Please, do you know any tutorial on implementing APIs with Cowboy or Plug?
I recommend taking a look at the readme for Plug. The Plug.Router section shows how to create route handlers for your API – it’s incredibly simple. If you have an application large enough to need to keep the business logic out of the routes (basically any non-trivial app), just create a models directory, put that logic in there, and use the route handlers similar to how you would use a controller in an MVC framework: control access, validate parameters, make calls to models.
One trick I like to do that’s not in the Plug docs is to create helpers for the HTTP responses:
defmodule Routes.Helpers do
import Plug.Conn
def ok!(conn, body) do
conn
|> send_resp(200, body)
|> halt
end
def not_found!(conn) do
conn |> error!(404, "Not Found")
end
end
That way you can respond to a route using code that’s a little easier to read in my opinion. It’s also a good place to handle JSON encoding if your API produces JSON responses (not shown here).
get "/hello" do
conn |> ok!("world")
end
match _ do
conn |> not_found!
end
That’s just my personal preference though. Do what makes sense for your project.
Regarding Cowboy, there’s a pretty extensive Cowboy User Guide. Just keep in mind that Cowboy is implemented in Erlang, so you’ll have to do some mental conversion from Erlang to Elixir when interfacing with it. It’s not too difficult, but you may want to be comfortable with the basics of Elixir first. Plug is probably going to be the easier of the two to get started with. Plug uses Cowboy under the hood in its typical configuration (and Phoenix uses Plug). An advantage to building directly on Cowboy is if you’re concerned about keeping your dependencies to a bare minimum (security requirements, maintainability, simplicity, etc.).
