Opened 19 months ago

Closed 18 months ago

Last modified 16 months ago

#14897 closed bug (invalid)

QuantifiedConstraints: Can't print type of quantified constraint

Reported by: Iceland_jack Owned by:
Priority: normal Milestone:
Component: Compiler Version: 8.5
Keywords: QuantifiedConstraints Cc:
Operating System: Unknown/Multiple Architecture: Unknown/Multiple
Type of failure: None/Unknown Test Case:
Blocked By: Blocking:
Related Tickets: Differential Rev(s):
Wiki Page:

Description

{-# Language QuantifiedConstraints, FlexibleInstances, UndecidableInstances, MonoLocalBinds #-}

class    (forall xx. Functor (f xx)) => Functor' f
instance (forall xx. Functor (f xx)) => Functor' f

fmap' :: Functor' f => (b -> b') -> (f a b -> f a b')
fmap' = fmap

load in ghci and check the type of fmap'

$ ... -ignore-dot-ghci Bug2.hs 
GHCi, version 8.5.20180128: http://www.haskell.org/ghc/  :? for help
[1 of 1] Compiling Main             ( Bug2.hs, interpreted )
Ok, one module loaded.
*Main> :t fmap'

<interactive>:1:1: error:
    No instance for (Functor (f xx)) arising from a use of ‘fmap'’
*Main> 

Simpler example

GHCi, version 8.5.20180128: http://www.haskell.org/ghc/  :? for help
Prelude> :set -XRankNTypes -XQuantifiedConstraints 
Prelude> let a :: (forall xx. Monoid (f xx)) => f a; a = mempty
Prelude> :t a

<interactive>:1:1: error:
    No instance for (Monoid (f xx)) arising from a use of ‘a’
Prelude>

I expected the same output as :t +v

Prelude> :t +v a
a :: (forall xx. Monoid (f xx)) => f a

Change History (3)

comment:1 Changed 19 months ago by Iceland_jack

I know printing isn't the problem, “over-eagerly resolves quantified constraint...” something

comment:2 Changed 18 months ago by simonpj

Resolution: invalid
Status: newclosed

What :t e does is to infer the type of the expression e and print it.

So with a :: (forall xx. Monoid (f xx)) => f a, it's as if you'd typed the top level decl

foo = a

What happens? We instantiate the type of a to get a Wanted constraint

[W] forall x. Monoid (f1 x)

and the type f1 a1, where f1 and a1 are unification variables. Now we try to simplify the constraints and infer the most general type for foo. Unsurprisingly, we fail with the reported error message.

If, rather than inferring the type of an arbitrary expression, you want to ask for the type (and other info) about an identifier, use :info a.

comment:3 Changed 16 months ago by RyanGlScott

Keywords: wipT2893 removed
Note: See TracTickets for help on using tickets.