Group Triples by Common Subject

[tmciver]

This work addresses the bug in #98.

[robstewart57]

@tmciver Thanks for this bug report and PR, and your efforts in improving the rdf4h library.

Looks good, I'll merge this.

It'd be good if, in addition to the "triples with the same subject should be grouped" test case (and other unit tests), we can also use property based tests using QuickCheck for far greater coverage of possible inputs.

E.g. for the bug that you report, the specification might be:

In the Turtle serialisation of a graph, every subject node appears in a subject position exactly once.

or

In the Turtle serialisation of a graph, a node can appear in a subject position at most once.

This can test for many more possible interleavings of triples with common subjects, in addition to:

<:something> <dc:title> \"Some title\" .
<:another> <rdf:type> <schema:Document> .
<:something> <rdf:type> <schema:Document> .

We probably can't use the Turtle parser to test this, since it desugars < subject , [ (predicate , object) ] > patterns into < subject , predicate , object > triples.

Hhmm..

[robstewart57]

@tmciver we should probably also have QuickCheck tests to validate properties that we believe should hold when parsers and serialisers are composed. E.g.

forall (g::RDF a) . triplesOf g
                    ==
                    triplesOf (parseString (TurtleParser Nothing Nothing)
                                   (writeRdfString (TurtleSerializer Nothing Nothing) g))

There's probably a few such properties we assume hold that might not. E.g. on the number of triples, the order of triples, ensuring namespaces are/aren't lost, etc.

[tmciver]

@robstewart57 I love the idea of using Quickcheck! That's something I can work on but I'd like to tackle #97 first since the produced Turtle is not valid until that one is resolved.