Re: RDF query testcases?

On Thursday, January 16, 2003, at 03:59  PM, Libby Miller wrote:

> query engine. But to be more general, we'd have to specify a file 
> format of
> some kind - I'm wondering whether RDF/XML is the one or whether
> N-triples would improve the matter or complicate things....what do you
> think?

yes indeed :-)

you might need some general manifest or schema which in addition to 
properties like query language, title, description, author, query text, 
num of results and query data specifies also the input format for the 
data; this is in some cases implicit into the query language self but 
if you want to have other tools to use your test data you have to stick 
to some vocabulary, something like: RDF/XML, N-triples, N3, XHTML, raw 
XML(?) . A solution could also be to stick to the mime-type of those.

another problem I have found relevant while building test-cases, is the 
possibility to specify more than one data source in the query, and in 
our implementation is possible. Our problem (at the moment) is that the 
resulting model is just a union one-by-one of triples and is not a real 
"RDF semantics aware" graph merge (bNodes smushing and stuff like 
that). Do you have got any idea or solution about that or is too 
complicated?

>
> Would you be happy for me to try to get your tests into my RDF manifest
> or something similar? (I don't have an RDF schema for that btw 
> Alberto).

yes, sure! I was thinking to try the other way round as well :-) add 
some properties in the use-cases and examples schema [1] for the tests 
and data.

one thing I realised is that RDF query test-cases might be useful for 
two things:

- query parser tests (syntax mostly and no real data)
- examples (real world examples with real data)

in fact we have separated the two in our implementation and the RDF 
Query and Rules survey is only about the second flavor. In your Inkling 
test suite you have got a mix of the two (e.g. queries/sqtest9.squish 
vs. queries/q1.sq ). In addition your manifest is assuming that the 
input data source to be queried is specified into the SquishQL query 
self, while in the general case you should specify also the input data 
in the manifest file I guess. Some tools do not allow to specify the 
input source directly in the syntax.

After looking at your test cases manifest I would add/modify the 
following properties:

- add queryURL (of the query text itself)
- add optional queryText (for inline query text)
- modify dataURL (change meaning and use it to point to the actual data 
source)
- add dataText property (for inline data)
- add dataFormat (mime-tyep or something)

it would be also nice at some moment to "link" all this RDF query tests 
stuff to Dave/Jan RDF Test Cases suite at some time; then have a common 
set of sources to test against RDF parsers, storages, query parsers and 
query engines.

Alberto


[1] 
http://rdfstore.sourceforge.net/2002/06/24/rdf-query/QueryUseCase.rdfs

Received on Thursday, 16 January 2003 12:02:22 UTC