[Haskell-cafe] new recursive do notation (ghc 6.12.x) spoils layout

[Dan Doel]

On Sunday 20 June 2010 9:24:54 pm Alexander Solla wrote:
> Why can't you just use let notation do deal with the recursion?  I
> thought "lets" in do blocks were just a little bit of syntactic sugar
> for "regular" let expressions, which do allow mutual recursion.  I
> could be totally wrong though.  I'm thinking of something like:
> 
> do a <- getChar
>     let b = c >>= return . f
>     let c = b >>= return . g
>     c >>= putChar
>     b

This is not what recursive do does. For instance:

  do rec l <- (:l) <$> getChar
     return l

will read one character, and return an infinite list with that character as 
the elements. By contrast:

  let l = (:) <$> getChar <*> l in l

will, for one, never complete, and two, call getChar an infinite number of 
times.

In general, it is based on the fixed point:

  mfix :: (MonadFix m) => (a -> m a) -> m a

where the function given may not be separable into a pure function and return 
at all, which is what would be necessary to implement using let, via the 
identity:

  mfix (return . f) = return (fix f)

The implementation of mfix is specific to the 'm' in question, not something 
you can write for all monads.

-- Dan