Make JSON schemas enter Object and Variant fields.

[jamadden]

No description provided.

[jzuech3]

LGTM!

[jamadden]

OK, cool.

Just be aware that we're traversing through a lot of things we didn't in the past. That means the schemas will get bigger, and I can't help but wonder if there may be some chance of disclosing data we don't actually expect to (though I haven't been able to come up with a specific scenario). It may be necessary to define the TAG_HIDDEN_IN_UI more often...in fact, I think we should probably query that field before we output key_type and value_type or the fields of a variant (right now it's only queried for the top level). Thoughts?

[jzuech3]

I think we do need to worry about the schemas getting bigger, but I believe it is necessary to give accurate info on the required types of underling sequence fields.

Agreed on making sure we also check hidden within variant. Do we need to make sure post_process_field is called in variant as well?

[jamadden]

Agreed on making sure we also check hidden within variant. Do we need to make sure post_process_field is called in variant as well?

I have no idea what it's used for (there's no implementation, docstring, or testing of it here), so since it is currently only called for top-level fields, my instinct would be to say "no, don't start suddenly calling it for nested fields."

[jzuech3]

I'm not sure of the intended use-case either; so we can probably ignore it for now. Yesterday, I did override post_process_field to get the title/description defs in for a client.

This PR looks good to me.

[jamadden]

Thanks!