Re: [Caml-list] GADT in an optional parameter

[Alain Frisch <alain@frisch.fr>]

On 07/25/2013 11:31 AM, David Allsopp wrote:
> Alain Frisch wrote:
>> The idea would be to have a different kind of optional argument, which
>> specifies as part of the function type an expression to be expanded
>> syntactically on the call site when the argument is not provided
>> explicitly.  (In your example, the Exn constructor would need to be
>> available from the call site.  -- Or do you interpret "Exn" as a typed
>> reference to the Exn constructor on the value declaration site?)
>>
>> If we assume that the "default" can be an arbitrary "parse tree"
>> (untyped) expression, I guess that type equality would require these
>> expressions to be strictly equal (modulo locations).
>
> That seems very powerful, but presumably similarly hard? I'm guessing that at the moment the .cmi file doesn't contain any indication of whether an optional argument has a default or not (because it's an implementation detail transformed in the code as Oleg explained). Would that be a reasonably easy change to make? It feels very weird suddenly needing code to be included in a type! Or is this is a better halfway house...

Indedd, it requires to store (untyped) parsetree fragments in types, 
which I'm sure will be considered as an abomination by most.  But from a 
technical point of view, I don't see any difficulty.

Alain