Check EYE RDF 1.2 is correct

[jeswr]

#1853 upgraded eye-js to use N3.js v2 - which implements RDF 1.2.

Prior to that PR the test was failing in the upgrade with the error attached below. What is unclear to me is whether the change I have made here is correct RDF 1.2 semantics; or whether there is a bug in the output of eye when using deductive_closure.

EYE is called with the following: main(['--nope', '--quiet', 'data_0.n3s', '--pass'])
The input to EYE is:

_:n3-6 <http://www.w3.org/1999/02/22-rdf-syntax-ns#reifies> <<(<http://example.org/socrates#Socrates> a <http://example.org/socrates#Human>)>> .

_:n3-6 <http://example.org/socrates#is> true .

<http://example.org/socrates#Human> <http://www.w3.org/2000/01/rdf-schema#subClassOf> <http://example.org/socrates#Mortal> .

{
  _:n3-8 <http://www.w3.org/1999/02/22-rdf-syntax-ns#reifies> <<(?S ?P ?O)>> . _:n3-8 <http://example.org/socrates#is> true . } => {?S ?P ?O . } . {?S a ?A . ?A <http://www.w3.org/2000/01/rdf-schema#subClassOf> ?B . } => {?S a ?B . } . 

Test Error:

● testing n3reasoner › should execute the n3reasoner [quad input quad output] [output: deductive_closure]

    expected two graphs to be isomorphic.

      Expected:
    [
      {"subject":"http://example.org/socrates#Socrates","predicate":"http://www.w3.org/1999/02/22-rdf-syntax-ns#type","object":"http://example.org/socrates#Human","graph":""},
      {"subject":"http://example.org/socrates#Socrates","predicate":"http://www.w3.org/1999/02/22-rdf-syntax-ns#type","object":"http://example.org/socrates#Mortal","graph":""},
      {"subject":"<<http://example.org/socrates#Socrates http://www.w3.org/1999/02/22-rdf-syntax-ns#type http://example.org/socrates#Human>>","predicate":"http://example.org/socrates#is","object":"\"true\"^^http://www.w3.org/2001/XMLSchema#boolean","graph":""},
      {"subject":"http://example.org/socrates#Human","predicate":"http://www.w3.org/2000/01/rdf-schema#subClassOf","object":"http://example.org/socrates#Mortal","graph":""}
    ]

      Actual:
    [
      {"subject":"http://example.org/socrates#Socrates","predicate":"http://www.w3.org/1999/02/22-rdf-syntax-ns#type","object":"http://example.org/socrates#Human","graph":""},
      {"subject":"http://example.org/socrates#Socrates","predicate":"http://www.w3.org/1999/02/22-rdf-syntax-ns#type","object":"http://example.org/socrates#Mortal","graph":""},
      {"subject":"_:b16_e_n3_MINUS_6_1","predicate":"http://www.w3.org/1999/02/22-rdf-syntax-ns#reifies","object":"<<http://example.org/socrates#Socrates http://www.w3.org/1999/02/22-rdf-syntax-ns#type http://example.org/socrates#Human>>","graph":""},
      {"subject":"_:b16_e_n3_MINUS_6_1","predicate":"http://example.org/socrates#is","object":"\"true\"^^http://www.w3.org/2001/XMLSchema#boolean","graph":""},
      {"subject":"http://example.org/socrates#Human","predicate":"http://www.w3.org/2000/01/rdf-schema#subClassOf","object":"http://example.org/socrates#Mortal","graph":""}
    ]

    Missing Quads (that don't contain Blank Nodes):
    [
      {"subject":"<<http://example.org/socrates#Socrates http://www.w3.org/1999/02/22-rdf-syntax-ns#type http://example.org/socrates#Human>>","predicate":"http://example.org/socrates#is","object":"\"true\"^^http://www.w3.org/2001/XMLSchema#boolean","graph":""}
    ]

    Additional Quads (that don't contain Blank Nodes):
    [

    ]

    Missing Blank Node Patterns:


