I agree that nthFix is mainly useful in proofs. Similarly for paramFix. But if we are having a separate library for "fix"-like functions, I don't see any harm in putting them there as well. The general idea of Control.Function looks good too. I used the name "traceFix" to avoid clash with the "trace" debugging primitive. It sounds too much like a debugging function in any case. -Levent. On Monday 04 March 2002 11:34 am, Ross Paterson wrote:
On Sun, Mar 03, 2002 at 11:38:44PM +0000, Levent Erkok wrote:
On Monday 25 February 2002 10:42 am, Simon Marlow wrote:
This sounds fine to me. It should probably go in Control.Fix. Levent: would you like to suggest a full contents for this library?
Sorry for the late reply. Modulo names, I think the attached library should do it.
I'd suggest that traceFix be renamed trace; it will be used by Control.Arrow. But is anyone ever going to use nthFix or paramFix? The former is only useful in proofs, while the latter is easily available as
paramFix :: ((a, u) -> u) -> a -> u paramFix f a = fix (curry f a)
How about a Control.Function module, with fix, trace and
compose :: [a -> a] -> a -> a compose = foldr (.) id
times :: Integral n => n -> (a -> a) -> a -> a times n f = compose (genericReplicate n f)
so then nthFix n f = times n f undefined. Maybe some of the prelude functions would belong in there too.