Re: Pre-CR review - Section 4 comments

On 23 Jan 2012, at 20:04, David McNeil wrote:
>> [[
>> The following conditions give rise to data errors: […] A term map that has a validatable RDF datatype as its specified datatype produces an ill-typed literal.
>> ]]
>> 
>>> I don't remember seeing a place in the spec that captures our intent regarding the intersection of SQL and XSD data types.
>> 
>> That intent was about which SQL types to handle in the natural mapping from SQL values to RDF literals. This is captured in Section 10.2 – it only handles the intersection of Core SQL 2008 and XSD.
> 
> Hmm... I thought the intent was to cover the data values.
> 
> The scenario I have in mind is:
> 
> TS is a SQL data type that maps to TX, an XSD data type.
> 
> The domain of valid values for TS is DS. The domain of valid values for TX is DX.
> 
> My understanding was that our intent for R2RML was to only define what happens for SQL data values of type TS that fall in the intersection of DS and DX.
> 
> Meaning that if a SQL data value of type TS had no valid representation in DX then the behavior of the mapping was undefined. In this case the R2RML processor could produce a data value which was outside of DX.

For all TS covered in the natural datatype mapping, TX has been chosen so that all possible values of TS are legal values of TX. So, the DX is always equal to, or a superset of, DS.

Perhaps we should make clearer that the quote above is unrelated to the natural mapping, but can only apply when the natural mapping is overridden with rr:datatype. Would it help if the quote was rephrased?

[[
The following conditions give rise to data errors: […] A _term map_ whose _natural RDF datatype_ is overridden with a _specified datatype_ and produces an _ill-typed literal_ (see _datatype-override RDF literal_).
]]

Best,
Richard



> 
> -David

Received on Tuesday, 24 January 2012 11:29:37 UTC