Add Limit test to document how the "limit" parameter is supposed to work

[gaurav]

I noticed some odd behavior with the way in which the limit parameter works, which can be best summarized by listing the results of running the same query against the SPARQL endpoint using the new test I've written in this PR.

Retrieved 14 results when limit=0
Retrieved 1 results when limit=1
Retrieved 2 results when limit=2
Retrieved 2 results when limit=3
Retrieved 2 results when limit=4
Retrieved 3 results when limit=5
Retrieved 6 results when limit=10
Retrieved 13 results when limit=20
Retrieved 14 results when limit=30
Retrieved 14 results when limit=40
Retrieved 14 results when limit=50
Retrieved 14 results when limit=60
Retrieved 14 results when limit=70
Retrieved 14 results when limit=80
Retrieved 14 results when limit=90
Retrieved 14 results when limit=100

This PR updates the SPARQL query generation code (in consultation with Jim) to correct this bug so we return the correct number of results. It retains the previously used code since there is functionality in there that we would probably like reuse in the future.

This PR previously introduced two new bugs (#517, #518) but then closes them in d4ccd3a.

[gaurav]

It looks like the problem is happening here:
https://github.com/NCATS-Tangerine/cam-kp-api/blob/46cf7fb272b705c23e1ecc5fd70e923cfb3e8bb7/src/main/scala/org/renci/cam/QueryService.scala#L69-L70

findInitialQuerySolutions() seems to apply the limit, returning the number of triples that we expect (up to 25 in the example included in this PR), but then extractCoreTriples() removes n0_type and n1_type from the initial query solutions, returning only the n0, e0 and n1 entries. This reduces the number of distinct triples to the number we get in the output, e.g. Retrieved 6 results when limit=10.

Interestingly, solutionTriples is only used for getting provenance information -- as far as I can tell, the rest of the code uses the initial query solutions only. So maybe this isn't where the limit is being applied? Hmm.

All triples returned when limit=10

Initial query solutions:

n1_type	e0	n1	n0_type	n0
http://purl.obolibrary.org/obo/GO_0006094	http://purl.obolibrary.org/obo/RO_0000056	http://model.geneontology.org/R-HSA-70263/R-HSA-70263	http://purl.obolibrary.org/obo/CHEBI_15361	http://model.geneontology.org/R-ALL-113557_R-HSA-70501
http://purl.obolibrary.org/obo/GO_0046034	http://purl.obolibrary.org/obo/RO_0000056	http://model.geneontology.org/R-HSA-70263/R-HSA-70263	http://purl.obolibrary.org/obo/CHEBI_15361	http://model.geneontology.org/R-ALL-113557_R-HSA-70501
http://purl.obolibrary.org/obo/GO_0046034	http://purl.obolibrary.org/obo/RO_0000056	http://model.geneontology.org/R-HSA-73621/R-HSA-73621	http://purl.obolibrary.org/obo/CHEBI_15361	http://model.geneontology.org/R-ALL-113557_R-HSA-909776
http://purl.obolibrary.org/obo/GO_0046034	http://purl.obolibrary.org/obo/RO_0000056	http://model.geneontology.org/R-HSA-73621/R-HSA-73621	http://purl.obolibrary.org/obo/CHEBI_15361	http://model.geneontology.org/R-ALL-113557_R-HSA-909780
http://purl.obolibrary.org/obo/GO_0019674	http://purl.obolibrary.org/obo/RO_0000056	http://model.geneontology.org/R-HSA-70263/R-HSA-70263	http://purl.obolibrary.org/obo/CHEBI_15361	http://model.geneontology.org/R-ALL-113557_R-HSA-70501
http://purl.obolibrary.org/obo/GO_0019674	http://purl.obolibrary.org/obo/RO_0000056	http://model.geneontology.org/R-HSA-73621/R-HSA-73621	http://purl.obolibrary.org/obo/CHEBI_15361	http://model.geneontology.org/R-ALL-113557_R-HSA-909776
http://purl.obolibrary.org/obo/GO_0019674	http://purl.obolibrary.org/obo/RO_0000056	http://model.geneontology.org/R-HSA-73621/R-HSA-73621	http://purl.obolibrary.org/obo/CHEBI_15361	http://model.geneontology.org/R-ALL-113557_R-HSA-909780
http://purl.obolibrary.org/obo/GO_0016223	http://purl.obolibrary.org/obo/RO_0000056	http://model.geneontology.org/R-HSA-909776	http://purl.obolibrary.org/obo/CHEBI_15361	http://model.geneontology.org/R-ALL-113557_R-HSA-909776
http://purl.obolibrary.org/obo/GO_0016223	http://purl.obolibrary.org/obo/RO_0002328	http://model.geneontology.org/R-HSA-909776	http://purl.obolibrary.org/obo/CHEBI_15361	http://model.geneontology.org/R-ALL-113557_R-HSA-909776
http://purl.obolibrary.org/obo/GO_0047305	http://purl.obolibrary.org/obo/RO_0000056	http://model.geneontology.org/R-HSA-909780	http://purl.obolibrary.org/obo/CHEBI_15361	http://model.geneontology.org/R-ALL-113557_R-HSA-909780

