overlapping instances and constraints

[Bulat Ziganshin]

Hello John,

Tuesday, February 28, 2006, 4:23:24 AM, you wrote:

>> i had plans to propose the same and even more:
>> 
>> instance C2 a b | a/=b

JM> I was thinking it would be all kinds of useful if we had two predefined
JM> classes

JM> class Eq a b
JM> class NEq a b

JM> where Eq has instances exactly when its two types are equal and NEq has
JM> instances exactly when its two types are not equal.

JM> Eq should be straightforward to implement, declaring any type
JM> automatically creates its instances. (sort of an auto-deriving). NEq
JM> might be more problematic as that would involve a quadratic number of
JM> instances so its implementation might need to be more special. but
JM> perhaps we can do with just 'Eq'.

with 'Eq' class we can't do anything that is impossible without it
:)))

the whole devil is to make general instance NON-OVERLAPPING with
specific one by EXPLICITLY specifying EXCLUSIONS with these "/=" rules:

class Convert a b where
  cvt :: a->b

instance Convert a a where  -- are we need Eq here? :)
  cvt = id

instance (NEq a b) => Convert a b where
  cvt = read.show



... yes, i recalled! my proposal was to allow "!" in instance headers:

instance C Int where ...
instance (!Int a, Integral a) => C a where ...
instance (!Integral a, Enum a) => C a where ...

adding your Eq class, it will be all we can do on this way

interesting, that the language theoretics can say about decidability,
soundness, and so on of this trick? :)

-- 
Best regards,
 Bulat                            mailto:Bulat.Ziganshin at gmail.com