Re: [Haskell-cafe] Typeclass constraints not inferred when defining a function

Hi, It’s because the monomorphism restriction turned off by default in GHCi sessions. If you put this at the beginning of your source file, it should work: {-# LANGUAGE NoMonomorphismRestriction #-} You can read up about it here: https://wiki.haskell.org/Monomorphism_restriction — Csongor

On 9 Feb 2017, at 18:20, Stephen Demos wrote:

Hi all,

A professor showed me a function today that looks like this -

`append = flip (foldr (:))`

GHC spits back an error when this is compiled -

``` • Ambiguous type variable ‘t0’ arising from a use of ‘foldr’ prevents the constraint ‘(Foldable t0)’ from being solved. Relevant bindings include append :: t0 a -> [a] -> [a] (bound at test.hs:4:1) Probable fix: use a type annotation to specify what ‘t0’ should be. ```

It's obvious from this error message the problem is that the type GHC infers doesn't include the typeclass constraint. In this case, it's Foldable, but you can easily think of other situations where the same error comes up with different typeclasses (a silly example that comes to mind is `fmap (const [])`). It's pretty easy to fix this error by just providing a type for the function that is more specific than the one GHC infers.

However, he showed me that defining the same function in ghci using a let binding doesn't yield the same error. In fact, when using `:t`, it shows that ghci successfully infers the typeclass constraint on the function -

```

...

:t flip (foldr (:)) flip (foldr (:)) :: Foldable t => t a -> [a] -> [a]

I was wondering if anyone could explain this apparent discrepancy to me.

Thanks!
Stephen Demos
_______________________________________________
Haskell-Cafe mailing list
To (un)subscribe, modify options or view archives go to:
http://mail.haskell.org/cgi-bin/mailman/listinfo/haskell-cafe
Only members subscribed via the mailman list are allowed to post.