proposal: add 'unsafeCoerce'

[Simon Marlow]

Ross Paterson wrote:
> On Mon, Nov 20, 2006 at 04:51:29PM +1100, Donald Bruce Stewart wrote:
> 
>>So the uses fall into 2 categories:
>>
>>  * FFI binding and raw pointer/foreign data manipulation
>>  * type equalities known, but not expressible statically
>>
>>Now, can we say something portable about these uses?
> 
> The second group, perhaps, but the first group (unsafeCoerce#) depend
> on GHC's internal representations.

I'd like to have a precise (sound, if not complete) description of when it's 
safe to use unsafeCoerce in GHC, but it needs some careful thought.  I *think* 
the following cases are guaranteed to work:

  * cast that changes a phantom type, or changes a type that is not
    reflected by a part of the value,
    eg. 'unsafeCoerce (Left 3) :: Either Int a' should be fine for any 'a',

  * casting a polymorphic type to the actual type of the runtime value.
    That is, you can safely cast a value to its correct type.  (eg. in
    Typeable.cast).

  * casting an unboxed type to another unboxed type of the same size.

The difficult cases arise when you want to cast a value to a type other than its 
real type, such as you need to do in the implementation of Dynamic for example. 
  I doubt that we could provide a portable implementation of Dynamic.

Casting boxed types to unboxed types or vice versa is pretty much guaranteed to 
end in tears.

Cheers,
	Simon