Re: Comments from Julien Homo - TCs

Hi Julien

Thanks for your work!

On Apr 18, 2012, at 2:30 PM, Julien Homo wrote:

> Hi Borris and Richard.
> 
> Thank you very much for your replies and for the update of TH software, the execution of the test suite works. 
> 
> I have some questions about it :
> 
> - You said in the R2RML and Direct Mapping Cases that "Databases that do not support the standard SQL 2008 datatypes cannot pass the tests".
> 
> I think that it is difficult to find a database that fully implements the SQL standard 2008. For example postGreSQL conforms to at least 160 mandatory features required for full Core conformance out of 179 (http://www.postgresql.org/docs/9.1/static/features.html), Oracle products provide full or partial conformance with Core SQL:2008 (http://docs.oracle.com/cd/E14072_01/server.112/e10592/ap_standard_sql003.htm) and I have not yet find information on mysql. Therefore, I was wondering if the WG had an ideal database to advise to pass the official tests ? 

No, we don't have it ….

> 
> In my case I have done mostly testing with PostgreSQL before being confronted with binary types. I tried also with MySQL but I have many syntax errors and particularly quotes around table names in SQL requests. 
> As you suggested, I can send my slightly modified test suite to the WG mailing lists for validation of support R2RML over non-SQL-compatible databases like PostgreSQL.

Yes, it would be great for us…. thanks!

> 
> -  In test cases, we can see that "Throughout all test cases, we use the base IRI <http://example.com/base/>: to turn relative IRIs generated by the Direct Mapping into absolute IRIs". However you don't mention base URI in direct mapping results ("@base <http://example.com/base/> .") and execution of test suite with TH software fails. Have to manually manage this ?
> 

My mistake, sorry.
I updated the DirectGraphs in the repository. You can try it now.

> - About TH software, db2triples in Direct Mapping mode use a deterministic mechanism to generate blank nodes with percent-encoding node ID as "_:1-deptCity-Cambridge.deptName-accounting.lead-8.name-pencil%20survey". TH       software apparently does not support this naming, an error is raised. I'm surprised because a Sesame triplestore accepts such a node. Do you agree with this TH software exception ?

TH software relies on Jena, and yes …. there are some issues with "." and "%" in the blank node labels with Jena.
I need to discuss this with the WG and I'll let you know.

Thanks again!

Boris