[GHC] #16208: map/coerce does not fire with all newtypes

21 Jan 2019

      #16208: map/coerce does not fire with all newtypes
-------------------------------------+-------------------------------------
           Reporter:  monoidal       |             Owner:  (none)
               Type:  bug            |            Status:  new
           Priority:  normal         |         Milestone:
          Component:  Compiler       |           Version:  8.6.3
           Keywords:                 |  Operating System:  Unknown/Multiple
       Architecture:                 |   Type of failure:  None/Unknown
  Unknown/Multiple                   |
          Test Case:                 |        Blocked By:
           Blocking:                 |   Related Tickets:
Differential Rev(s):                 |         Wiki Page:
-------------------------------------+-------------------------------------
 Consider a slightly modified version of T2110, compiled with -O:

 {{{
 #!haskell
 {-# LANGUAGE GADTs #-}
 {-# LANGUAGE MagicHash #-}
 {-# LANGUAGE RankNTypes #-}

 import GHC.Exts
 import Unsafe.Coerce

 newtype Age a b where
   Age :: forall a b. Int -> Age a b

 foo :: [Int] -> [Int]
 foo = map id
 fooAge :: [Int] -> [Age a b]
 fooAge = map Age
 fooCoerce :: [Int] -> [Age a b]
 fooCoerce = map coerce
 fooUnsafeCoerce :: [Int] -> [Age a b]
 fooUnsafeCoerce = map unsafeCoerce

 same :: a -> b -> IO ()
 same x y = case reallyUnsafePtrEquality# (unsafeCoerce x) y of
     1# -> putStrLn "yes"
     _  -> putStrLn "no"

 main = do
     let l = [1,2,3]
     same (foo l) l
     same (fooAge l) l
     same (fooCoerce l) l
     same (fooUnsafeCoerce l) l
 }}}

 This code correctly prints "yes" four times, as required by #2110.
 However, changing the order of type arguments in the definition of Age to:

 {{{
   Age :: forall b a. Int -> Age a b
 }}}

 causes the test to fail in one case: `map Age` is no longer simplified to
 `Age`. The reason is that this change causes the newtype `Age` to have a
 wrapper, and the map/coerce rule is not detecting it (see Note [Getting
 the map/coerce RULE to work] and Note [Data con wrappers and GADT syntax])

 This ticket is a prerequisite to linear types (since in linear types, all
 newtypes have wrappers).

-- 
Ticket URL: http://ghc.haskell.org/trac/ghc/ticket/16208
GHC http://www.haskell.org/ghc/
The Glasgow Haskell Compiler