`Type missing from the context` warning when using prefix

[fpotier]

When a prefix is defined in the context and referenced later:

RoCrateMetadataContext context =  new RoCrateMetadataContext();
context.addToContext("rdfs", "https://www.w3.org/2000/01/rdf-schema#");
...
builder = new DataEntityBuilder();
builder.addType("rdfs:Property");

The following warning is emitted:
type rdfs:Property is missing from the context!

#226 and this made me think, would it make sense to rely on titanium-json-ld for validation/flattening?

[Pfeil]

Thanks again for your many contributions, this is really valuable input for us.

I did not yet look into the issue, but if I can confirm this, it is clearly a bug. As far as I get it from your description, it is only a warning and therefore not too urgent, right?

#226 and this made me think, would it make sense to rely on titanium-json-ld for validation/flattening?

In other words, to allow adding any kind of value, but then flatten it at some point? Could be done, I assume. I wonder if there was a thought behind the way it is done now (the initial developer is not on the project anymore). If we do not have to sacrifice validation or user experience (or can even improve it), I think there stands nothing against it. We'd happily accept a PR (even with a proof-of-concept). As the priority is not super high, I would not do it already in the next release except we have a surprisingly quick fix. I wonder if such a solution would end in a 3.0.0 of the library. We'd have to look carefully if we break any API expectations if we put it into a 2.x release.

I'll report with more details when I had some time to look into it.

[ThomasJejkal]

Sorry to chime in, but I just tried to reproduce the issue and I was not able to. I used the following code snippet the create a basic crate:

RoCrate roCrate = new RoCrateBuilder("name", "description", "2024-02-09T08:21:41Z", "licenseIdentifier")
                .addValuePairToContext("rdfs", "https://www.w3.org/2000/01/rdf-schema#")
                .addDataEntity(
                        new FileEntity.FileEntityBuilder()
                                .setId("survey-responses-2019.csv")
                                .addProperty("name", "Survey responses")
                                .addType("rdfs:Property")
                                .addProperty("contentSize", "26452")
                                .addProperty("encodingFormat", "text/csv")
                                .build()
                )
                .build();

As you can see, the context entry and the property are provided to the builder. Afterwards, I wrote the crate to disk and received the following metadata document:

{
  "@context" : [ "https://w3id.org/ro/crate/1.1/context", {
    "rdfs" : "https://www.w3.org/2000/01/rdf-schema#"
  } ],
  "@graph" : [ {
    "name" : "name",
    "description" : "description",
    "@id" : "./",
    "@type" : "Dataset",
    "license" : {
      "@id" : "licenseIdentifier"
    },
    "datePublished" : "2024-02-09T08:21:41Z",
    "hasPart" : {
      "@id" : "survey-responses-2019.csv"
    }
  }, {
    "about" : {
      "@id" : "./"
    },
    "conformsTo" : {
      "@id" : "https://w3id.org/ro/crate/1.1"
    },
    "@id" : "ro-crate-metadata.json",
    "@type" : "CreativeWork"
  }, {
    "name" : "Survey responses",
    "contentSize" : "26452",
    "encodingFormat" : "text/csv",
    "@id" : "survey-responses-2019.csv",
    "@type" : [ "rdfs:Property", "File" ]
  } ]
}

Looks fine to me and there was also no warning. I agree, that some more elaborated JSON-LD handling using a dedicated library could be beneficial, but that's another story. ;-)

[Pfeil]

Ok, so I dived deeper into this and came to the following conclusions. I used this code to reproduce the issue:

  @Test
  void addType_withContextAvailable_printsNoWarning() throws IOException {
    RoCrate crate = new RoCrate();

    RoCrateMetadataContext context =  new RoCrateMetadataContext();
    context.addToContext("rdfs", "https://www.w3.org/2000/01/rdf-schema#");

    crate.setMetadataContext(context);
    crate.addDataEntity(
            new DataEntity.DataEntityBuilder()
                    .addType("rdfs:Property")
                    .build()
    );
    Writers.newFolderWriter().save(crate, ".tmp/mytest.crate");
  }

And while using a builder works, this is just because there is a check missing. See the TODOs at the bottom. Also, we obviously lack documentation in this section. The actual issue is, though, that we do currently not support validating prefix terms. We currently assume that a term is a complete term, which is obviously not correct.

A workaround would be to either add each term separately the way you did it or just use the term "Property" and add the URL directly (addToContextFromUrl). In the latter case you will get a warning that ro-crate-java does not understand the format of the context (no ttl support so far), though.

Please check the workaround achieves the result you like to have until this is fixed.

TODOs which should fully fix this issue:

• fix builders do not check the entities being added (fixing also means the context for an entity needs to be there in beforehand to avoid warnings!)
• document context methods properly
• implement prefix support (I think the titanium json ld library will help with this and hopefully replace most of our context code)
• support getting context terms from turtle format

[Pfeil]

Issue fixed in #267, will land in v2.1.0.