Re: Functions to compute URIs

On Wed, 2011-06-08 at 09:50 -0500, David McNeil wrote:
> At Semtech this week I have talked to two separate people who were
> interested in R2RML and who thought it was important for R2RML to
> support the conversion of data in the relational database to URIs. I
> believe they were talking about both: 
> * the case where the data in the database contains the data necessary
> to produce the URI (e.g. "John Smith" -> "http://people/john_smith").
> There was a desire expressed to do this without writing the SQL or
> SPARQL expression to perform the conversion.
> * the case where the data in the database does not provide sufficient
> data to produce the URI (e.g. a code of "js25" needs to be translated
> to "http://people/john_smith").
> Of course R2RML has the template feature which can partially be used
> to address the first case (but not without writing a SQL/SPARQL
> expression to do case conversion on the input data), but not the
> second.

If the information is not in the database, how do you expect R2RML to
know about it?

> In the past I believe we have talked about satisfying this type of
> need (in the context of ISSUE-7), by defining a framework for plugging
> in user provided conversion functions without specifying the detailed
> form which those plugins would take.

My understanding is that RDB2RDF technologies shouldn't know *anything*
about what's outside the database. This work should be done with other
technologies, eg. rule engines.

More generally, I don't think it's a good idea to put such things in
RDB2RDF when we already have so many specs and technologies addressing
these issues.

My two cents,


> -David

Received on Wednesday, 8 June 2011 15:10:40 UTC