chouzar

Getting the Module name of where a function is executed

I would like to get from “where” a particular function is executed. Is this possible in elixir? For example:

defmodule Sender do

  def to(pid) do
    send(pid, build_message()) 
  end

  defp build_message() do
    %{
      module: __MODULE__, # Can this be obtained at runtime?
      from: self(),
      message: "Hola!"
    }
  end

end

In the future I would like to separate the build_message function into its own thing so it doesn’t need to know the __MODULE__s value (which I guess its defined at compile time).

Macros might be a solution here but have been avoiding them wonder if anyone has any pointers on this?

4 comments

#beginners-questions

0 2729 4

2020-10-03 01:00:16 UTC

First 4 of 4 Posts

hauleth

Could you describe your use case exactly? Because on such abstract example it is hard to reason and maybe provide you better answer.

In general it seems like XY problem.

Post #1

chouzar

Thanks @hauleth appreciate your comments and you’re probably right with the XY problem but let me better explain my case (which admitelly is not much of a use case at this point).

The idea behind comes from wanting to better standardize messaging between different system “components”; concretely I’m trying to apply a standard to messages between my LiveView and LiveComponent modules.

I have this struct defined:

defmodule Action do
  defstruct [
    module: Unknown,
    message: "",
    data: %{}
  ]

  # Along with some constructor helpers
  # ...
end

When sending messages between LiveComponents to LiveViews since both share the same pid, I’m thinking it would be nice to differentiate between them via the component’s module name:

# Although I'm using the Liveview framework as an example 
# this probably not that important, it could be viewed as 
# any process or genserver trying to send and catch a message.

# From the live component
defmodule LiveComponent.SendEmail do
  use Web, :live_component

  # mount, render, process callbacks

  # send being a a helper function
  defp send() do
    send(self(), %Action{
      module: __MODULE__, 
      message: "send-to", 
      data: %{to: "email@email.com", body: "Hello!"}
    }
  end
end

# From the liveview or process that receives the event
defmodule Web.Demo.TreatmentDisplayLive do
  use Web, :live_view

  # mount, render, process callbacks

  # receive the message from the livecomponent
  @impl true
  def handle_info(msg = %{module: LiveComponent.SendEmail}, socket) do
    case msg.message do
      "send-to" -> 
        # process info...

      _other ->
        # process diferently...
    end
  end

I understand even this example will seem nonsensical, what is the point? why not just send a tuple? What I think this achieves is:

A little bit of “extra metadata” for pattern matching purposes, it is on the programmer’s end to enforce a match of this module.
Programmers reading the code might find it easier to parse and know where a message is coming from (module-wise speaking).

I’m currently just experimenting (not a clear path), but would like to expand this idea and see where it goes.

Post #2

al2o3cr

A macro would be useful here - Ecto.Repo does exactly that to pass along the name of the module that’s calling use Ecto.Repo in all the functions:

https://github.com/elixir-ecto/ecto/blob/83d5b7df6faa77b29e88cd665ede812710577eb8/lib/ecto/repo.ex#L267-L269

Post #3