Monday 21 September 2009

Real World Haskell (2)

Chapter 2 of Real World Haskell gives a quick overview of types and functions. I think I've already picked up most of the bits in here, so this is just some notes on the things I found new / interesting.

One of Haskell's distinguishing features is its strong type system. Real World Haskell says the three interesting aspects are:

  1. Strong types - The type system guarantees that the program does not contain certain errors (treating an integer as a float). In comparison, something like C allows you to reinterpret the bytes as you see fit (weak typing)
  2. Static Types - At the point the code is built, the compiler knows the type of every expression and value. In comparison dynamically typed languages only know the type of an expression at runtime.
  3. Type Inference - The compiler can automatically infer the types of expressions. Type definitions are optional and Haskell can deduce the required types (however, it seems to be good practise to provide the type signature as a documentation aid).

Lists and tuples are the primary way of grouping elements. Items in a list must all be the same type, items in a tuple can be different types. One interesting point made was that each tuple is a unique type and therefore writing a function to "get the second element from any tuple" is almost impossible. Looking at the types of fst :: (a,b) -> a and snd :: (a,b) -> b, it's easy to see why. In order to get the second element from any tuple, we'd have to define it for an infinite range of tuples (a,b,c), (a,b,c,d) and so on.

When Haskell evaluates an expression it uses normal-order evaluation. Normal-order evaluation simply evaluates expressions as they are needed. This is also known as a lazy evaluation strategy because expressions that don't contribute to the result aren't evaluated. This avoids the special case treatment that short-circuiting operators get in other languages.