Since unification is so generic, should 'null' values be omitted?

[alaendle]

This is more a question that came to my mind will working with a relatively complex json file (following this schema: https://www.eclipse.org/unide/specification/v3/process-message/#messageDetail)

E.g. consider a simple example json like:

{
    "values": [{"a": "a"},{"b": "b"}]
}

Do we really want to write it back like:

{
    "values": [{"a": "a", "b": null},{"a": null, "b": "b"}]
}

A simple hack would be to just use this snippet to strip null value pairs before creating the object in to JSON (and remove the toEncoding implementation). If a PR is able to better explain my intention just let me know and I provide a draft.
https://gist.github.com/alanz/2465584#file-simple-hs-L21-L22

I know that this might start an endless discussion about the semantic of "null" - but that is not my point, I think a (great, thanks!) tool like json-autotype couldn't support any corner case; so I wondered what is the most pragmatic interpretation in such a case. And since I haven't found any related issue I created this one. But please feel free to just close this issue if you consider this thing as already discussed.

[mgajda]

Good point.
JavaScript API distinguishes them, but best practices don't.

[mgajda]

Should be an option --discard-nulls.