Checking whether a given content type or Lang can be used to read/write a certain Jena model class

[namedgraph]

I'm trying to find a ResultSet-based equivalent to the RDFLanguages class but not able to.

There is RDFParserRegistry and RDFWriterRegistry. There is also ResultSetReaderRegistry and ResultSetWriterRegistry.
But there is only RDFLanguages, not ResultSetLanguages. That seems inconsistent and asymmetric. Would it make sense to add such a class?

More specifically, what can I use as an equivalent of RDFLanguages.contentTypeToLang() but for ResultSet langs?

[rvesse]

See ResultSetLang

[namedgraph]

But that is like Lang :) A media type definition together with their instances, but not their registry.

[afs]

Use RDFLanguages.contentTypeToLang.

ResultSetLang has constants which are Lang and these are registered in RDFLanguages.

ResultSetReaderRegistry.register takes a Lang. There are no ResultSetLang objects.

    Lang x = RDFLanguages.contentTypeToLang("application/sparql-results+xml");
    System.out.println(x);

[afs]

Lang are symbols (unique token; typed and not strings) together with HTTP information.

Ask the question as "Can I use this Lang for <purpose>?".

"text" for example, is only for writing.

[namedgraph]

I feel like I'm struggling to explain myself 😄

I need a method that checks whether a given content type or Lang can be used to read/write a certain Jena model class.

• for Model, there's RDFLanguages::isTriples
• for Dataset, there's RDFLanguages::isQuads
• for ResultSet, there is none?

Do you see what I mean? Unless there's a better way to do it altogether.

[namedgraph]

OK wait, I think I figured it out:

• for Model: RDFParserRegistry::isTriples and something from RDFWriterRegistry?
• for Dataset: RDFParserRegistry::isQuads and something from RDFWriterRegistry?
• for ResultSet: ResultSetReaderRegistry::isRegistered and ResultSetWriterRegistry::isRegistered

That looks like a better approach overall, but then RDFWriterRegistry has no methods to distiguish between triple and quad Langs?

[afs]

triples/quads isn't an aspect of a write format (though it might be for some custom RDFFormat). It applies to both reading and writing. Ask RDFLanguages (arguable, remove from RDFParserRegistry).

You can write triples to quad formats.
JSON-LD confuses the situation - it is commonly talked about as graph syntax but it is really a quad syntax.

ResultSets are not split by triple/quad capabilities. Either it is registered (has a factory) for a use or not.

[namedgraph]

triples/quads isn't an aspect of a write format (though it might be for some custom RDFFormat). It applies to both reading and writing. Ask RDFLanguages (arguable, remove from RDFParserRegistry).

Andy I understand what you're saying, but it's not really what I'm asking about. Let me explain my use cases:

• building an Accept header value. Here I need a registry of readable triple/quad/result set langs because I want to prioritise langs in each group differently.
• in JAX-RS response body writing happens in providers that depend on the return object type, e.g. MessageBodyReader<Model>, MessageBodyReader<Dataset> or MessageBodyReader<ResultSet>. Each of those providers needs to know which langs can be used based on the generics class. Again it requires a registry of readable/writable types that apply to each class.

You can write triples to quad formats.

I understand this, but not sure how it's related to my use case. If quads were requested, JAX-RS will write a Dataset and needs to use a quad format. Returning a triple format (even if there are no named graphs in the dataset) would violate the conneg.

ResultSets are not split by triple/quad capabilities. Either it is registered (has a factory) for a use or not.

Yes this is clear as well, I don't think I was trying to claim ResultSet is related to quads or triples.

[namedgraph]

To summarise, my main 2 quibbles are:

• the classes RDFParserRegistry, RDFWriterRegistry, ResultSetReaderRegistry, ResultSetWriterRegistry are all Lang registries but are not aligned. Would it be reasonable for them to implement a single interface such as

  interface LangRegistry {
    Collection<Lang> registeredLangs();
  }

  ? If you agree, I can create an issue.
• RDFLanguages::isTriples and RDFLanguages::isQuads, but no RDFLanguages::isResultSet. And why does RDFLanguages need a registry if parsers and writers already have their own Lang registries?

[afs]

interface LangRegistry

They are static classes - not interfaces. We have just released Jena5; changing to objects is a breaking change.

If you want parallel capabilities via objects with LangRegistry interface, build it. Add the setup code as ServiceLoader and initialize after Jena. That could also add the "Accept" prioritization.

RDFLanguages::isResultSet could be done - make it a call to (new) ResultSetLang.isRegistered (+machinery).

I can create an issue.

Are you going to provide a PR?

[namedgraph]

I can try :) Maybe I could add singleton objects but keep the static methods.

[afs]

Seems over-engineering.

• Add RDFLanguages::isResultSet - make it a call to (new) ResultSetLang.isRegistered.
• interface LangRegistry is a functional interface so implement three functions (one liners) calling the appropriate RDFLanguages::is????? function.
• Pass the right function to code that needs a general lookup algorithm specialised to a chosen context.

[namedgraph]

@afs seems like this has been addressed by #2510? Thank you for that :)

I still feel like a contentTypeToLang method for ResultSetLang is missing though?

[namedgraph]

Ah OK, I see they're part of RDFLanguages now - which was not the case in 4.7.0:
https://github.com/apache/jena/pull/2510/files#diff-4efee2e599cd106d504a1687b6fff298e6fb0b24d41e4db707d88088f76d2afcR102