Re: abstract syntax for R2RML mapping

On Tue, 2011-06-14 at 11:10 -0400, Eric Prud'hommeaux wrote:
> I hacked up a "grammar" for an R2RML mapping (scare quotes 'cause it's
> an abstract syntax -- there are no lexical tokens to e.g. tell if the
> first Name is a TableOwner or a LogicalTable). 

Well, looking at what you did, it's more an abstract model than an
abstract syntax. Indeed, it is useful to just consider RDF/Turtle as
just one serialization format.

Anyway, ericP++


> Anyone can hack on this. Note that there's no history, so if you're
> going to do something radical, it's polite to either save under a new
> name or copy the old bnf someplace safe like an email.
> I tried to be as faitful to the spec text as possible, which should hilight some possible questions, e.g.
> [2] TriplesMap   ::= (TableOwner)? LogicalTable SubjectMap (PredicateObjectMap)* (RefPredicateObjectMap)*
> [4] LogicalTable ::= TableName
>                    | SqlQuery
> Shouldn't the TableOwner be optional only when the LogicalTable is a
> TableName, i.e.
> [2] TriplesMap   ::= LogicalTable SubjectMap (PredicateObjectMap)* (RefPredicateObjectMap)*
> [4] LogicalTable ::= TableOwner? TableName
>                    | SqlQuery
> ?
> Anyways, I think working out an abstract syntax will help many people
> towards understand R2RML.

Received on Tuesday, 14 June 2011 15:23:10 UTC