Question

I am playing around with the aho-corasick algorithm to try and get a little better with F#, and I ran across a problem with the Trie implementations, they are all mutable or can’t be tail call optimized.

The basic issue from what I can see is that immutable data structures have to be built “bottom up” since you can’t change what they point to, so your options are either make them mutable, or find out the nodes as you go along(i.e. recurse in construction).

Is there any way of making an immutable trie data structure with tail call optimizations on the construction?(and not lose efficiency by copying.)

Answer 1

The tail call optimiation can be eliminated by using a continuation. Here’s a sample where the key and value are string and int respectively

type Trie = 
 | Data of string * int * Trie * Trie 
 | Leaf 

let Insert start key value = 
  let rec inner current withNode = 
    match current with
    | Data (currentKey, currentValue, left, right) ->
      if key < currentKey then
        inner left (fun left -> Data (currentKey, currentValue, left, right))
      else 
        inner right (fun right -> Data (currentKey, currentValue, left, right))
    | Leaf -> withNode (Data (key, value, Leaf, Leaf))
  inner start (fun x -> x)

Eliminating the copies is a bit more difficult if you want to stick with immutable structures