Re: [GHC] #8537: confusing error message for specialize pragma

22 Nov 2013

      #8537: confusing error message for specialize pragma
-------------------------------------+------------------------------------
        Reporter:  carter            |            Owner:
            Type:  bug               |           Status:  new
        Priority:  normal            |        Milestone:
       Component:  Compiler          |          Version:  7.6.3
      Resolution:                    |         Keywords:
Operating System:  Unknown/Multiple  |     Architecture:  Unknown/Multiple
 Type of failure:  None/Unknown      |       Difficulty:  Unknown
       Test Case:                    |       Blocked By:
        Blocking:                    |  Related Tickets:
-------------------------------------+------------------------------------

Old description:
...
I get the following confusing error/warning about SPECIALIZE
{{{
SPECIALISE pragma for non-overloaded function `$cfmap'
}}}
with the following example module
{{{
{-# LANGUAGE PolyKinds   #-}
{-# LANGUAGE BangPatterns #-}
{-# LANGUAGE DataKinds #-}
{-# LANGUAGE TypeFamilies #-}
{-# LANGUAGE GADTs #-}
{-# LANGUAGE DeriveDataTypeable #-}
{-# LANGUAGE TypeOperators #-}
{-# LANGUAGE FlexibleInstances #-}
{-# LANGUAGE FlexibleContexts #-}
{-# LANGUAGE FunctionalDependencies #-}
{-# LANGUAGE UndecidableInstances #-}
{-# LANGUAGE ScopedTypeVariables #-}
{-# LANGUAGE StandaloneDeriving #-}
{-# LANGUAGE CPP #-}
{-# LANGUAGE TemplateHaskell #-}
{-# LANGUAGE NoImplicitPrelude #-}
module Numerical.Types.Shape where
import Control.Applicative
import Data.Foldable
import Data.Monoid
import Data.Functor
import Prelude (seq, ($!),($),Show(..),Eq(),Int)

...
data Nat = S !Nat  | Z
    --deriving (Eq,Show,Read)

...
{-
not doing the  HLIST style shape because I don't want to have
any pattern matchings going on.
Also would play hell with locality quality in the address translation
hackery,
because there'd be an extra load to get those ints!
-}
infixr 3 :*
{-
the concern basically boils down to "will it specialize / inline well"
-}
data Shape (rank :: Nat) a where
    Nil  :: Shape Z a
    (:*) ::  !(a) -> !(Shape r a ) -> Shape  (S r) a
--deriving instance (Show (Shape rank a))
-- deriving instance Eq (Shape rank a)

...
-- #if defined( __GLASGOW_HASKELL__ ) &&  ( __GLASGOW_HASKELL__  >=
707)
    --deriving instance Typeable (Shape rank a)
    -- #endif

...
instance Functor (Shape Z) where
fmap  = \ f Nil -> Nil
    {-# INLINABLE fmap #-}
{-# SPECIALIZE fmap :: (Int ->Int )-> (Shape Z Int)-> (Shape Z Int)
#-}
}}}
New description:

 I get the following confusing error/warning about SPECIALIZE
 {{{
   SPECIALISE pragma for non-overloaded function `$cfmap'
 }}}

 with the following example module

 {{{
 {-# LANGUAGE PolyKinds   #-}
 {-# LANGUAGE BangPatterns #-}
 {-# LANGUAGE DataKinds #-}
 {-# LANGUAGE TypeFamilies #-}
 {-# LANGUAGE GADTs #-}
 {-# LANGUAGE DeriveDataTypeable #-}
 {-# LANGUAGE TypeOperators #-}
 {-# LANGUAGE FlexibleInstances #-}
 {-# LANGUAGE FlexibleContexts #-}
 {-# LANGUAGE FunctionalDependencies #-}
 {-# LANGUAGE UndecidableInstances #-}
 {-# LANGUAGE ScopedTypeVariables #-}
 {-# LANGUAGE StandaloneDeriving #-}
 {-# LANGUAGE CPP #-}
 {-# LANGUAGE TemplateHaskell #-}
 {-# LANGUAGE NoImplicitPrelude #-}

 module Numerical.Types.Shape where

 import Control.Applicative
 import Data.Foldable
 import Data.Monoid
 import Data.Functor
 import Prelude (seq, ($!),($),Show(..),Eq(),Int)

 data Nat = S !Nat  | Z

 infixr 3 :*

 data Shape (rank :: Nat) a where
     Nil  :: Shape Z a
     (:*) ::  !(a) -> !(Shape r a ) -> Shape  (S r) a

 instance Functor (Shape Z) where

     fmap  = \ f Nil -> Nil
     {-# INLINABLE fmap #-}

     {-# SPECIALIZE fmap :: (Int ->Int )-> (Shape Z Int)-> (Shape Z Int)
 #-}

 }}}

--

Comment (by simonpj):

 The `$cfmap` part is not very clever, but the actual message is fine isn't
 it?  There isn't any useful specialisation we can do on `fmap` at type
 `(Shape Z)` is there?  I suppose you might say that the pattern match is
 exhaustive so we could omit the test for `Nil` vs `(:*)`, but that's all.

 Simon

--
Ticket URL: http://ghc.haskell.org/trac/ghc/ticket/8537#comment:1
GHC http://www.haskell.org/ghc/
The Glasgow Haskell Compiler