Custom JsonTypeIdResolver is not invoked when object is part of generic type

[nstdio]

Search before asking

• I searched in the issues and found nothing similar.

Describe the bug

Please see the reproducible

Version Information

3.0.4

Reproduction

The JacksonTest#nestedMap and JacksonTest#nestedGenericRecord are failing.

import static org.junit.jupiter.api.Assertions.assertEquals;

import com.fasterxml.jackson.annotation.JsonTypeInfo;
import com.fasterxml.jackson.annotation.JsonTypeInfo.Id;
import java.util.Map;
import org.junit.jupiter.api.Test;
import tools.jackson.core.type.TypeReference;
import tools.jackson.databind.DatabindContext;
import tools.jackson.databind.SerializationFeature;
import tools.jackson.databind.annotation.JsonTypeIdResolver;
import tools.jackson.databind.json.JsonMapper;
import tools.jackson.databind.jsontype.impl.TypeIdResolverBase;

public class JacksonTest {

  private final JsonMapper mapper = JsonMapper.builder()
      .enable(SerializationFeature.INDENT_OUTPUT)
      .build();

  @Test
  void root() {
    //given
    var bar = new Bar("test");
    var expected = """
        {
          "@type" : "BAR",
          "any" : "test"
        }""";

    //when
    var actual = mapper.writeValueAsString(bar);

    //then
    assertEquals(expected, actual);
  }

  @Test
  void nestedMap() {
    //given
    var map = Map.of("bar", new Bar("test"));
    var expected = """
        {
          "bar" : {
            "@type" : "BAR",
            "any" : "test"
          }
        }""";

    //when
    var actual = mapper.writeValueAsString(map);

    //then
    assertEquals(expected, actual);
  }

  @Test
  void nestedRecord() {
    //given
    record Box(Foo value) {

    }
    var box = new Box(new Qux(1));
    var expected = """
        {
          "value" : {
            "@type" : "QUX",
            "any" : 1
          }
        }""";

    //when
    var actual = mapper.writeValueAsString(box);

    //then
    assertEquals(expected, actual);
  }

  @Test
  void nestedGenericRecord() {
    //given
    record Box<T>(T value) {

    }
    var box = new Box<>(new Qux(1));
    var expected = """
        {
          "value" : {
            "@type" : "QUX",
            "any" : 1
          }
        }""";

    //when
    var actual = mapper.writeValueAsString(box);

    //then
    assertEquals(expected, actual);
  }

  @Test
  void nestedGenericRecordExplicitWriter() {
    //given
    record Box<T>(T value) {

    }
    var box = new Box<>(new Qux(1));
    var expected = """
        {
          "value" : {
            "@type" : "QUX",
            "any" : 1
          }
        }""";

    var writer = mapper.writer().forType(new TypeReference<Box<Foo>>() {
    });

    //when
    var actual = writer.writeValueAsString(box);

    //then
    assertEquals(expected, actual);
  }

  @JsonTypeInfo(use = JsonTypeInfo.Id.CUSTOM, property = "@type")
  @JsonTypeIdResolver(FooIdResolver.class)
  public sealed interface Foo permits Bar, Qux {

  }

  record Bar(String any) implements Foo {

  }

  record Qux(int any) implements Foo {

  }

  public static class FooIdResolver extends TypeIdResolverBase {

    @Override
    public String idFromValue(DatabindContext ctxt, Object value) {
      return idFromValueAndType(ctxt, value, value.getClass());
    }

    @Override
    public String idFromValueAndType(DatabindContext ctxt, Object value, Class<?> suggestedType) {
      return switch (value) {
        case Bar _ -> "BAR";
        case Qux _ -> "QUX";
        default -> throw new IllegalStateException("Unexpected value: " + value);
      };
    }

    @Override
    public Id getMechanism() {
      return Id.CUSTOM;
    }
  }
}

Expected behavior

No response

Additional context

No response

[pjfanning]

Can you provide the output to save people time on having to reproduce the issue?

[nstdio]

For JacksonTest#nestedMap

{
  "bar" : {
    "any" : "test"
  }
}

For JacksonTest#nestedGenericRecord

{
  "value" : {
    "any" : 1
  }
}

[pjfanning]

Could you try JsonMapper.builderWithJackson2Defaults() to see if that helps? It could be that one of the configurable features has had its default setting changed in Jackson 3.

[nstdio]

Does not help.

[pjfanning]

And this worked in an older version of Jackson?

[nstdio]

No idea.

[pjfanning]

No idea.

So we'll assume that this never worked and that it's possible that you are not using the feature as it was designed to be used.

I've assigned Copilot to try to investigate but anyone who wants to try this for themselves, feel free to work on your own solution.

[nstdio]

So we'll assume that this never worked and that it's possible that you are not using the feature as it was designed to be used.

No, obviously it should work otherwise functional difference between nestedGenericRecord and nestedGenericRecordExplicitWriter is unexplainable.

[pjfanning]

So why shouldn't you have to create an explicit writer like in nestedGenericRecordExplicitWriter? For me, the nestedGenericRecord issue is an enhancement request and not urgent since we have a way to make this work.
You may also be able to work around the Map case by creating an explicit writer with a TypedReference for Map<String, Box<Foo>>.

[nstdio]

So why shouldn't you have to create an explicit writer like in nestedGenericRecordExplicitWriter?

Because framework does serialization.

For me, the nestedGenericRecord issue is an enhancement request and not urgent since we have a way to make this work.

there's no way If your suggested solution is to "So why shouldn't you have to create an explicit writer like in nestedGenericRecordExplicitWriter"

[pjfanning]

So why shouldn't you have to create an explicit writer like in nestedGenericRecordExplicitWriter?

Because framework does serialization.

For me, the nestedGenericRecord issue is an enhancement request and not urgent since we have a way to make this work.

there's no way If your suggested solution is to "So why shouldn't you have to create an explicit writer like in nestedGenericRecordExplicitWriter"

Even if we make changes for this, it could be a while until they are reviewed and merged and ultimately released. I've given some advice on what you can do to workaround the issue for today.
If you are unhappy with Jackson, there are plenty of other libs out there.