module Lazy_m:sig
..end
Lazy
There are two advantages to reimplementing the lazy module in pure Ocaml. Advantage number one is speed: I've measured ~140 clocks to force a standard lazy value the first time, and ~80 clocks to force it the second time. If the lazy computation you're avoiding is creating a simple cons cell, this is horribly expensive. The following implementation is like ~30 clocks to force the lazy value the first time, and single-digit clocks to force it the second time.
The second one is that we can make lazy values a monad. This is a correctness issue, as a common mistake with laziness is not being lazy enough. This is much easier to get right if you're doing monadic binding.
There are two downsides to doing it this way. One, you can't use
lazy keyword. And two, this implementation uses a little more memory
per lazy value (it currently uses 5 words/lazy value).
type 'a
t
val of_val : 'a -> 'a t
val of_fun : (unit -> 'a) -> 'a t
If the function raises an exception, all futures forces of the
lazy value will also raise the same exception.
val force : 'a t -> 'a
If the function that produces the value throws an exception, this
function will throw the same exception.
val is_val : 'a t -> bool
include Monad