@JsonValue not considered when evaluating inclusion

[chrylis]

I have an object (a links holder) that is essentially a container around a java.util.Map. The map is annotated with @JsonValue, and this behaves as expected (the map is dumped as a JSON object inline).

However, I only want this included when the map is non-empty, so I annotated the getLinks() method with @JsonInclude(NON_EMPTY). However, it appears that the inclusion filter is applied before the value substitution, so it is always included even as {}.

I'm using Jackson 2.11.2 (managed via Spring Boot).

@Data
class Wrapper {
  @JsonValue
  Map<String, String> values;
}

@Data
class TopLevel {
  @JsonInclude(NON_EMPTY)
  Wrapper nested;
}

[cowtowncoder]

Interesting. Fixing this might be as simple (?) as adding functional isEmpty() in JsonValueSerializer....

[cowtowncoder]

Actually, while the basic idea is to indeed implement isEmpty(), details get rather tricky (due to existing workarounds).
I think I can solve this but it's not a good starter problem for anyone else.

[chrylis]

Thanks! Should I plan to bump to 2.12.0 when it's out to see the new output?

[cowtowncoder]

I am planning to release one more rc, 2.12.0-rc2 within 1 week.

[mbenz89]

Seems not to work on my side. I'm using 2.12.3 and the @JsonValue annotated getter is still serialized to "null"/default-value if I set JsonInclude to NON_NULL/'NON_DEFAULTS` in the ObjectMapper properties or view annotating the relevant value field.

My POJO looks like this:

class Container {
@JsonValue @JsonInclude(Include.NON_NULL)
 public getValue(){
 // compute value
 }
}

Any ideas @cowtowncoder ?

[cowtowncoder]

@mbenz89 This is a bit different combination than issue here and is not supported. Technical reason is because getValue() is not an accessor in POJO sense (but provides alternate surrogate/delegate representation to serialize); inclusion would be determined by annotation either on returned type (class annotations) or on property that refers to this value type.
If you want, you can file a new issue to request (RFE) for this combination to work; I don't know if that would be easy to implement or not.

[mbenz89]

@cowtowncoder I see. Thanks a lot for the elaboration!