On 20 Apr 2009, at 10:57, Peter Verswyvelen wrote:

On Mon, Apr 20, 2009 at 10:23 AM, Thomas Davie <tom.davie@gmail.com> wrote:

Really? Is it any less referentially transparent than unamb already is - i.e. it's referentially transparent, as long as the two values really are equal.

I think it is. Suppose we call the function hnf :: a -> Bool. hnf might return a different result for the same argument, since the evaluation of the argument might be happening on a different thread, so the result of hnf depends on the time when it is evaluated. Or am I missing something here?

Sure, so hnf would give us a non-determined result, but I don't think that makes unamb any less referentially transparent – the same value is always returned, and always reduced at least to hnf.