Fix parsing of node IDs with local middleware in modern schemas (v1.4)

[avitex]

Fixes #95

[avitex]

In the process of refactoring the tests by pulling the supporting schema's into one (see below), I'm coming up with another error that I think I've narrowed down to my macro usage (not currently in this PR).

defmodule Absinthe.Relay.Node.ParseIDsTest.SupportingSchema do
  defmacro __using__(flavor) do
    quote do
      use Absinthe.Schema
      use Absinthe.Relay.Schema, unquote(flavor)
      # ...
    end
  end
end

defmodule Absinthe.Relay.Node.ParseIDsTest.SchemaClassic do
  use Absinthe.Relay.Node.ParseIDsTest.SupportingSchema, :classic
end

defmodule Absinthe.Relay.Node.ParseIDsTest.SchemaModern do
  use Absinthe.Relay.Node.ParseIDsTest.SupportingSchema, :modern
end

This test fails with Unknown argument \"input\" on field \"updateParentLocalMiddleware\" of type \"RootMutationType\".", and I have confirmed with Absinthe.Schema.types/1 that it is indeed the case.
Note manually pasting the schema body into SchemaClassic and SchemaModern works.

test "for modern schema" do
  encoded_parent_id = Base.encode64("Parent:1")
  encoded_child1_id = Base.encode64("Child:1")
  encoded_child2_id = Base.encode64("Child:1")
  result =
    """
    mutation FoobarLocal {
      updateParentLocalMiddleware(input: {
        parent: {
          id: "#{encoded_parent_id}",
          children: [{id: "#{encoded_child1_id}"}, {id: "#{encoded_child2_id}"}, {id: null}],
          child: {id: "#{encoded_child2_id}"}
        }
      }) {
        parent {
          id
          children { id }
          child { id }
        }
      }
    }
    """
    |> Absinthe.run(SchemaModern)

  expected_parent_data = %{
    "parent" => %{
      "id" => encoded_parent_id, # The output re-converts everything to global_ids.
      "children" => [%{"id" => encoded_child1_id}, %{"id" => encoded_child2_id}, nil],
      "child" => %{
        "id" => encoded_child2_id
      }
    }
  }
  assert {:ok, %{data: %{"updateParentLocalMiddleware" => expected_parent_data}}} == result
end

[avitex]

As this PR works with it's current test, shall I move this to a separate issue?

[bruce]

@avitex If it's for what you think is a different issue, then definitely pull it out. Let's keep these as narrow as possible.