[Haskell-cafe] Re: Haskell Cheat Sheet?

[ChrisK]

I disagree -- see below

Dan Weston wrote:

> I suggest that it be removed and the real Control.Monad.Fix.fix function
> be defined in its own section, with an side-by-side comparison with a
> named recursive function. This would be useful because the type
> 
> fix :: (a -> a) -> a
> 
> is highly confusing, suggesting to newcomers a usage like:
> 
> f = fix (+1)
> 
> which is undefined (and seems to be "missing an argument"), when
> invariably its type is in practice restricted to:
> 
> fix :: ((a -> b) -> (a -> b)) -> (a -> b)
> 
> which is much more suggestive (but nowhere to be found in the docs).
> 
> Dan Weston
> 

Useful counterexample:

-- import Data.Function(fix)

fix :: (t -> t) -> t
fix f = let f' = f f' in f'

one :: [Int] -> [Int]
one = (1:)

ones = [Int]
ones = fix one

This emphasizes that Haskell's fix can define non-function types, which the Y
combinator in strict languages is not useful for.  (Unless someone would like to
correct me...)

-- 
Chris