feat: add eval test cases for empty graph and isImpliedBy

[jeswr]

Test cases addressing:

• Is :a :b {} the same triple as :a :b true #185
• Evaluation test cases for <= #157

There may be a problem with some of these test cases as some:

• Contain literals in the subject position
• Contain variables

Neither of which is actually permitted in the n-quads spec. A solution for those test cases could be to have the same content for the result but contained in another n3 document. What do you think @william-vw ?

[william-vw]

@jeswr

A solution for those test cases could be to have the same content for the result but contained in another n3 document.

Just to be sure, you mean keeping the results in an n3 results file instead of n-triples / n-quads?

[jeswr]

Just to be sure, you mean keeping the results in an n3 results file instead of n-triples / n-quads?

Correct

[william-vw]

@jeswr yes I think that makes a lot of sense!

[jeswr]

@jeswr yes I think that makes a lot of sense!

So it turns out that doesn't work on files like this because N3 doesn't fully support n-quads syntax, in particular having a graph term like this one.

Do you think it is worth opening up discussion as to whether this syntax should be supported in the spec @william-vw @josd .

[william-vw]

@jeswr sorry I'm missing something - why do you want to use the n-quads syntax in these tests (as you say, it is not supported by the n3 syntax)?

[jeswr]

Generally the results of eval tests are given in an n-quads format (in particular see the rdf-tests repo) as it is straightforward to parse and as such is a good syntax in which to unambiguously provide the expected result of parsing.

In terms of what I'm proposing, it is that in N3 the following:

@prefix : <http://example.org/> .
"a" :b { :c :d :e } .

Be considered equivalent to:

@prefix : <http://example.org/> .
"a" :b _:b0 .
:c :d :e _:b0 .

Parsers over in the RDF/JS world like N3.js already interpret the first statement as the following RDF/JS quads:

const result = [
quad(
  literal("a"),
  namedNode("http://example.org/b"),
  blankNode("b0"),
),
quad(
  namedNode("http://example.org/c"),
  namedNode("http://example.org/d"),
  namedNode("http://example.org/e"),
  blankNode("b0"),
)]

[josd]

In terms of what I'm proposing, it is that in N3 the following:

@prefix : <http://example.org/> .
"a" :b { :c :d :e } .

Be considered equivalent to:

@prefix : <http://example.org/> .
"a" :b _:b0 .
:c :d :e _:b0 .

They are different, the former uses a graph term, the latter uses a graph by reference.
I have added initial support for n-quads in eye and this example round trips fine:

$ cat nq2.n3
@prefix : <http://example.org/> .

"a" :b _:b0 .
:c :d :e _:b0 .

$ eye --quiet --nope nq2.n3 --pass
@prefix : <http://example.org/>.

"a" :b _:e_b0_1.
:c :d :e _:e_b0_1.

One can of course compose graph terms with the following rule in

$ cat nq3.n3
@prefix log: <http://www.w3.org/2000/10/swap/log#>.
@prefix graph: <http://www.w3.org/2000/10/swap/graph#>.
@prefix : <http://example.org/> .

"a" :b _:b0 .
:c :d :e _:b0 .
:f :g :h _:b0 .

{   ?A ?B ?C.
    ({?D ?E ?F} {?D ?E ?F ?C} ?L) log:collectAllIn ?SCOPE.
    ?G graph:list ?L.
} => {
    ?A ?B ?G.
}.

$ eye --quiet --nope nq3.n3 --pass
@prefix log: <http://www.w3.org/2000/10/swap/log#>.
@prefix graph: <http://www.w3.org/2000/10/swap/graph#>.
@prefix : <http://example.org/>.

"a" :b _:e_b0_1.
"a" :b {
    :c :d :e.
    :f :g :h.
}.
:c :d :e _:e_b0_1.
:f :g :h _:e_b0_1.
:c :d :e {
    :c :d :e.
    :f :g :h.
}.
:f :g :h {
    :c :d :e.
    :f :g :h.
}.

[william-vw]

Generally the results of eval tests are given in an n-quads format (in particular see the rdf-tests repo) as it is straightforward to parse and as such is a good syntax in which to unambiguously provide the expected result of parsing.

Why would n-quads be better to provide expected results - what would the graph label indicate? I'm not finding n-quads result files in the folder you reference (aside from the rdf-n-quads folder, but those are not result files). Not trying to be difficult here, I am just wondering what the added value would be for an individual results file.

It's a much bigger discussion, for sure, but I feel that graph terms can be used to group triples belonging to the same graph (these could be made "referentially transparent" as well).

[jeswr]

I've updated the files @william-vw .

I'll also respond to #202 (comment) in due course when I have time to write a full response.

[william-vw]

looks good (aside from minor comment)

[william-vw]

unsure what is meant by complex use case?

[jeswr]

Not sure either - resolved in 59f2ae7.

I don't have permissions to merge on this repo so will leave that to you.