> For now, this has nothing to do with lazy evaluation. Such implementation will be slow in every language. It happens because add doesn’t use tail-call recursion.
This is why I'll never really trust Haskell to be usable in a serious context. Haskellers will talk your ear off about how such restricted abstractions enable the compiler to optimize it to hell and back, and perfectly idiomatic Haskell can be compiled to almost-perfect machine code, but in point of fact GHC basically never does, and you still have to care about things like tail calls and design APIs to do things like continuation passing.
This is why I'll never really trust Haskell to be usable in a serious context. Haskellers will talk your ear off about how such restricted abstractions enable the compiler to optimize it to hell and back, and perfectly idiomatic Haskell can be compiled to almost-perfect machine code, but in point of fact GHC basically never does, and you still have to care about things like tail calls and design APIs to do things like continuation passing.