Hi!

I’m trying to avoid redundant testing (that means, tests that are made in the same way all the time but with slightly different parameters). For example, I have this all the time when I want to test that the call to a controller was ok:

setup do
  count = 3
end

test "responds with right schema", %{conn: conn, swagger_schema: schema} do
    conn
    |> create_authenticated_conn()
    |> get(~p"/something",  %{a: 1})
    |> validate_resp_schema(schema, "Schema")
    |> json_response(200)
    |> assert_length(count)
end

The things that change are the method call (get, post…), the params, the schema to evaluate, and, optionally if I want to make more assertions (as checking the length of the result array).

So, I thought of creating a macro that simplifies all this work, parameterizing the values that need to be parameterized and also allowing a do: block element to add additional checks. I end up with this:

defmacro test_ok(schema_name, method, url, params \\ nil, do: block) do
  quote do
    test "responds with right schema", %{conn: conn, swagger_schema: schema} do
      response =
        conn
        |> create_authenticated_conn()
        |> unquote(method)(unquote(url), unquote(params))
        |> validate_resp_schema(schema, unquote(schema_name))
        |> json_response(200)

      evaluate_response(response, block)
    end

    defp evaluate_response(var!(response), block) do
      unquote(block)
    end
  end
end

If I ignore the response part and the block part (evaluate_response method) all works fine. As soon as I want to add to the macro the block part and pass the response to inside the body, I have two problems. First, let’s put an example of code:

test_ok("Schema", :get, ~p"/something") do
  assert_length(response, count)
end

Then, the first problem is about the response not being recognized inside the block code. And the other is count not being available neither, although in the way normal test macro, is recognized perfectly.

What is my mistake here?