tests related to importing of blank nodes

[hartig]

This PR adds several tests related to the spec text that we are still working on in PR #2. That is, these tests are about blank node identifies, reading them from one or multiple Turtle files, from SPARQL queries, or from both Turtle files and SPARQL queries.

@kasei please review. I think most of them are quite straightforward. The only one that is probably controversial is :bnodes-sparql-10 which uses the following query.

ASK {
	BIND( "[_:b, 42]"^^cdt:List AS ?list1 )
	BIND( "[_:b, 43]" AS ?str )
	BIND( STRDT(?str, cdt:List) AS ?list2 )

	BIND( cdt:get(?list1,1) AS ?e1 )
	BIND( cdt:get(?list2,1) AS ?e2 )

	FILTER( isBLANK(?e1) )
	FILTER( isBLANK(?e2) )
	FILTER( ?e1 = ?e2 )
}

I am unsure myself about this one. It just came to my mind that something like this is possible and there may be expectations for this to work. Yet, strictly speaking, this test should not pass; not according to the current spec text that we have in PR #2 because that spec text focuses only on cdt:List and cdt:Map literals that are given directly with their lexical forms. So, perhaps, the correct version of this test should say FILTER( ?e1 != ?e2 ) instead of FILTER( ?e1 = ?e2 ) !?

[hartig]

@kasei I added more tests (see commit 0c47b81). These new tests cover cases in which the CDT literals are nested such as the following two examples.

ex:s ex:p "[_:b, 42, [_:b] ]"^^cdt:List .

ex:s ex:p "[_:b, 42, '[_:b]'^^<http://w3id.org/awslabs/neptune/SPARQL-CDTs/List> ]"^^cdt:List .

[kasei]

I will review them soon.

The only one that is probably controversial is :bnodes-sparql-10 which uses the following query.

Interesting test. I'm not sure STRDT (or casting/constructor functions) should do bnode rewriting. Maybe there's a reason to think it shouldn't allow any lexical form with a bnode in it? Otherwise you let in the accidentail skolemization again. I'm sure I'll have more to say after going through the tests.

For what it's worth, my implementation currently fails 9 out of the 73 tests.

[hartig]

I have worked on my implementation the last two days and it passes all these tests now, except for the STRDT test.

I didn't necessarily mean that we need to include this test. It is just something that came to my mind and I wanted to bring it up. We may also include the inverse of this test.

[kasei]

OK. I've now gone over all of the tests. My code now also passes all tests except bnodes-sparql-10, which I think we need to discuss more.

[hartig]

Thanks for checking the tests! I have removed bnodes-sparql-10 for the moment, to merge this PR. After that I will create a separate PR with bnodes-sparql-10.