`declaringClass` of "enum-as-POJO" not removed for `ObjectMapper` with a naming strategy

[tynakuh]

Happens in 2.9.9.

When you change ObjectMapper to use a different naming strategy, i.e:

ObjectMapper mapper = new ObjectMapper();
mapper.setPropertyNamingStrategy(PropertyNamingStrategy.SNAKE_CASE);

When serialising/using writeToString(Object value) for an object such as:

@JsonFormat(shape = JsonFormat.Shape.OBJECT)
public enum Foo
{
    BAR("Foo", 0),
...

"declaring_class" will be added to the JSON payload.

Normally this is removed at:
com/fasterxml/jackson/databind/ser/BasicSerializerFactory.java:1139 but because in
com.fasterxml.jackson.databind.introspect.BasicBeanDescription#removeProperty (It's only usage as far as I can see) we aren't using InternalName, declaringClass is not removed. Potential fix here is to change prop.getName() to prop.getInternalName(), a quick test of that seemed to work in my case.

[cowtowncoder]

Interesting. Thank you for reporting this (and apologies for slow follow-up) -- sounds like a bug indeed! I'll verify that suggested fix works: I suspect it should in some form: does for this property, but changing public method to change lookup might break other handling. But either addition of secondary lookup or (if need be) adding new method should work.

[cowtowncoder]

Interestingly enough, I can reproduce this on 2.10 branch (including 2.10.2) but NOT on 2.11.
So I think something got fixed since 2.10, although none of issues fixed so far seems related.
Still, adding test to verify fix I consider this resolved for 2.11 (and 3.0 via master); will add further notes if I find something explaining which change resulting in fix,.

[LukasKarasek]

Thanks
I delete my post because first I thought it was fixed in 2.10.1 but then I realize that this is still ongoing ticket with this bug.
I did a workaround on our side and will wait for fixed version.
L.

[cowtowncoder]

Ok. At this point I think I will not try fixing it for 2.10.x partly since it is resolved for 2.11, and partly since I suspect there is non-trivial risk for regression for 2.10.x. And mostly I guess as I really hope to get 2.11.0 released relatively soon (early Feb 2019).

[LukasKarasek]

OK, thanks , I'll try when it will release .