Re: [rdf4j-dev] Relocate InverseMagicProperty

[Jeen Broekstra <jeen.broekstra@xxxxxxxxx>]

On 6 Oct. 2016, at 03:31, Jacek Grzebyta <grzebyta.dev@xxxxxxxxx> wrote:

Dear All,

Regards https://github.com/jgrzebyta/rdf4j/tree/issues/%23220-lucene-matches

I suggest to relocate InverseMagicProperty from rdf4j-spin to rdf4j-queryalgebra-evaluation package to keep it together with TupleFunction.

In the above task Mark Hale created an instance of magic property (spin function) which should implement InverseMagicProperty interface. Because of that I added rdf4j-spin dependency into rdf4j-sail-lucene-api. After I found a cyclic reference error:

I guess that makes sense. It’s an API breakage, but then again the SPIN stuff is still branded beta, so I can live with shifting things a bit. Perhaps leave a deprecated empty subclass in the old location. 

More generally I’m thinking that a lot of the TupleFunction implementations have applicability beyond SPIN reasoning. Should we perhaps move all them into the queryalgebra-evaluation package so that we can also (optionally) support their use in “normal”  SPARQL queries (when using an extended evaluation strategy)? Or are there obstacles to that that I haven’t seen? 

Jeen