Re: [Caml-list] Difference between [ `A ] and [< `A ]

[Jacques Garrigue <garrigue@math.nagoya-u.ac.jp>]

On 2011/05/11, at 1:46, Dario Teixeira wrote:

> Hi,
> 
> I've seen OCaml code "in the wild" where both of the following signatures
> are present: (the type parameter for 't' is a phantom type)
> 
>  val foo: [< `A ] t -> unit
>  val bar: [ `A ] t -> unit
> 
> But is there any practical difference between [ `A ] and [< `A ] given
> that there is only one element in the set?

In this particular case the two types are almost equivalent.
The only counterexample I could find is unifying with the following
private row type:

  type leA = private [< `A]

leA is unifiable with [< `A] but not with [`A].

The difference becomes more significant when there is an argument.
For instance, [<  `A of int] and [< `A of bool] are unifiable, giving [< `A of int & bool],
but [`A of int] cannot be unified with [`A of bool].

Note that some old versions of OCaml did some "singleton promotion", i.e.
[< `A of int] was automatically converted to [`A of int].
This was removed as an unnecessary complication, and also because you
might actually want to distinguish the two for private row types.

Jacques Garrigue