DataLoader KV does not resolve fields in batch

Hi,

Im trying to get DataLoader working with an external service.

This is my field resolver

object :recent_conversation do
    # other fields here
    field :sender, non_null(:user), description: "Last sender for the conversation" do
      resolve fn %{sender_id: sender_id}, _args, %{context: %{loader: loader}} ->
        loader
        |> Dataloader.load(:user, :user, sender_id)
        |> Dataloader.run()
        |> on_load(fn loader ->
          loader
          |> Dataloader.get(:user, :user, sender_id)
          |> case do
            %{} = user ->
              {:ok, user}
          end
        end)
      end
    end
  end

And here is my dataloder KV source

def data() do
    Dataloader.KV.new(&fetch/2)
  end

  def fetch(_batch_key, users) do
    IO.puts("here")
    IO.inspect(users)

    loaded_users = users |> MapSet.to_list() |> load_users()

    users
    |> Enum.with_index()
    |> Enum.reduce(%{}, fn {id, idx}, results ->
      Map.put(results, id, Enum.at(loaded_users, idx))
    end)
  end

  defp load_users(ids) when is_list(ids) do
    IO.puts("Loading users #{inspect(ids)}")

    case ApiClient.get_users_bulk(ids) do
      {:ok, %{users: users}} ->
        users

      {:error, _err} ->
        {:error}

      {:api_error, _err} ->
        {:error}
    end
  end

I prefer to have a generic data loader which can be used to load users by passing different fields, so I avoided using batch resolution(API accepts user IDs in batch)

When I’m running the following query, I see multiple Loading users logs(multiple API calls)

query {
  myRecentConversations{
    id
    content
    createdAt
    unreadCount
    sender{
      id
      username
      avatar
    }
  }
}

Why dataloader is not loading all users per this GraphQL query in a batch? I’ve used Facebook Dataloader in JS projects and it loads all data in a batch when running in a single graphql query context unlike this.

Is there something I can do to make this execute a batch call to API?

This is your issue. You shouldn’t be calling Dataloader.run yourself, just pipe straight to on_load(fn loader ->) and then Absinthe will call Dataloader.run internally after the whole batch has been accumulated.

1 Like

Thanks, it worked :slight_smile:

Are there other improvements I can do to this code?
My main reason was to use the same loader with different fields, thus I didn’t use dataloader macro.