Hi Joachim, As discussed previously, I do think we need a way to store representational coercions. In my roles branch (which is getting close to pushing, I believe), I have created eqReprPrimTyCon, which is like eqPrimTyCon, but it is the type of coercions witnessing representational equality. The Id stored in a CoVarCo can usefully have a type headed by eqReprPrimTyCon. If you're really eager to take a look, I use github to coordinate among computers; you can see the roles branch at github.com/goldfirere/ghc. Richard On Jul 24, 2013, at 8:31 AM, Joachim Breitner wrote:
Hi,
Am Dienstag, den 23.07.2013, 18:58 +0000 schrieb Simon Peyton-Jones:
If you add -XIncoherentInstances *just to the module that has instance IsNT a a*, then it'll work fine I think. This says "pick this instance even though an instantiation of the constraint might match a more specific instance". You don't need the flag in importing modules.
That does not seem to be the case: Enabling the flag just for the special "IsNT a a" instance does not work:
Prelude GHC.NT> :t castNT :: (Either Int a) -> (Either Age a)
<interactive>:1:1: Overlapping instances for IsNT a1 a1 arising from a use of ‛castNT’ Matching instances: instance [incoherent] IsNT a a -- Defined at <interactive>:12:10 instance (IsNT a a', IsNT b b') => IsNT (Either a b) (Either a' b') -- Defined at <interactive>:7:1 (The choice depends on the instantiation of ‛a1’ To pick the first instance above, use -XIncoherentInstances when compiling the other instance declarations) In the expression: castNT :: (Either Int a) -> (Either Age a)
(as correctly specified by the error message.)
But now the solution is easy to see: When deriving IsNT, simply set the incoherent flag for every instance, independent of any active pragmas.
Greetings, Joachim
-- Joachim “nomeata” Breitner mail@joachim-breitner.de • http://www.joachim-breitner.de/ Jabber: nomeata@joachim-breitner.de • GPG-Key: 0x4743206C Debian Developer: nomeata@debian.org
_______________________________________________ ghc-devs mailing list ghc-devs@haskell.org http://www.haskell.org/mailman/listinfo/ghc-devs