Global middleware configuration

[tanguilp]

Just an idea, I'd be glad to have feedback here: in my apps and libs I often allow configuration options to add Tesla middlewares either globally or for a specific module. For instance:

dev.exs:

config :my_lib, :tesla_middlewares, [Tesla.Middleware.Logger]

config :my_lib, :tesla_middlewares_for_some_module, [Another.Tesla.Middleware]

and then in my code I build the list of middlewares:

my_module.ex:

defp middlewares() do
  [Tesla.Middleware.JSONDecode, Some.Other.Needed.Middleware]
  ++ Application.get_env(:my_lib, :tesla_middlewares_some_domain, [])
  ++ Application.get_env(:my_lib, :tesla_middlewares, [])
end

The global :tesla_middlewares allows applying some interesting things globally:

• enabling / disabling logging in dev or for debugging in prod
• setting application ID header to all outbound requests
• having telemetry for all outbound requests
• etc.

Now my problem is that when having an app using different libs, one would need to configure the same middlewares for each lib:

config :my_app, :tesla_middlewares, [Tesla.Middleware.Logger]

config :my_lib_1, :tesla_middlewares, [Tesla.Middleware.Logger]

config :my_lib_2, :tesla_middlewares, [Tesla.Middleware.Logger]

config :my_lib_3, :tesla_middlewares, [Tesla.Middleware.Logger]

Hence the following idea: couldn't Tesla automatically add middlewares declared using a configuration option (for example: config :tesla, :middlewares) to all clients? Unless there's already a way of achieving that? This also could be a convention / best practice. What do you think?

[teamon]

Configuring things globally is generally not the best idea - it's one of those things that seems nice, but turns out to be painful in the long run.

ref #167