Re: [GHC] #14878: Can't witness transitivity ((.)) of isomorphism of Constraints

5 Mar 2018

      #14878: Can't witness transitivity ((.)) of isomorphism of Constraints
-------------------------------------+-------------------------------------
        Reporter:  Iceland_jack      |                Owner:  (none)
            Type:  bug               |               Status:  new
        Priority:  normal            |            Milestone:
       Component:  Compiler          |              Version:  8.5
      Resolution:                    |             Keywords:
                                     |  QuantifiedConstraints, wipT2893
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):

 I don't think this is a bug. Consider this example:

 {{{#!hs
 {-# LANGUAGE AllowAmbiguousTypes #-}
 {-# LANGUAGE ConstraintKinds #-}
 {-# LANGUAGE GADTs #-}
 {-# LANGUAGE QuantifiedConstraints #-}

 data Dict c where
   Dict :: c => Dict c

 f :: ( c
      , c => d
      )
   => Dict d
 f = Dict
 }}}

 This works, because there is exactly one matching local instance (`c =>
 d`) for `d`. What about this example, which is closer to what you are
 writing?

 {{{#!hs
 {-# LANGUAGE AllowAmbiguousTypes #-}
 {-# LANGUAGE ConstraintKinds #-}
 {-# LANGUAGE GADTs #-}
 {-# LANGUAGE QuantifiedConstraints #-}

 data Dict c where
   Dict :: c => Dict c

 f :: ( a, b
      , a => c, b => c
      , c => d
      )
   => Dict d
 f = Dict
 }}}

 There is one matching local instance (`c => d`) for `d`, so we try to
 deduce `c`. But there are multiple matching local instances for `c`: `a =>
 c` and `b => c`. Which one does GHC pick? As noted in the
 [https://github.com/Gertjan423/ghc-
 proposals/blob/e16828dbcd59d0ca58573c81fc6cea671875e6e2/proposals/0000
 -quantified-constraints.rst#125overlap quantified constraints proposal],
 if GHC is ever in doubt about which local instance to pick, it simply
 rejects the code.

-- 
Ticket URL: http://ghc.haskell.org/trac/ghc/ticket/14878#comment:1
GHC http://www.haskell.org/ghc/
The Glasgow Haskell Compiler