Module Base_quickcheck__.Generator

Generates random functions that use the given observer to perturb the pseudo-random state that is then used to generate the output value. The resulting functions are therefore deterministic, assuming the observer is deterministic.

Produces any of the given values, weighted uniformly.

Chooses among the given generators, weighted uniformly; then chooses a value from that generator.

Returns the current size parameter.

Produces a generator that ignores the size parameter passed in by Base_quickcheck and instead uses the given ~size argument. Most often used with size to reduce the size when dispatching to generators for subparts of a value.

For example, here is a use of with_size and size to create a generator for optional lists. We are careful to generate None even at non-zero sizes; see the note above about not using size as a lower bound.

let optional_list generator =
  let open Let_syntax in
  match%bind both size bool with
  | (0, _) | (_, false) -> return None
  | k, _ ->
    let%map elements = with_size ~size:(k-1) (list generator) in
    Some elements

Produces a list of sizes that distribute the current size among list elements. The min_length and max_length parameters can be used to bound the length of the result.

This is the distribution used by generators such as list to divide up size among elements.

This function is designed so that elements of list are always generated at strictly smaller size than the list itself. The technical invariant is: if size_list is generated by with_size ~size:n (sizes ~min_length ()), then:

(List.length size_list - min_length) + (List.sum (module Int) size_list)
<= n

Produces values for which f returns true. If f returns false, retries with size incremented by 1. This avoids filter getting stuck if all values at a given size fail f; see the note above about not using size as a lower bound.

When f produces Some x, produces x. If f returns None, retries with size incremented by 1, as with filter.

Ties the recursive knot to produce generators for recursive types that have multiple clauses, separating base cases from recursive cases. At size 0, only base cases are produced; at size n > 0, the base cases are produced at size n along with the recursive cases at size n-1. Raises if the list of base cases is empty or if the list of recursive cases is empty.

For example, here is a use of recursive_union to create a generator for an expression datatype.

type exp =
  | Int  of int
  | Bool of bool
  | If   of exp * exp * exp
  | Add  of exp * exp

let exp_generator =
  recursive_union
    [
      map int ~f:(fun i -> Int i);
      map bool ~f:(fun b -> Bool b);
    ]
    ~f:(fun exp ->
      let open Let_syntax in
      [
        (let%map a = exp and b = exp and c = exp in If (a, b, c));
        (let%map a = exp and b = exp in Add (a, b));
      ])

Like recursive_union, without separate clauses or automatic size management. Useful for generating recursive types that don't fit the clause structure of recursive_union.

For example, here is a use of fixed_point to create a generator for N-ary trees. No manual size management is needed, as Generator.list guarantees to generate list elements at strictly smaller sizes than the list itself.

type tree = Node of tree list
let tree_generator =
  fixed_point (fun tree ->
    map (list tree) ~f:(fun trees -> Node trees))

Creates a t that forces the lazy argument as necessary. Can be used to tie (mutually) recursive knots.

Produces one of the given values, chosen with the corresponding weight. Weights must be non-negative and must have a strictly positive sum.

Produces one of the given generators, chosen with the corresponding weight, then chooses a value from that generator. Weights must be non-negative and must have a strictly positive sum.

Like recursive_union, with explicit weights for each clause. Weights must be non-negative and the recursive case weights must have a strictly positive sum.

Passes in additional "salt" used to perturb the pseudo-random state used to generate random values. Generators' output is intended to be deterministic for any initial pseudorandom state, so perturb can be used to generate a new generator with the same distribution that nonetheless produces different values from the original for any given pseudo-random state.

Creates a generator that calls the given function with the current size parameter and pseudorandom state.

Generates a random value using the given size and pseudorandom state. Useful when using create and dispatching to other existing generators.