[Haskell-cafe] Typeclass Constraint Causing Typechecker to Fail

[James M]

Below I have a contrived example. Please do not take this to be real world
code.

f :: (Show a) => Int -> (Int -> a) -> Int -> IO a
f i g x = do
    print i
    return $ g (x + i)

foo :: Bool -> IO (Either [Int] [String])
foo b = do
    let helper = f 2
    if b then
        Left  <$> sequence (fmap (helper negate) [0,1])
    else
        Right <$> sequence (fmap (helper show) [0,1])

The above will fail stating

Example.hs:14:35:

    Couldn't match type ‘Int’ with ‘[Char]’

    Expected type: Int -> IO String

      Actual type: Int -> IO Int

    In the first argument of ‘fmap’, namely ‘(helper show)’

    In the first argument of ‘sequence’, namely

      ‘(fmap (helper show) [0, 1])’


Example.hs:14:42:

    Couldn't match type ‘[Char]’ with ‘Int’

    Expected type: Int -> Int

      Actual type: Int -> String

    In the first argument of ‘helper’, namely ‘show’

    In the first argument of ‘fmap’, namely ‘(helper show)’


However, if I simply change f's type signature to not have the typeclass
constraint, the type checker is happy:
f :: Int -> (Int -> a) -> Int -> IO a

Another possibility to remove the problem is to remove the let statement
and instead put the entire expression in.

foo :: Bool -> IO (Either [Int] [String])
foo b = do
    if b then
        Left  <$> sequence (fmap (f 2 negate) [0,1])
    else
        Right <$> sequence (fmap (f 2 show) [0,1])

This seems to be a bug in the typechecker, and maybe it is a well known
issue.

Can someone please confirm that this is a bug and whether or not it is
known?

James
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mail.haskell.org/pipermail/haskell-cafe/attachments/20160419/de77d8e0/attachment.html>