Error in fromRdf/0027-out.jsonld

[marcelotto]

The new fromRdf#t0027 test expects its result to be wrapped in a @graph container. Maybe I'm missing something, but that seems wrong to me.

Also: Why is the @type specified on the values of the boolean-native object, when the corresponding values in the input are valid boolean values? Same on number-native.

[anatoly-scherbakov]

Hi @marcelotto, thanks for looking into this!

1. I reviewed this against 8.4 Serialize RDF as JSON-LD Algorithm
   , I think you are right. Will check this out.

2. 8.6 Data Round Tripping seems to override conversion of RDF literals into value objects indeed.

I will prepare a PR, we will maybe discuss this on the next meeting.

---

I was the one who wrote this test and I do remember that I was considering the two alternatives: value object vs a native JSON value and somehow I chose the former. I do not recall the details anymore.

[anatoly-scherbakov]

For reference, I think this is the PR #625

[w3cbot]

This was discussed during the #json-ld meeting on 09 April 2025.

View the transcript

w3c/json-ld-api#648

<gb> Issue 648 Error in fromRdf/0027-out.jsonld (by marcelotto)

anatoly-scherbakov: when I was converting literals into JSON-LD
… when they used native types, I was converting them to Value Objects
… but that's not what the spec is going for
… the spec says to use native JSON types
… and I made a PR to address that

w3c/json-ld-api#650

<gb> Pull Request 650 #648 `0027-out.jsonld`: `@graph` and value objects (by anatoly-scherbakov)

gkellogg: the first sentence was about `@graph`
… it was always used in 1.0
… but generally eliminated in 1.1
… except for certain use cases
… if JSON-LD is a single object, then you would not expect to see an `@graph` containing it
… so I think there are two things in here
… the specifics of `@graph` and what types of JSON values are emitted vs. expanded values
… you basically want more review, correct?

anatoly-scherbakov: yes. exactly
… some re-review would also be appreciated

gkellogg: so the linked issue is 648

anatoly-scherbakov: and the PR is 650

gkellogg: thanks
… we do also have some things marked as propose close

---

[anatoly-scherbakov]

I double checked the (2) above. Check out RDF to Object conversion → Algorithm

1. If value is an IRI or a blank node identifier, return …
2. Otherwise value is an RDF literal:
   2.1) Initialize a new empty map result.
   …
   2.11) Return result.

There are no early returns between 2.1 and 2.11 which I can see. This indeed will be a value object. Test seems correct in that regard. Sorry for the confusion above.