JsonUnwrapped on multiple fields has problems with deserialization

[kariem]

I am in the process of trying to understand why I cannot correctly configure Jackson serialization to do what I want. I have already asked this on StackOverflow, but I haven't found a solution yet.

I have a simple bean which holds a string and a separate map of its localizations

public class LocalizedField {

  private String value;
  private Map<String, String> value_l10n;

  // ... getters and setters
}

I want to use this in another class

public class MyFancyClass {

	@JsonUnwrapped
	private LocalizedField name;
	@JsonUnwrapped
	private LocalizedField description;

    // ... getters and setters
}

This should convert to the following in JSON

{
  "name": "name",  // name.value
  "name_l10n": {   // name.value_l10n
    "de": "name",
    "fr": "nom"
  },
  "description": "description",  // description.value
  "description_l10n": {          // description.value_l10n
    "de": "Beschreibung",
    "fr": "description"
  }
}

Using a contextual serializer that sets property name looks ok for the serialization.

However, I have a problem when trying to deserialize this. I always get the string in value instead of the whole node that represents MyFancyClass. I would like to just read the two properties (name and name_l10n) to construct a LocalizedField.

I have prepared a sample project here: kariem/jackson-multiple-fields

Any hints on what I could try?

[JooHyukKim]

I always get the string in value instead of the whole node that represents MyFancyClass. I would like to just read the two properties (name and name_l10n) to construct a LocalizedField.

Could you elaborate further how exactly is deserialized?

[JooHyukKim]

@kariem Try @JsonAnySetter?

[kariem]

Could you elaborate further how exactly is deserialized?

Not sure, if I understand the question correctly. If I set the debug breakpoint to the point where I can see the current node via DeserializationContext.readTree(JsonParser), I receive the text node.

In the current deserializer, this would be here

So, for the input

{
  "name": "the name",  // name.value
  "name_l10n": {       // name.value_l10n
    "de": "der Name",
    "fr": "le nom"
  }
}

I get would get "the name".

If I execute the statement again (ctxt.readTree(p)), I seem to get the full JSON node.

[cowtowncoder]

Oh. Keep in mind that for @JsonUnwrapped there is extensive buffering, so input may not look quite like actual input document (JsonParser being used is nested inner class of TokenBuffer, I think). Custom deserializers cannot really be combined with unwrapping either.

[kariem]

What options other than custom deserializers would I have then?

[cowtowncoder]

We would need full reproduction, not just snippets.

I suspect in general use of more than one @JsonUnwrapped (at least ones without prefix) will probably not work well, but it is difficult to know without seeing all the pieces.

[kariem]

I have linked https://github.com/kariem/jackson-multiple-fields what is missing there?

[JooHyukKim]

Jackson is light enough to not require a personal repo for reproduction so it is highly preferable to provide reproduction like this.

I think that's what he means? Or just missed the link.

[kariem]

I have just created FasterXML/jackson-databind#5182

Does that work?