This is about the same as the other paper I linked to, I think, but I'm interested in actually connecting the fully general trie construction to Template Haskell and other facilities to reduce the coder's overhead of using these tries in practice to a minimum.
Louis Wasserman
wasserman.louis@gmail.com
The goal is similar, but I'm attempting to automatically infer the appropriate map type for any algebraic datatype -- and while I'm at it, the TrieMap package aims to include all the methods Data.Map offers.
Perhaps you want to read about the type-indexed tries in "Generic Haskell: Applications": http://www.cs.uu.nl/~johanj/publications/ghpractice.pdf
Sean