Bug: API Objects with a Single Scalar Field Are Interpreted As Scalars

[sherl0cks]

This may be related to the use of Kotlin data classes in my code.

For the following API and model

@GraphQlClientApi(endpoint = "https://api.com/graphql")
interface Client {
    fun query(foo : String): Response
}

data class Project( var apiObject : ApiObject)

data class ApiObject(
    var id: String = "",
    var name: String = "" // This second field must be added or graphql client thinks that `Project` is a `String` due to its "scalar constructor"
) 

In this case, if ApiObject only has id, graphql request with be apiObject, instead of apiObject{ id }. Adding a second field will result in apiObject{ id name }

Seems caused by this chain:

smallrye-graphql/client/implementation/src/main/java/io/smallrye/graphql/client/typesafe/impl/GraphQlClientProxy.java

Line 86 in a9f7313

if (type.isScalar()) {

smallrye-graphql/client/implementation/src/main/java/io/smallrye/graphql/client/typesafe/impl/reflection/TypeInfo.java

Line 122 in a9f7313

|| scalarConstructor().isPresent();

smallrye-graphql/client/implementation/src/main/java/io/smallrye/graphql/client/typesafe/impl/reflection/TypeInfo.java

Line 136 in a9f7313

Stream.of(getRawType().getConstructors()).filter(this::hasOneStringParameter))

[phillip-kruger]

@t1 any thoughts ?

[t1]

I think this is exactly what we discussed in eclipse/microprofile-graphql#251: a single-string constructor is currently considered as defining a custom wrapper class.

I will do a change for that case in the client.

[phillip-kruger]

Yea this was my worry in applying that by default. I think we need to apply this with some annotation so that it's a more explicit decision from the user.