A quick guide to GADTs and why you ain't gonna need them
A quick guide to GADTs and why you ain't gonna need them
I’m no OCaml expert and mostly I’m casually browsing.
The arguments presented read quite compelling. What are your thoughts? Does the conclusion make sense? Have you had any real world experience w/ GADTs that you could share?
I’ve tried something recently in a project with Dream.
Suppose you have an url like /page/param1/param2/ : you want to be able to do three things:
Of course, some pages will have two arguments, some other three, and you have to find a way represent this in the type system.
For this, I’ve used a gadt:
type ('a, 'b) t = | Static : string * ('a, 'b) t -> (unit -> 'a, 'b) t | Int : string * ('a, 'b) t -> (int64 -> 'a, 'b) t | String : string * ('a, 'b) t -> (string -> 'a, 'b) t | End : ('a, 'a) tThe string is the parameter name in the url (id, login, …) and the gadt make the representation for this parameter in the type system. This gives me a way to declare some urls in the application:
val root : (unit -> 'a, 'a) t (** The path to root / *) val topic_url : (string -> 'a, 'a) t (** The path to /:topic *) val thread_url : (string -> int64 -> 'a, 'a) t (** The path to /:topic/:thread *)Then I have some some functions wihch handle this gadt:
val repr : (unit -> ('a, 'b) t) -> string (** Represent the route associated with this path *) val unzip : (unit -> ('a, 'b) t) -> (string -> string) -> 'a -> 'b (** Extract the parameters from the request. [unzip path extract f] will apply the function extract for all the arguments in the path, and gives them to f The function [extract] is given in argument in order to break dependency circle. This should be something like: let extract_param request name = Dream.param request name |> Dream.from_percent_encoded *)handler in your case.