Enumeration value in JSON-LD without @context

[KalleOlaviNiemitalo]

What's the right way to represent a schema.org enumeration value in JSON-LD without using @context? Does it follow from some specification?

• RDF label:

  {
      "@type": "https://schema.org/CreateAction",
      "https://schema.org/actionStatus": "CompletedActionStatus"
  }

• Value object:

  {
      "@type": "https://schema.org/CreateAction",
      "https://schema.org/actionStatus": {
          "@value": "CompletedActionStatus",
          "@type": "https://schema.org/ActionStatusType"
      }
  }

• @id:

  {
      "@type": "https://schema.org/CreateAction",
      "https://schema.org/actionStatus": {
          "@id": "https://schema.org/CompletedActionStatus"
      }
  }

Analysis

The https://schema.org/actionStatus property (JSON-LD) has a range that includes https://schema.org/ActionStatusType (JSON-LD), which is a subclass of https://schema.org/StatusEnumeration (JSON-LD) and thus also https://schema.org/Enumeration (JSON-LD).

The https://schema.org/CompletedActionStatus enumeration value (JSON-LD) is an instance of https://schema.org/ActionStatusType and has "CompletedActionStatus" as its label.

If I understand correctly, the https://schema.org/CompletedActionStatus node would be a suitable value for the https://schema.org/actionStatus property. The string "CompletedActionStatus" would not be suitable as is, and would have to be coerced. However, it is not clear to me whether this coercion would be done as part of converting from JSON-LD to tuples, or in a subsequent processing step. It is also not clear whether the coercion would be based on the http://www.w3.org/2000/01/rdf-schema#label property of the enumeration members, on the path components of the URLs, or on hardcoded knowledge about the https://schema.org/actionStatus property and the https://schema.org/ActionStatusType enumeration type.

If "@context": "https://schema.org/" were used

The JSON-LD context at https://schema.org/docs/jsonldcontext.jsonld declares "actionStatus" as meaning http://schema.org/actionStatus, but does not specify a @type for coercing the property value:

schemaorg/data/releases/22.0/schemaorgcontext.jsonld

Line 24 in 12e9bbb

"schema": "http://schema.org/",

schemaorg/data/releases/22.0/schemaorgcontext.jsonld

Line 1433 in 12e9bbb

"actionStatus": { "@id": "schema:actionStatus"},

The context also declares "CompletedActionStatus" as meaning http://schema.org/CompletedActionStatus:

schemaorg/data/releases/22.0/schemaorgcontext.jsonld

Line 250 in 12e9bbb

"CompletedActionStatus": {"@id": "schema:CompletedActionStatus"},

If the context specified "@type": "@vocab" for the http://schema.org/actionStatus property, then that would cause the JSON-LD consumer to treat the property value "CompletedActionStatus" as a term and expand it to http://schema.org/CompletedActionStatus. The context doesn't currently do so, though. And of course this cannot help when the JSON-LD document does not use @context.

[WeaverStever]

The schema validator seems to extract "CompletedActionStatus" to http://schema.org/CompletedActionStatus

<script type="application/ld+json">
{
  "@context": "https://schema.org",
  "@type": "ListenAction",
  "agent": {
    "@type": "Person",
    "name": "John"
  },
  "object": {
    "@type": "MusicGroup",
    "name": "Pink!"
  },
  "participant": {
    "@type": "Person",
    "name": "Steve"
  },
  "location": {
    "@type": "Residence",
    "name": "Ann's apartment"
  },
  "instrument": {
    "@type": "Product",
    "name": "iPod"
  },
  "actionStatus": "CompletedActionStatus"
}
</script>

[KalleOlaviNiemitalo]

That validator seems to have some built-in knowledge about the http://schema.org/actionStatus property even when @context is not used. It coerces "CompletedActionStatus" to http://schema.org/CompletedActionStatus from the following input, too:

{
  "@type": "http://schema.org/ListenAction",
  "http://schema.org/agent": {
    "@type": "http://schema.org/Person",
    "name": "John"
  },
  "http://schema.org/object": {
    "@type": "http://schema.org/MusicGroup",
    "name": "Pink!"
  },
  "http://schema.org/participant": {
    "@type": "http://schema.org/Person",
    "name": "Steve"
  },
  "http://schema.org/location": {
    "@type": "http://schema.org/Residence",
    "name": "Ann's apartment"
  },
  "http://schema.org/instrument": {
    "@type": "http://schema.org/Product",
    "name": "iPod"
  },
  "http://schema.org/actionStatus": "CompletedActionStatus"
}

AFAICT, this coercion is not part of the JSON-LD specification. But I'd like to understand whether it is

• standardised at schema.org,
• standardised as part of RDF, or
• ad-hoc implemented by individual schema.org consumers in order to make sense of web pages that they crawl.

[gkellogg]

JSON-LD does not perform any vocabulary processing, so the fact that the range of actionStatus is Action is not used when expanding the JSON-LD, or turning it into RDF. The only way this is done is if the term definition for actionStatus in an included context is defined with "@type": "@id" (or "@type": "@vocab"), which it apparently is not, thus "CompletedActionStatus" must be interpreted as a literal value.

This does not mean that other tools can't use the stated range to infer something different, but the Linter, for one, will not. Generally, schema.org has a high tolerance for accepting strings where an IRI (or object) might be expected to the point where properties are often defined to have a textual range, so the schema.org context is not overly aggressive in using @type in a term definition. Certainly, it would not be hard to infer that the literal value "CompletedActionStatus" was really intended to represent http://schema.org/CompletedActionStatus using the available information.

[WeaverStever]

This look relevant...
https://jazz.net/wiki/bin/view/LinkedData/UseUrisForEnums

(Edited: for explaining enumerations in RDF)

[rrlevering]

For the record, the schema.org validator is essentially doing the non-standardised third option which is part of a "coercion" layer (read "Anything non-standard that is common on the web") We have a copy of the schema definition from schema.org and we normalize "" -> "https://schema.org/" if it's a valid enum value for any enumeration in the range.

[danbri]

Let’s call it normalization or canonicalization, to avoid misunderstandings. I prefer the former as the same data might be normalized differently for different consuming environments. In practice all schema.org consumers need to do quite a lot of this…

…

On Mon, 27 Nov 2023 at 15:22, Ryan Levering ***@***.***> wrote: For the record, the schema.org validator is essentially doing 3) which is part of a "coercion" layer (read "Anything non-standard that is common on the web") We have a copy of the schema definition from schema.org and we normalize "" -> "https://schema.org/" if it's a valid enum value for any enumeration in the range. — Reply to this email directly, view it on GitHub <#3355 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AABJSGNG4ILKIUSQ4FMEJIDYGSV4NAVCNFSM6AAAAAA27O7OVSVHI2DSMVQWIX3LMV43SRDJONRXK43TNFXW4Q3PNVWWK3TUHM3TMOBSGI4TI> . You are receiving this because you are subscribed to this thread.Message ID: ***@***.*** com>