This might be a slightly controversial point but I don’t think Elixir code needs to be very different from OO code. A user struct is not very different from an immutable user object. The domain core being functionally pure is I link a good idea because it makes units very easy to test.
By the time you step out to application services and infrastructure, your code will no longer be pure. However at this level a Elixir program will probably be coordinating multiple process and they encapsulate state so again many of the OO lessons are still relevant.
So in summary I think that the onion architecture in Elixir is a good idea.
