Tatu Saloranta

Was fixed in `jackson-databind` 2.12.0, actually.

Ok, I see what the intent is. Not sure how to address it yet, but whatever approach taken would go in 3.0 (since that's next after 2.9).

@mydata Generally documentation should match implementation unless indicated otherwise. So I'd start with that.

I am not sure I understand this exactly, but I suspect the difference is for number where `long` property value will become `LongNode` whereas regular `1` from JSON would become...

While I agree this is unexpected, I tried to explain why this occurs, and reasons why this is nor easy to fix nor likely to be changed in near future....

Yes, agreed. Timely RFE as well, due to: https://github.com/FasterXML/jackson-databind/pull/1364 which is for the same thing, but applying generally, via `@JsonFormat`. There are some open questions on possibly configuration, how to...

If something is done, something is noted on the issue typically. I am not aware of any related work. PRs welcome.

I am not sure how this can really work unfortunately as there is no obvious "empty" representation of "untyped" value (nominal type of `java.lang.Object`). I guess theoretically it could be...

Hmmh. Not sure about call flow, use of `EmptyValueProvider`... The core concept with databind is that serializers and deserializers have the main responsibility for handling, over centralized entities. So I...

Interesting... while semantics are slightly different (in CSV output everything is essentially a String), this could be closed enough. The only question I guess would be whether existence of CsvGenerator.Feature.ALWAYS_QUOTE_STRINGS...