I wrote a 3D mesh reconstruction algorithm using feature point extraction in Haskell. The hardest part was graph algorithms since those can be pretty tricky to

I wrote a 3D mesh reconstruction algorithm using feature point extraction in Haskell. The hardest part was graph algorithms since those can be pretty tricky to express in functional languages. But Even with 3D output to OpenGL, image loading and manipulation, etc.. I never touched a monad transformer. I think you are trying too hard. Just embrace IO for the effects and move everything else to pure code.