On Mon, 17 Jul 2006 10:58:48 +1200, Vivian McPhail wrote: > Haskell is often cited as an elegant and 'mathematical' language in which to > program. This post is to voice my support for a modified Numeric Prelude > that would be consistent with [1] and/or [2]. It would be an improvement if > the numeric classes were more structured along the lines of algebraic > properties (similar to the monad laws) of various entities. > > I remember griping about this when I first learned Haskell. > > While the typical user may not need to know about the underlying > mathematical structure of the objects that they are working with, it would > be good if the system were consistent for those that do. > > [1] http://haskell.org/docon/ > [2] http://cvs.haskell.org/darcs/numericprelude/
The numeric prelude package (though the documentation is hard to follow with all the classes called C and types called T) seems to consist of two layers: - a bunch of unary classes refactoring the Haskell 98 numeric class hierarchy, and - multi-parameter classes, building on the former classes, for vector spaces and related concepts. A refactoring like the former would be useful on its own, and would provide a base for adding the other classes later. Of course the algebraic properties won't be satisfied by floating point arithmetic. _______________________________________________ Haskell-prime mailing list [email protected] http://www.haskell.org//mailman/listinfo/haskell-prime
