Re: Proposal for the datatype mapping

On Tue, Nov 22, 2011 at 8:16 AM, Richard Cyganiak <richard@cyganiak.de>wrote:

> 4. When generating IRIs in R2RML, the output SHOULD be the canonical
> natural RDF literal, although non-canonical forms are still conforming.
>

Yesterday we talked about how this "SHOULD be the canonical natural RDF
literal" guidance would also be applied to plain literal and blank node
identifiers. It occurred to me that another case where this should apply is
when values are converted to xsd:strings. For example, a SQL integer being
converted to an xsd:string in the RDF output.

Looking at it from the other direction if the mapping author wants to
produce the canonical RDF literal form of a SQL integer as an xsd:string or
a plain literal without a language tag... how do they do that? The closest
way I can see to do it is to add a language tag to the mapping so that the
number SHOULD come out as a canonical RDF string with the language tag. I
don't think converting it in SQL will help because there is not a SQL
function to convert a value to a canonical RDF form.

So it seems to me that we should add words to the spec for the following:
1) when a datatype-override is used to convert a non-string SQL value to an
xsd:strings value then the output value SHOULD be the canonical natural RDF
literal
2) allow users to specify that a SQL value should be mapped to a plain
literal without requiring a language tag

-David