JRDF can now do the following relational operations on RDF: project, restrict, union, semi-difference, natural join, anti-join, full and left outer join. It does most of the good things like retaining relations all the way through so it can perform the operations without ordering restrictions. However, the query engine doesn't do any optimization yet. It also doesn't have to deal with duplicates so there's no need for DISTINCT and UNION is predictable. The SPARQL grammar is a subset of the current one and it only supports SELECT, join (.), UNION and OPTIONAL. There are also no null values.
There's also a Swing based GUI (that's a bit like Twinkle) that allows the submission of SPARQL queries and renders them in tabular fashion.
I'm thinking that I might add a difference operator (-) and produce an OPTIONAL like query that is not order dependent. I had initially planned to use minimum union (outerunion and tuple subsumption (removing rows with more null values)). However, it does seem that the existing full outerjoin will work with tuple subsumption too - although it does look like more work.
There's also some interesting aspects of the grammar too.
Due to time constraints the level of code quality isn't quite up to what I'm happy with so at the moment there's no release and there's probably still bugs left to squash. For the truely keen the current source code is available in Subversion.