it is not obvious how to make them but it is priority #1

but release early, release often as they say

now, we have a sprint planning tomorrow, we'll start by introducing a new

type

construct in the data model, that can be used as embedded documents in mongo, but that work means the client must support this too, so I'd say the feature might not be in 3.3.0 still due to having a two team effort here

and this introspection is the first idea how it should be done, please report all bugs and ideas in the issue tracker so we know how you're using it

the interesting problems with introspecting embeds: • how do we name it, my idea is something like model name + field name • should we do deduplication, if two models have almost similar embeds, should we have two types or merge it to one • how would the naming work in the second example mongo introspection is quite best effort basis, we cannot do everything as we can with postgres for example

let's see, if we have

{ name: "foo", data: { something: "bar" } }

so we can get the model name from the collection, but it will have one embed in the

data

this is still easy, but what if we have another model that has an embed with

{something: "bar"}

and we get into the model of deduplication

we either create many similar types, let the user to merge them, or implement deduplication where we merge similar types together

this gets also really hairy, the other embed might have

{ something: "foo", other: 2 }

, so we can have

type UserData {
  something String,
  other Int?
}

but I think we go with the kind of dumb solution first, iterate from there