    Additional Blank Node Patterns:
    _:b16_e_n3_MINUS_6_1 : [
      {"subject":"_:b16_e_n3_MINUS_6_1","predicate":"http://www.w3.org/1999/02/22-rdf-syntax-ns#reifies","object":"<<http://example.org/socrates#Socrates http://www.w3.org/1999/02/22-rdf-syntax-ns#type http://example.org/socrates#Human>>","graph":""},
      {"subject":"_:b16_e_n3_MINUS_6_1","predicate":"http://example.org/socrates#is","object":"\"true\"^^http://www.w3.org/2001/XMLSchema#boolean","graph":""}

The question is whether the result we are now testing for is correct, or whether EYE does not correctly implement RDF 1.2

cc @josd

[josd]

The parsing of RDF 1.1 and RDF 1.2 in EYE is experimental and the only support will be for https://w3c.github.io/N3/spec/
Same for https://github.com/eyereasoner/eyeling and so we would really encourage to further improve https://github.com/eyereasoner/eyeling/blob/main/tools/n3gen.js

Usage:
  n3gen <file.ttl|file.trig|file.srl>

Converts RDF 1.2 Turtle/TriG and SHACL 1.2 Rules to N3.

Examples:
  n3gen file.ttl > file.n3
  n3gen file.trig > file.n3
  n3gen file.srl > file.n3

concretely

cd /tmp
git clone https://github.com/eyereasoner/eyeling
cd eyeling/
npm run test:n3gen
Cloning into 'eyeling'...
remote: Enumerating objects: 2955, done.
remote: Counting objects: 100% (386/386), done.
remote: Compressing objects: 100% (272/272), done.
remote: Total 2955 (delta 192), reused 251 (delta 106), pack-reused 2569 (from 2)
Receiving objects: 100% (2955/2955), 2.09 MiB | 7.79 MiB/s, done.
Resolving deltas: 100% (1786/1786), done.

> eyeling@1.10.6 test:n3gen
> node test/n3gen.test.js

== Running n3 conversions for 9 inputs (git worktree mode)
node v24.12.0
OK 01 annotation.ttl -> annotation.n3 (34 ms)
OK 02 bind.srl -> bind.n3 (37 ms)
OK 03 collection.ttl -> collection.n3 (36 ms)
OK 04 context-association.trig -> context-association.n3 (45 ms)
OK 05 family.srl -> family.n3 (33 ms)
OK 06 filter.srl -> filter.n3 (38 ms)
OK 07 reifies.ttl -> reifies.n3 (36 ms)
OK 08 snaf.srl -> snaf.n3 (34 ms)
OK 09 triple-term.ttl -> triple-term.n3 (35 ms)

== Total elapsed: 428 ms (0.43 s)
OK All n3 conversions passed (9/9)

[jeswr]

Agent-assisted triage (for @jeswr). Verified the post-#1853 expectation is correct RDF 1.2 semantics; EYE's output is not buggy here:

1. In RDF 1.2 Turtle, annotation/quoted-triple syntax desugars to a fresh reifier: << :s :p :o >> :is true is sugar for _:r rdf:reifies <<( :s :p :o )>> . _:r :is true . Triple terms may only appear as the object of rdf:reifies; a triple term in subject position — the old "Expected" output in the report — is RDF-star CG semantics that was dropped in RDF 1.2.
2. deductive_closure (--pass) must preserve input triples, so EYE echoing _:r rdf:reifies <<(...)>> and _:r :is true (with the bnode relabelled, fine under isomorphism) is the correct result; only the pre-1.2 expectation was wrong.
3. Cross-check: N3.js v2.1.1 itself parses << :s :p :o >> :is true to a blank-node reifier (not a quoted-triple subject), so parser and reasoner agree on the 1.2 model.

Per @josd, RDF 1.1/1.2 parsing in EYE remains experimental with official support targeting the N3 spec (n3gen in eyeling is the recommended conversion path). Suggest closing: the test as changed in #1853 is correct RDF 1.2; optionally add a README note that eye-js RDF 1.2 handling rides on experimental EYE support.

Per @jeswr: closing, and a follow-up PR will document the level of RDF 1.2 support in EYE (experimental, per josd) in the README, keeping the existing dataStar tests as the guard.