On Sat, Feb 19, 2011 at 4:49 PM, wren ng thornton <wren@freegeek.org> wrote:

How does, or will, this package differ from Milan Straka's http://hackage.haskell.org/package/hashmap ?

It's a continuation of that work. The unordered-containers implementation is more space efficient and a bit faster. Milan also uses a Patricia trie in his implementation. I plan to switch to a hash-array mapped trie (HAMT) in the future. It's about 25% faster for lookups but at the moment quite a bit slower for inserts. Once we got some new faster array copy primops in GHC we can hopefully improve the insert time a lot. A HAMT also has much lower space overhead per entry. Johan