Enable writing of json files

[daniel-sol]

There are several datatypes in the fmu domain that as a data structure best fits in as a dictionary. The first types that spring to mind:

• parameter.txt
• observations
• alias lists (Although there has been discussed that these rather belongs as metadata for the case)

[perolavsvendsen]

Is the data class "dictionary"? Or "keyvalues"? Or "json"? I'm thinking of consistency with existing classes, e.g. table, surface, cpgrid etc.

To clarify, we will end up with is something like this: myfile.yml .myfile.yml.yml

[daniel-sol]

I have suggested to have class as dictionary. See PR #355

[perolavsvendsen]

I am wobbling a bit but I think I am leaning more towards using json (or yaml) as class, not dict. I worry about confusion with the dict data type in Python. A Python dict can contain "anything", but in this case we are constrained by what can actually be represented as json.

But, that being said - to materialize a JSON from Python, you have a dict instance. And when parsing JSON with Python, you end up with a dict instance.

This is conceptually tricky, as it mixes class and format. This is much easier with e.g. table/csv. I think we are missing the analogue to table for this. JSON essentially contains two structures: name/value pairs, and (ordered) lists. But I think it will be confusing to start annotating those.

🤯

[daniel-sol]

Well, I beg to differ. I don't think json or yaml, or yml for that matter are the right terms. These are formats, not data structures. But I have changed the classname to dictionary to keep it distinct from pythons dict.