Re: [GHC] #14070: Allow ‘unsafe’ deriving strategy, deriving code with ‘unsafeCoerce’

1 Aug 2017

      #14070: Allow ‘unsafe’ deriving strategy, deriving code with ‘unsafeCoerce’
-------------------------------------+-------------------------------------
        Reporter:  Iceland_jack      |                Owner:  (none)
            Type:  feature request   |               Status:  new
        Priority:  normal            |            Milestone:
       Component:  Compiler          |              Version:  8.0.1
      Resolution:                    |             Keywords:
                                     |  QuantifiedContexts, deriving
Operating System:  Unknown/Multiple  |         Architecture:
                                     |  Unknown/Multiple
 Type of failure:  None/Unknown      |            Test Case:
      Blocked By:                    |             Blocking:
 Related Tickets:                    |  Differential Rev(s):
       Wiki Page:                    |
-------------------------------------+-------------------------------------

Comment (by RyanGlScott):

 GHC can "coerce" that just fine (roles notwithstanding) if you do it the
 way `GeneralizedNewtypeDeriving` generates code (pretend I'm using
 `coerce` instead of `unsafeCoerce`):

 {{{#!hs
 {-# LANGUAGE ImpredicativeTypes #-}
 {-# LANGUAGE InstanceSigs #-}
 {-# LANGUAGE RankNTypes #-}
 {-# LANGUAGE ScopedTypeVariables #-}
 {-# LANGUAGE TypeApplications #-}

 import Unsafe.Coerce

 type Lens s a = forall f. Functor f => (a -> f a) -> (s -> f s)

 class X f where
   x :: Lens (f a) a

 newtype WrappedX f a = WrapX (f a)

 instance X t => X (WrappedX t) where
   x = unsafeCoerce @(forall a. Lens (t a) a)
                    @(forall a. Lens (WrappedX t a) a)
                    x
 }}}

 I put "coerce" in quotes because we're not actually coercing the `f`.
 Rather, the two occurrences of `f` tucked underneath the `Lens`es get
 unified. But you were writing this instance in a somewhat unorthodox way
 where you manually applied `@t` and `@a` to `x`, but neglected `f` (`f`
 isn't in scope the way you wrote it, but it's hiding underneath the
 `Lens`). Because of this, GHC was unable to conclude that the `f` in `x'`
 and the `f` in the instance signature were the same.

 Happily, writing instances the way `GeneralizedNewtypeDeriving` does (by
 visibly applying the full type signatures to `coerce`) doesn't suffer from
 this problem.

-- 
Ticket URL: http://ghc.haskell.org/trac/ghc/ticket/14070#comment:11
GHC http://www.haskell.org/ghc/
The Glasgow Haskell Compiler