XSD.string is a curious special case about rdf

[artob]

Not sure what you consider the problem to be in RDF::Literal handling languages? Nothing has changed on that in RDF.rb 0.2.1.

If it's just a case of misremembering the RDF specs, the short of it is that languages are attributes of literals (not of strings) just the same way as are datatypes, and that these attributes are mutually exclusive. It is not permitted to have a literal that has both a datatype and a language tag, which means that there can be no xsd:string literal that has a language tag. It's one or the other. This also means that though one could casually (though not according to the spec) consider plain literals to have an implicit xsd:string datatype, this is certainly not the case for language-tagged literals.

In any case, I highly doubt I'm going to want to subclass Ruby's String class in RDF.rb, but nothing is stopping you from doing any of this in Spira. Experimental/disruptive changes should be done as a set of monkey-patches in other gems, after which it can be discussed what belongs in RDF.rb and what doesn't. Gregg is setting a good example to follow with regards to this strategy.

That said, we could add an RDF::Literal::String class for xsd:strings specifically, it's just another datatype to support and there are a number of them we haven't got as yet. Just as long as it's understood it has nothing to do with plain or language-tagged literals.

[gkellogg]

Note that Literal("s", :datatype => RDF::XSD.string) is different that Literal("s"). Also, it is illegal for a literal to have both a datatype and a language, other than for XMLLiteral.

One way I dealt with this is to treat :language congruent to :datatype.

Also, you might simply do an object method to create a :literal on string, rather than create a new subclass.

[bhuga]

Well shoot, I learned something today. All of this time I thought that the language was an attribute on XSD.string, with the syntax omitting the explicit datatype just some sugar, but a careful trip to the RDF concepts document has sorted me out. Live and learn.

In this case, RDF::Literal::String for XSD strings won't help me, because what I'm trying to do is come up with a way for language-tagged literals to make the round trip to a Ruby object and back, and I thus need an object which keeps track of a language attribute, but otherwise behaves just like a string--not an RDF::Literal, but something that RDF::Literal returns for #object on language-tagged literals.

Anyways, I see the messiness of including that in RDF.rb proper, and if you're not interested in that tradeoff, well, that's why I asked before I implemented it there. I'll cram it into Spira and we'll see how it goes.