Question about SPARQL tests

Hi everyone,

I'm writing a tool that goes through the existing SPARQL tests and
executes them automatically. I ran into a couple of issues, and while
asking Andy about it he pointed out that I should post to the group,
so it gets in with other errata.

According to the test case README, Query-Evaluation Tests will have an
mf:action, and this will have "at a minimum" both qt:data and qt:query
properties. Also, the property qt:graphData may also exist, indicating
NAMED GRAPHs to use.
http://www.w3.org/2001/sw/DataAccess/tests/README

Andy commented:
> There are tests which have neither - because the dataset is described
> by FROM/FROM NAMED.

However, I've found that data-r2/graph/manifest.ttl has 2 tests in it
that only includes qt:graphData and no qt:data property (these are
tests :dawg-graph-02 and :dawg-graph-03).
http://www.w3.org/2001/sw/DataAccess/tests/data-r2/graph/manifest.ttl

Andy commented:
>I remember a discussion ... a dataset always has a default graph (that's
> in the spec) - it's assumed to be empty unless mentioned in the test
> suite, which is the case here.  Maybe it would be better to load an
> empty graph because the query spec does not force it to be empty
> (being the merge of all named graphs being the obvious counter
> example).
>
> graph-03 does not touch the default graph so it does not matter.
>
> BTW The dataset description might be in the query (FROM etc) - there
> some queries like this.

I'm fine with Andy's rationale, however it *does* conflict with the
description of the test case structure, which is an issue for an
automated tool.

I also agree with Andy's suggestion of loading an empty graph for the
default graph. It's not an issue in my case (Mulgara starts with an
empty default), but the definition should be pinned down more
carefully.

Regards,
Paul Gearon

Received on Friday, 12 June 2009 17:02:06 UTC