Re: [Caml-list] On variants, integers and testing

[Cedric Cellier <rixed@happyleptic.org>]

-[ Fri, Mar 16, 2018 at 12:12:56PM +0100, Bahman Movaqar ]----
> However, when it comes to testing, things are not that simple. Now that I
> have exported only 1 function in the .mli, I can only test that 1 function.
> Is this the proper way[2]?

This is one of the reasons some people don't like to use mli files.

Another approach is to define (and document) the signatures in ".ml" file
(named for instance "yadayada_intf.ml") and then the implementation can be
written and exposed fully, knowing that you can create another one and
constraint it to the signature later on. Example:

--- yadayada_intf.ml --- 8< --- 8< ---
(** This is library yadayada *)

module type S =
sig
  (** The type of yadayadas *)
  type t

  (** Create a yadayada from a xyz: *)
  val create : xyz -> t

  (** Call this to perform a foo with your yadayada created with [create]: *)
  val foo : t -> wooz
end
--- >8 --- >8 ---

Then you implement the actual thing in yadayada_impl.ml:

--- yadayada_impl.ml --- 8< --- 8< ---
(* Implementation for yadayada *)

type t = ...
let create x = ...
let foo t = ...
--- >8 --- >8 ---

That implementation you can test from "yadayada_tests.ml" for instance, since
none of it is hidden.

Then, you can provide users with a simplified module Yadayada.M, defined in
"yadayada.ml":

--- yadayada.ml --- 8< --- 8< ---
module M = Yadayada_impl : Yadayada_intf.S
--- >8 --- >8 ---

Plenty of variations are possible of course, but the names "t" for the main
type, "S" for the main signature and "M" for the main module (or "Make" if it's
a functor) are common.

The drawback of this technique is that it tends to lead toward a design with
plenty of modules and functors, which makes reading the code more complex (and
is also harder to inline efficiently).

Note that for inline tests you have the qtest
(https://github.com/vincent-hugot/qtest) tool that I find quite nice for inline
tests as long as you do not try to shoehorn it into the large functional test
framework it is not designed to be.


-- 
Caml-list mailing list.  Subscription management and archives:
https://sympa.inria.fr/sympa/arc/caml-list
Beginner's list: http://groups.yahoo.com/group/ocaml_beginners
Bug reports: http://caml.inria.fr/bin/caml-bugs