Just a thought off the top of my head. The Phoenix Router is “just a plug”. It tends to be the Plug at the end of the pipeline, but it is still a Plug. So it seems you could develop a Plug that looks up a new route in a dynamic data source (an ETS table, or database).

Thank you for reply!
I was thinking this way too…
I already tried to create a Plug and added that to the Pipeline.
I thought that I can analyse requested url and match through the regexp.
Once matched forward it to required function (handler).

But the problem is the following:

• it gives the error, something like undefined route and gives me a list of valid routes.

I suppose the question is - during Plug execution, how to connect to the point when Phoenix analyses existing routes and give it the new route taken from DB or User Input ?

You cannot use the Router of plug or phoenix for the routing in this case. You need to create a wildcard route, which simply matches any url that finds it’s way to your endpoint and use custom plug to implement the routing to controllers or whatever needs to receive the request if the url is valid. The routers of plug and phoenix are based on compiled pattern matching functions, so they cannot be dynamic (or even support regex).

Thank you for reply.
I understand that I cannot use that as I already tried without success.
I would prefer not to create a wildcard route due to performance issues.

The question is - how to achieve the desired behaviour without losing performance.

• It is possible to restart Router independently of Phoenix - I mean recompile it live ?
• Just that peace of functionality, not the whole Phoenix ?

There wouldn’t be any performance issues if done correctly. I have a plug for webhooks that are generated at runtime to which I forward from the phoenix router module.

It works roughly like this

# in the router.ex
scope "/" do
  # ...
  forward "/webhooks", MyAppWeb.Plugs.WebHooks
end

# in plugs/web_hooks.ex
def call(%Plug.Conn{path_info: path_info, params: params} = conn, _opts) do
  if handler = get_webhook_handler(path_info) do
    handler.handle!(params)    
  end
  
  send_resp(conn, :ok, [])
end

Do you have performance issues? If not I’d not worry to much until you have. If you do then what have you tried?

No, I don’t currently have that.

I just remember there was a post by Chris McCord stating that using a wildcard route would lead to performance problems.

Thank you very much for your example, I really appreciate it !

Probably this is something I should use for now, and if there would be any issues, then look for something else.

One more time, thank you everyone for help !

A Plug accepts a Plug.Conn and returns a different Plug.Conn resource.

Part of the Plug.Conn structure is the “path_info”. Your plug could look at the Path Info and decide what function to call, pass he connection to it, and halt the propagation of the connection (so the the router never sees it).

Or, it could return a new Plug.Conn, with different path_info, let it propagate to the router, and the router could use its regular mechanism on the new path_info.