Resolve type variables in resolver return types

[oryan-block]

Checklist

• Pull requests follows the contribution guide
• New or modified functionality is covered by tests

Description

Same problem as #880, but for return types. MethodFieldResolver used the return type of a method inherited from a generic class without resolving the class's type variables, so it saw T where the scanner saw Future<Any> or EventStream. Two cases broke because of it:

1. A non-CompletionStage future behind a type variable got handed to graphql-java as the value, with no error. I found this while reviewing Resolve type variables in resolver parameter types #882.

open class Names<T>(private val name: T) {
    fun name(): T = name
}

class Query : Names<Future<Any>>(FutureTask<Any> { "generic" }.also { it.run() }), GraphQLQueryResolver

{ name } with name: String returned "java.util.concurrent.FutureTask@4554de02[Completed normally]". As a union it failed in the type resolver instead. The check that rejects these futures only treats the value as a wrapper when the method's return type is Future, and it read T. Now it fails with the same error as a method declared to return Future<Any>:

Resolver method 'scenarios.Names.name' resolved to a java.util.concurrent.FutureTask, which graphql-java can't wait on. Return a CompletionStage, such as a CompletableFuture, from the method or its generic wrapper transformer instead.

2. A subscription method whose return type is a type variable bound to a Publisher subtype failed the build:

class EventStream(name: String) : Publisher<Event> by ...

open class Events<P>(private val events: P) {
    fun onEvent(): P = events
}

class EventSubscription : Events<EventStream>(EventStream("inherited")), GraphQLSubscriptionResolver

SchemaClassScannerError: Two different classes used for type Event:
- scenarios.EventStream: return type of method public final java.lang.Object scenarios.Events.onEvent()

(or "No method or field found" depending on the schema). asPublisher, which turns EventStream into Publisher<Event>, ran before the type variables were resolved, so it got P and did nothing. Now it builds and streams the event.

The fix: getReturnType() is now a lazy returnType that resolves the type variables with GenericTypeResolver#resolveTypeVariables, same as getMethodParameterType since #882. getResolvedReturnType (incl. asPublisher) and the Future check both start from it. It's lazy because FieldResolverScanner#findFieldResolver builds a resolver per search and may only use the first one.

This is the first of a few PRs that put the scanner and the data fetchers on one code path for a resolver's Java types. Nothing that builds and works today should change. No public API change, only private members of the internal MethodFieldResolver.

Release note: Resolver methods inherited from a generic class now resolve the class's type variables in their return type too. A non-CompletionStage Future behind a type variable now gets the same error as a declared one instead of returning its toString(), and subscription methods returning a Publisher subtype through a type variable now build.

Testing

Folded both cases into existing tests: inheritedSearch in ReactiveTest and onInheritedPublisher in SchemaParserTest's publisher subtypes test. Both failed before the change. mvn clean verify passes on JDK 25, on JDK 25 against graphql-java 25.1, and on JDK 11.

Also checked through the public API with the classes above, on graphql-java 26.1 and 25.1, before and after. Before: name was the FutureTask string and the subscription schema didn't build. After: the error above, and {"onEvent":{"name":"inherited"}}. A CompletableFuture behind a type variable resolves the same in both.

Schema build benchmark (JMH, graphql-java 25.1) didn't get slower: large schema 37.5 ± 1.0 ms vs 39.3 ± 1.1 on master, same allocation per op.

[sonarqubecloud]

Quality Gate passed

Issues
0 New issues
0 Accepted issues

Measures
0 Security Hotspots
0.0% Coverage on New Code
0.0% Duplication on New Code

See analysis details on SonarQube Cloud