Hoopl (data flow analysis) would also make a difference. I did a very similar project at my university in Haskell and Hoopl definitely saved us from writing quite a bit of code. We also used parser combinators in the frontend, which I think saved us time too.