I'm currently porting a Numpy project into haskell as a self-improvement project, and it's been 69% great. But then I find some library that I have to implement from scratch when I just had it in python. That's a pretty big problem, and I'm willing to put up with it, but most people aren't.