Fix use native types flag with non-native values

[niklasl]

This addresses #670.

• Add test case
• Implement fix in algorithm

---

Preview | Diff

[niklasl]

Unless there is a better idea for refactoring the spec algorithm in this place (see #670), I'll probably implement the fix similarly to how I'll fix it in TRLD: niklasl/trld@a66feee (i.e. add a done converting flag).

[niklasl]

I made a possibly cleaner approach in niklasl/trld@5fe0981 (explained in that commit message). Any feedback welcome; I'll probably proceed with fixing the spec algorithm accordingly in the coming week.

[niklasl]

I've now implemented what I see as an acceptable fix, both in TRLD: niklasl/trld#5 (which now passes all fromRdf tests including the one added in this PR); and as a proposed spec change in this PR.

It is a slight refactoring of the RDF to Object Conversion algorithm, to address this issue whilst also keeping or seeking to improve the clarify of the algorithm.

[w3cbot]

This was discussed during the #json-ld meeting on 24 June 2026.

View the transcript

Pull Request 671 Fix use native types flag with non-native values (by niklasl) [needs discussion]

niklasl: this issue has been around forevever; I think that implementations do not follow the spec (they would encounter issues),
… but instead implement the suggested change in this PR
… note that the HTML diff is easier to follow than the PR diff

bigbluehat: I agree on the separate PR for the change on the "quote" thing

<Zakim> pchampin, you wanted to react to anatoly-scherbakov

pchampin: the CI issue is due to an problem I created with common files,
… which w3c/json-ld-api#693 (not yet merged) is supposed to fix

<gb> Pull Request 693 fix bug in Rakefile after removing common/ (by pchampin)

---

[niklasl]

Oh my, the github bug that changes newline form struck again. (It seems to occur, sometimes, when a suggestion is applied through the web interface; possibly if there is some mix of line ending forms already in the doc.) I'll take a look, and importantly rework the last commit (and force-push that).

[niklasl]

OK; the change this PR proposes looks clean again now (the culprit was \r\n-line-endings from b028727; itself a change suggestion).

I recommend squash-and-merge for when we merge this, to have a clean change commit for this in the main branch.

[TallTed]

the culprit was \r\n-line-endings from b028727; itself a change suggestion

That change suggestion came from me.

It appears that GitHub has badly tweaked something in how it handles such change suggestions. I certainly didn't intend to change the line endings, and I can find nothing about that commit that indicates the line endings were being changed.

If anyone figures out a switch that I can flip to avoid such issues going forward, please let me know!