suggesting to use similar manifest files from the SPARQL 
test cases?

Boris


  [1] http://www.w3.org/2006/03/test-description

Received on Sunday, 28 November 2010 23:53:30 UTC