On 23/09/11 18:19, Tim Harsch wrote:
Thanks Andy, I have additional questions now.  I've now checked out 
Fuseki-0.2.1-SNAPSHOT from maven and ran tests with it.  The behavior I noticed 
in Fuseki-0.2.0 continues.  I configured it to run with a TDB database (which 
has many named graphs in it) and if I attempt to use FROM/FROM NAMED in a query 
passed in I get:
"Error 400: Query may not include a dataset description (FROM/FROM NAMED)"

It should work so that needs fixing.

A quick hack workaround is remove the check :-) in SPARQL_QueryDataset.validateQuery

and it works for me. The only reason not to do this as a complete fix is that it means different things to different storage layers. Maybe it' needs to be an option in the config file (per service) but I'm not rushing to add lots of options as it all gets confusing.

Which is what led me to giving Joseki a spin, cause the last I thought heard 
from the mail list was that support of FROM/FROM NAMED in Fuseki is TBD (sorry 
don't know where I read it, I may be in error).  I think it gets rejected by 
the servlets.

So I'm confused by your response about Joseki, when you say "the one case of having 
FROM-web loading AND a fixed dataset isn't supported on the same service endpoint".

I am attempting to allow changing the graph the query executed on
through FROM/FROM NAMED, but I do not wish RDF datasets pulled from the
web.  Which I think is this:
       joseki:allowExplicitDataset       "true"^^xsd:boolean ;
       joseki:allowWebLoading            "false"^^xsd:boolean ;

I would expect "the one case of having FROM-web loading AND a fixed dataset isn't 
supported on the same service endpoint" to look like this config
       joseki:allowExplicitDataset       "false"^^xsd:boolean ;
       joseki:allowWebLoading           "true"^^xsd:boolean ;


The problem is that Joseki does all the FROM processing; it overrides the storage layer processing that might happen. The design predates (by a long time)

This is just one of the several reasons why I started Fuseki as a redesign of Joseki. New capabilities (SPARQL Graph Store Protocol), less need for arbitrary processors, better balance of responsibilities.

        Andy

Reply via email to