SPARQL: Drop support for $shapesGraph and $currentShape as pre-bound variables

[HolgerKnublauch]

https://www.w3.org/TR/shacl/#sparql-constraints-prebound stated that SPARQL engines MAY support the optional variables $shapesGraph and $currentShape. This is reinforced by test case shapesGraph-001.ttl (which is incorrect because not all implementations need to support these variables).

I suggest we delete these variables altogether for 1.2. Having an optional feature such may cause Failures is not a good idea. Also I don't think this was used much by anyone. The original idea was to have them in the spec so that all Core constraint components can be expressed via SPARQL, but that didn't work anyway.

[TallTed]

which is incorrect because not all implementations need to support these variables

I'm not sure what you mean by "incorrect". Results of testing a MAY should simply be a report of whether that implementation supports that MAY; i.e., it does not impact conformance.

I am concerned about dropping such a MAY from the spec, as it seems that going forward, implementations which did implement the MAY might then be considered non-conformant... though perhaps by virtue of there being no statement, MAY or otherwise, whatever an implementation might do would be conformant....

Having an optional feature such may cause Failures is not a good idea

Perhaps, but how might such an optional feature "cause Failure"? If you're just speaking of the shapesGraph-001.ttl test, again I say, failing this test is just a question of optional feature support; it does not indicate nonconformance.

[HolgerKnublauch]

@TallTed implementations that did support these variables can continue to do so, and their behavior isn't going to change. But I don't believe anyone else supported this ever. Implementations may also have other SPARQL-related native features (e.g. GeoSPARQL support), and we are just silent about them. I believe we can do the same here.

The test framework would need to be extended to allow for such optional features, but I really don't think this marginal feature is worth doing that.

[afs]

See also #183 about extending SHACL to the validation of RDF datasets.

I don't think $shapesGraph and $currentShape help there. At best, they will reappear in some form; at worse, keeping them gets in the way.