In this chapter, we learned how lazy evaluation works, what weak head normal form is, and how to control it by increasing strictness with different methods. We considered the peculiarities of right-fold, left-fold, and strict left-fold, and in which situations one fold strategy works better than another. We introduced the concept of CAF along with memoization techniques, utilized the worker/wrapper pattern, and used guarded recursion to write clean and efficient recursive programs.
We used the :sprint command in GHCi to inspect unevaluated thunks and the Runtime System option -s to inspect the heap usage and GC activity of compiled programs. We took a look at inlining, stream fusion, and the performance costs of partial functions and polymorphism.
In the next chapter, we will take a look at other basic data and control structures, such as different array structures and some monads. But first, we will learn about the performance semantics of Haskell data types and related common optimization techniques.