Re: type and subtype

["Dmitry A. Kazakov" <mailbox@dmitry-kazakov.de>]

On Sun, 24 Jul 2005 16:40:26 GMT, Jeffrey Carter wrote:

> Ludovic Brenta wrote:
>> 
>> Two subtypes of the same type are compatible with one another.  One
>> can convert between them using an implicit type conversion (i.e. no
>> special syntax required).
> 
> There is no type conversion between subtypes of the same type, since 
> they are the same type. There are checks that the value of the source 
> matches the constraints of the destination.

This sounds much like hairsplitting. However it is an important issue for
the future.

I think that Ludovic's view is more consistent: checks are just a part of
the conversion which otherwise is an identity function, because *presently*
the representation is required to be same. But representation is an
implementation detail and thus cannot count. Moreover it is very desirable
to allow subtypes having different representations. For example: statically
constrained subtypes with removed constant discriminants from their values;
T'Class constrained to some specific type S could remove the type tag from
its objects, etc. Such subtypes would require more-than-checks conversions.

-- 
Regards,
Dmitry A. Kazakov
http://www.dmitry-kazakov.de