On 2006-03-24, Henning Thielemann
On Fri, 24 Mar 2006, Aaron Denney wrote:
Basically, my big objection is that it's hard to define many useful operations on them that are statically safe.
Why not defining the Torsor class you suggested?
Torsor is not quite the right word -- it's just that one of the contexts for non-negative numbers is very similar to one fairly standard example of torsors -- pointers and offsets.
Now granted, the numeric hierarchy should be broken up a bit (hmm, I should finish my strawman proposal for Haskell'), but even then I see problems.
Hm, is there something going on?
A strawman proposal, not yet posted anywhere.
Without breaking compatibility? But class instances become invalid if the hierarchy is modified.
No, compatibility will be broken. Hopefully not for most uses -- I don't think most people define new instances, and those that do will be able to do so more reasonably, so hopefully wouldn't mind.
If there is some progress towards a refined numeric class hierarchy I want to point again to http://cvs.haskell.org/darcs/numericprelude/ http://cvs.haskell.org/darcs/numericprelude/src/Algebra/Core.lhs I hope I don't annoy you. :-)
Not at all. That is one of the things I looked at a while ago, that has inspired a lot of my decisions -- but I'm more willing to rename things that I think have silly names. And there are a few minor details, like allowing only for euclidean domains rather than principal ideal domains. I will probably actually put two proposals up, with one allowing more generality via MPTCs and FDs (which I truly hope make it into the standard). -- Aaron Denney -><-