Solution triples is a HashSet() consisting of:

• Triple(IRI(http://model.geneontology.org/R-ALL-113557_R-HSA-909780),IRI(http://purl.obolibrary.org/obo/RO_0000056),IRI(http://model.geneontology.org/R-HSA-909780))
• Triple(IRI(http://model.geneontology.org/R-ALL-113557_R-HSA-70501),IRI(http://purl.obolibrary.org/obo/RO_0000056),IRI(http://model.geneontology.org/R-HSA-70263/R-HSA-70263)),
• Triple(IRI(http://model.geneontology.org/R-ALL-113557_R-HSA-909780),IRI(http://purl.obolibrary.org/obo/RO_0000056),IRI(http://model.geneontology.org/R-HSA-73621/R-HSA-73621))
• Triple(IRI(http://model.geneontology.org/R-ALL-113557_R-HSA-909776),IRI(http://purl.obolibrary.org/obo/RO_0002328),IRI(http://model.geneontology.org/R-HSA-909776))
• Triple(IRI(http://model.geneontology.org/R-ALL-113557_R-HSA-909776),IRI(http://purl.obolibrary.org/obo/RO_0000056),IRI(http://model.geneontology.org/R-HSA-909776))
• Triple(IRI(http://model.geneontology.org/R-ALL-113557_R-HSA-909776),IRI(http://purl.obolibrary.org/obo/RO_0000056),IRI(http://model.geneontology.org/R-HSA-73621/R-HSA-73621)))
• Triple(IRI(http://model.geneontology.org/R-ALL-113557_R-HSA-909780),IRI(http://purl.obolibrary.org/obo/RO_0000056),IRI(http://model.geneontology.org/R-HSA-909780))
• Triple(IRI(http://model.geneontology.org/R-ALL-113557_R-HSA-70501),IRI(http://purl.obolibrary.org/obo/RO_0000056),IRI(http://model.geneontology.org/R-HSA-70263/R-HSA-70263)),
• Triple(IRI(http://model.geneontology.org/R-ALL-113557_R-HSA-909780),IRI(http://purl.obolibrary.org/obo/RO_0000056),IRI(http://model.geneontology.org/R-HSA-73621/R-HSA-73621))
• Triple(IRI(http://model.geneontology.org/R-ALL-113557_R-HSA-909776),IRI(http://purl.obolibrary.org/obo/RO_0002328),IRI(http://model.geneontology.org/R-HSA-909776))
• Triple(IRI(http://model.geneontology.org/R-ALL-113557_R-HSA-909776),IRI(http://purl.obolibrary.org/obo/RO_0000056),IRI(http://model.geneontology.org/R-HSA-909776))
• Triple(IRI(http://model.geneontology.org/R-ALL-113557_R-HSA-909776),IRI(http://purl.obolibrary.org/obo/RO_0000056),IRI(http://model.geneontology.org/R-HSA-73621/R-HSA-73621)))

[gaurav]

Jim is concerned that this code might also misbehave if:

s	p	o	stype	ptype
1	p1	2	A	B
1	p1	2	C	B
3	p1	4	A	B

Note that in this case there are two different individuals that indicate the same relationship (A p1 B). They could be from different sources.

The way to find this would be a unit test that provides tests the difference pieces of the run() workflow, e.g. getTRAPIEdgeBindingsMany().

[balhoff]

Looks, good, just a few possible changes.