This is mostly educational. The BDDs do not share nodes and this might introduce inefficiencies.

An important (for me, in teaching) feature is that I can immediately draw the BDD to an X11 window (via graphviz). For example, to show the effect of different variable orderings, try this in ghci:

import qualified Prelude as P
import OBDD
let f [] = false; f (x:y:zs) = x && y || f zs
display P.$ f P.$ P.map variable [1,2,3,4,5,6]
display P.$ f P.$ P.map variable [1,4,2,5,3,6]

If you want better performance, use CUDD Haskell bindings (low level, high level).