I have a custom resolver in one of my graph ql types that requires all nested values to be loaded before it can run a certain calculation. I went through the dataloader documentation and github source and figured out how to use the Dataloader.load and on_load. However I cannot figure out how to load a nested relationship.

The in efficient solution is to do Repo.get inside the resolver but that creates an N+1 query problem. I know one solution is to use a batch and write a customer query but I am hoping to leverage the other fields already being loaded.

In example below, how do I load likes ?

Ecto Schema:

schema "User" do
 *** other fields ***
  has_many :posts, Module.Posts

schema "Posts" do
   has_many :likes,  Module.Likes

schema "Likes" do
  field :type, :string

Absinthe Type Definition:

object :like do
 field(:type, :string)
end

object :posts do 
  field(:likes, :like, resolve: dataloader(Module)
end

object :users do
    *** other fields ***
 field(:posts, :post, resolve: dataloader(Module))
    )

 field(:some_calc_value, :integer,
      resolve: fn parent, _, %{context: %{loader: loader}} ->
        loader
        |> Dataloader.load(Module, :posts, parent)
        |> Dataloader.load ( ????????)  // HOW DO I LOAD LIKES FOR EACH POST ?
        |> on_load(fn loader ->
         posts = 
          loader
          |> Dataloader.get(Module, :posts, parent)

         likes = ???? How DO I Get the likes ? 

          calculate_value(posts, likes)
          |> (&{:ok, &1}).()
        end)
      end
    )