Re: Datatype handling added to R2RML spec (ISSUE-29, ISSUE-48, ISSUE-51)

On 26/08/2011 0:56, Richard Cyganiak wrote:
> There's a new section in R2RML that discusses the handling of datatypes:
> http://www.w3.org/2001/sw/rdb2rdf/r2rml/#datatype-conversions
>
> The definitions and tables in that section are referenced from various places throughout the spec, in particular:
> http://www.w3.org/2001/sw/rdb2rdf/r2rml/#typed-literals
> http://www.w3.org/2001/sw/rdb2rdf/r2rml/#generated-rdf-term
>
> Some noteworthy features:
>
> - it is now an error to have a logical column that cannot be CAST to string
> - unknown/user-defined types are CAST to string and mapped to plain literal
> - rr:datatype is implicit if unspecified, that is, date column will be xsd:date even without explicit rr:datatype
> - explicit rr:datatype that doesn't match the implicit type results in typed literal where the lexical form is simply the result of CAST to string
> - implementations MUST support literal well-formedness checks for the 8 supported XSD types
> - implementations MAY support literal well-formedness checks for additional types
> - ill-typed literal is a data error
> - I ignored INTERVAL because it's so freaking complicated
> - binary types get mapped to xsd:base64Binary
>
> Please review and comment!
>
>
> Process: I've set ISSUE-29, ISSUE-48 and ISSUE-51 to Pending Review and propose:
>
> [[
> PROPOSAL: Close ISSUE-29; all places that interpret SQL values as string now use "conversion to string"
> ]]
+1
> [[
> PROPOSAL: Close ISSUE-48 as it is addressed in http://www.w3.org/2001/sw/rdb2rdf/r2rml/#datatype-conversions
> ]]
+1
>
> [[
> PROPOSAL: Close ISSUE-51; the spec now demands that “supported RDF types” MUST raise data error, and other types MAY raise data error
> ]]
+1
>
Boris

Received on Monday, 29 August 2011 14:42:01 UTC