This is a bit of a pain, because hooks have to be built and started before the worker actually is... so the irony is the worker doesn't even exist during start up.

So... Need to think of another way around it. Considering an option to pass into a hook of requires_worker: true which causes the worker to ping { :worker, worker } on startup to allow the hook to configure itself accordingly.

This is kinda gross, but it does lend itself to backwards compatibility well.

defmodule MyModule.WorkerHook do
  use Cachex.Hook

  def init(args) when is_map(args) do
    { :ok, args }
  end

  def handle_modify({ :worker, worker }, state) do
    { :ok, Map.put(state, :worker, worker) }
  end

  def handle_notify(_action, state) do
    { :ok, IO.inspect(state) }
  end
end

Ok I have this working now... I'll just neaten it up a bit and get a PR in. It's not the nicest interface, but I expect it to be less frequently used, so it shouldn't matter too much.

Hit a bug with overriding handle_call/2 to give myself access to handle_modify/2 (namely that it breaks too easily). For this reason, just going to provide changes to handle_call/2 and document well.