[Haskell-cafe] Re: CAF's in Haskell

[Jonathan Cast]

On 26 Dec 2007, at 12:30 PM, Peter Verswyvelen wrote:

> Well I certainly hope the standard defines that both fred and bob  
> will only
> be evaluated once, because my programs depend on that :)

If your programs depend on lazy evaluation, they can't be Haskell  
98.  Any complete reduction method is sound for Haskell 98.  If  
you're using an extension (such as unsafePerformIO) that isn't  
Haskell 98, check your compiler documentation --- although you'll  
find that there is no guarantee that a CAF is evaluated only once  
(and in fact small CAFs may not be).

jcc