This library encourages you to do memoization in three separate steps:
(1) Create a memoizable function
(2) Create or select an appropriate memoizer
(3) Run the memoizer on the memoizable function
Let's start with the first. When you create a memoizable function, you should use the 'self' convention, which is that the first input to the function is 'self', and all recursive calls are replaced with 'self'. One common convention that goes well with the 'self' convention is using a helper function 'go', like so:
' fib :: Memoizable (Integer -> Integer) fib self = go where go 0 = 1 go 1 = 1 go n = self (n-1) + self (n-2) '
Now for the second. For this example, we need a Memoizer that can handle an 'Integer' input, and an 'Integer' output. 'Data.MemoCombinators' provides 'integral', which handles any 'Integral' input, and any output. 'Data.MemoUgly' provides 'memo', which can memoize any function 'a -> b', given an 'Ord' instance for 'a'.
Third, let's run our memoizers! Since we have decoupled the definition of the memoized function from its actual memoization, we can create multiple memoized versions of the same function if we so desire.
' import qualified Data.MemoUgly as Ugly import qualified Data.MemoCombinators as MC
fibUgly :: Integer -> Integer fibUgly = runMemo Ugly.memo fib
fibMC :: Integer -> Integer fibMC = runMemo MC.integral fib '
You could easily do the same with 'Data.MemoTrie.memo', 'Data.Function.Memoize.memoize', etc.
Using this technique, you can create local memoized functions whose memo tables are garbage collected as soon as they are no longer needed.