Step 3: Grading with generators for Ocaml simple built-in types

You can find the examples below in the exercises/sampler-built-in-types directory (branch: step-3).

As see previously, you can either give manually inputs for the tested functions or you can ask the grader to automatically generate inputs.

For simple built-in types, the grader actually do most of the work for you: you only need to precise the number of inputs sets you want to be generated and the grader will use predefined sampler.

Which types have a predefined sampler ?

Before beginning, we should make an important warning : there is no default sampler for functional types, tuples (or type composed of tuple) for example. Here is a list of the signatures of predefined samplers :

  val sample_int : int sampler
  val sample_float : float sampler
  val sample_list : ?min_size: int -> ?max_size: int -> ?dups: bool -> ?sorted: bool -> 'a sampler -> 'a list sampler
  val sample_array : ?min_size: int -> ?max_size: int -> ?dups: bool -> ?sorted: bool -> 'a sampler -> 'a array sampler
  val sample_option : 'a sampler -> 'a option sampler
  val sample_string : string sampler
  val sample_char : char sampler
  val sample_bool : bool sampler

How to use a default sampler

When the function to test has inputs of type with predefined sample, you have nothing to do. In the example below, five tests are automatically generated.

let exercise_1 =
    [%ty: int -> int] "identity"
    ~gen:5 [0]

However, with this method, you have no control on how the inputs are generated. If, for example, you want a function of type int -> int -> int to be tested for inputs between 12 and 42, you need to give the grade function the sampler you want the same way you will to for type without predefined sampler. There are actually two ways do to that:

Method 1 : using the ~sampler argument

The more general way do provide a sampler is to use the optional argument ~sampler that has type:

unit -> <arg1 type> * <arg2 type> * <arg3 type> etc..

let exercise_2 =
    [%ty: int -> int -> int] "pi1"
    ~sampler:(fun () -> (Random.int 31 + 12, Random.int 31 + 12) )

Method 2 : redefining the corresponding sampling function.

Another way is to define a sampling function of type unit -> <arg1 type> * <arg2 type> * <arg3 type> etc. using the naming convention : sample_<type>. In this case, nothing needs to be add to the grade function call.

let sample_int = Random.int 31 + 12

let exercise_3 =
    [%ty: int -> int -> int] "pi1"

More avanced examples

You can find the examples below in the exercises/advanced-examples-step-3 directory (branch: step-3).

There is nothing new to learn in this part, there are only more examples of how to build a sampler for more complexed types. In particular, there are examples with:

let sampler_4 () = let sampler_tuple () = (sample_int (), sample_int ()) in (sample_option sampler_tuple) ()

let exercise_4 = test_function_1_against_solution [%ty: (int * int) option -> int] “opt_add” ~gen:5 ~sampler:sampler_4 []

* functional type

let sampler_5 () =
  let sampler_f () = match Random.int 3 with
    | 0 -> succ
    | 1 -> pred
    | _ -> fun _ -> 0 in
  sampler_f (), sample_int ()

let exercise_5 =
    [%ty: (int -> int) -> int -> int] "apply"

let sampler_6 =
  sample_array ~min_size:1 ~max_size:10 sample_int

let exercise_6 =
    [%ty: int array -> int list] "array_to_list"