Possible regression when using @JsonIdentityInfo

[hgeraldino]

When I try to serialize/deserialize instances from a class marked with @JsonIdentityInfo using the same ObjectMapper, I get this JsonMappingException: "com.fasterxml.jackson.databind.JsonMappingException: Already had POJO for id (java.lang.Integer) [com.fasterxml.jackson.annotation.ObjectIdGenerator$IdKey@3f0456e4".

See: https://gist.github.com/hgeraldino/ed389bfacc579d1eb0c6

This very same test works on the 2.3.0 version, but breaks on the most recent 2.4.1 version. It doesn't matter if I try to deserialize different objects, as long as they have the same property value the application breaks.

Is this an issue or something somewhat expected?

[cowtowncoder]

It is hard to say for sure, but it could be actual fix. Object Ids need to be unique.
2.4 did add full support (finally!) for forward-references, so it is possible that things were passing with 2.3, due to missing checks and handling for cases where reference to an id precedes id itself.

But there are other possibilities; like it being a bug, or that handling or definition of scope (of uniqueness of object id) is incorrect.

[cowtowncoder]

Quick look at gist suggests that maybe there is a bug. I will try to play bit more with this tonight or tomorrow, to see what is going on. I did not see duplicate definitions in sample JSON. But if there is a problem, I am guessing it'd be more with scoping of id uniqueness.

[hgeraldino]

Yes, I also think it has to do with the scope. If you try to deserialize the JSON once things just work, but if you try to deserialize it again (or deserialize another JSON string that represents a different object but with the same property) it breaks.

There should be a way for resetting the _items Hashmap on the SimpleObjectIdResolver class after a de/serialization operation completes on an object.

[cowtowncoder]

Ah! That does ring a bell. It would indeed be scoping, but not in-document scoping (that I was thinking), but inter-document scoping (uninentional, i.e. bug).

So: I think #491 actually resolves this for 2.4.1.1 (micro-release), and upcoming 2.4.2.
If you can try building from master, or use 2.4.1.1 as dependency, problem should be resolved.

[hgeraldino]

Oh you nailed it! Problem is indeed solved on 2.4.1.1

Thank you!!!!!!!!!!!

[cowtowncoder]

Awesome!