W3C home > Mailing lists > Public > public-rdf-dawg-comments@w3.org > October 2012

Re: Test without data?

From: Jeen Broekstra <jeen.broekstra@gmail.com>
Date: Mon, 29 Oct 2012 14:24:48 +1300
Message-ID: <508DDAE0.10801@gmail.com>
To: public-rdf-dawg-comments@w3.org
On 25/10/12 23:40, Andy Seaborne wrote:
> 
> 
> On 23/10/12 08:09, Jan Wielemaker wrote:
>> Hi Andy,
>>
>> On 10/22/2012 10:35 PM, Andy Seaborne wrote:
>>>
>>>
>>> On 22/10/12 14:41, Jan Wielemaker wrote:
>>>> Hi,
>>>>
>>>> I'm looking into the SPARQL 1.1 test "constructwhere04 - CONSTRUCT
>>>> WHERE". This refers to data from <data.ttl>, but the manifest does not
>>>> tell us to load that file. Am I missing something (so far I always
>>>> had)?
>>>>
>>>>      Thanks --- Jan
>>>>
>>>
>>> The query itself has a FROM clause:
>>
>> Yes, I've seen that. However, I consider FROM to be the name of a graph,
>> not an instruction to load a graph from a URI. We have the update part
>> for that. Most other tests involving FROM or USING seem to load the
>> graph explicitly using a data statement in the test manifest. Could we
>> have that here too?
> 
> To quote SPARQL 1.0:
> 
> [[
> Each FROM clause contains an IRI that indicates a graph to be used to
> form the default graph.
> ]]
> 
> that might mean load - sometime systems do other things - but it means
> that the graph is placed into the dataset to be queried.  Load is a
> common way of doing that.

The point I believe Jan is making is that in most of the test suite,
it's the job of the manifest file to set up each test's environment
(query file, expected result, and dataset). This test (and, I believe, a
few others) is an exception to that rule, the result of which is that
suddenly there are two possible ways in which the dataset can be defined
for each test, and moreover, the second way requires preprocessing the
actual query being tested.

To put it another way: using this construction in the test suite  may
not, strictly speaking, be wrong, but it does make life harder for
testing, and as far as I can see, it does not test a feature that is
actually required to be implemented in the fashion the test seems to
assume (as you say yourself, nowhere is it specified that a SPARQL
engine must resolve a FROM clause and load the data, but this test seems
to assume that it does).


Regards,

Jeen
Received on Monday, 29 October 2012 01:25:19 GMT

This archive was generated by hypermail 2.2.0+W3C-0.50 : Monday, 29 October 2012 01:25:20 GMT