
On Tue, 11 May 2010, Ivan Lazar Miljenovic wrote:
Henning Thielemann
writes: I do not see why there is the need for any type extension, at all. Consider cabal-sort, a very basic program, that is Haskell-98 today, will no longer run in Hugs and JHC (untested so far) because it uses FGL's topological sort when FGL upgraded to associated types.
How should it be able to specify a fixed type value for the vertex type? We can't specify that g :: * -> * -> * -> * because the vertex type should _not_ be a parameter in that sense (since for many graphs it won't be, and we need some way of specifying what it is).
If I understand you right, you say that explicit type parameters for labels are ok, because I can set them to () if not needed, but for an alternative node type you find an explicit type parameter inappropriate?