Cannot use dynamic graphql client to invoke mutation with non-null parameter

[frisket]

I'm attempting to call a graphql API that has a non-null parameter defined.

Eg, with a schema like this:

type Film {
title: String
}

mutation {
updateTitle(name: Film!) {
title
}
}

I cannot invoke the "updateTitle" API because I get a name validation error trying to build the graphql client.

The specific exception I get is:

java.lang.IllegalArgumentException: Invalid name: Film!
        at io.smallrye.graphql.client.core.utils.validation.NameValidation.validateName(NameValidation.java:68)
        at io.smallrye.graphql.client.core.VariableType.varType(VariableType.java:17)
        at io.smallrye.graphql.client.core.Variable.var(Variable.java:77)

Attempting to invoke as so:

Document document = Document.document(
	                                       operation(
					                OperationType.MUTATION,
							List.of(var("rec", "Film!")),
					                field("updateTitle", arg("name", var("rec", "Film")))
					        ));

Response response = dynGraphqlClient.executeSync(document, Map.of("rec", film));

I am able to send the same query manually via other tools (eg: Altair) to the endpoint succesfully.

Found this older ticket that seems similar: quarkusio/quarkus#18194

[mskacelik]

Hi @frisket, try using io.smallrye.graphql.client.core.VariableType.nonNull method to wrap the Film type instead of using an explicit !
Something like this should work (my reproducer):

    Variable inputVar = Variable.var("arg1", nonNull("String")); // ScalarType.GQL_STRING
    Document query = document(
            operation(vars(inputVar), field("returnSomething", arg("input", inputVar))
        )
    );
    Response response = client.executeSync(query, Map.of("arg1", "Hello"));

The problem is that we added naming validators, and they do not accept characters such as !, [, ]. But this might be considered a bug (because I did not think of use explicitly like that (same with lists)), so I might try to add an additional naming rule for types. @jmartisk wdyt?

[jmartisk]

I would say that using these wrapper methods to denote non-null and arrays is intentional and good, it makes the code a bit more type-safe, no? I would say that we could update the IllegalArgumentException's message in a way that if it detects that the user wrapped a type in [] or added !, it will tell the user to use the proper methods instead of this.

[mskacelik]

Yeah, the error messages for the validators are lacking and don't say anything of importance in general.

[mskacelik]

Yeah, the error messages for the validators are lacking and don't say anything of importance in general.

PR: #2260