Generate n-ary binary tree in haskell

Question

I try to create random trees from this structure:

data Tree a = Leaf a
            | Funktion (Tree a) (Tree a)
            | Lambda (Tree a)
            deriving (Show)

The problem I have is that I don't even know how to generate a tree with the depth of (for example) 2 which only has "Lambda" as nodes. If someone could help me with generating this simple tree (depth 2) I could generate them randomly. If I implement a function like this:

build (Tree a) 0 = Leaf "A"
build (Tree a) n = build (Lambda a) (n-1)

It won't work since the function build itself expects a Tree as input. Actually I need trees which have the nodes Lambda or Funktion, but first of all I need to understand how to generate a simple version of this structure.

Answer 1

It sounds like you want something like

build :: Natural       -- ^ Number, n, of Lambda nodes
      -> a             -- ^ Value, v, to store in the Leaf
      -> Tree a        -- ^ Lambda^n (Leaf v)
build 0 a = Leaf a
build n a = Lambda (build (n - 1) a)

So that build 4 "A" will produce

Lambda (Lambda (Lambda (Lambda (Leaf "A"))))

Unfortunately, the rest of your question (about generating random trees) really requires substantially more context to answer.

Answer 2

You're close - your build function at the moment will always return a Leaf , as the base case of the recursion doesn't do anything with its argument. You don't actually need the argument:

build :: Integer -> Tree String
build 0 = Leaf "A"
build n = Lambda $ build (n-1)

This would produce what your build function seems to be intending, that is a simple "tree" of given depth n composed of a Leaf node and Lambda nodes. Eg:

λ> build 2
Lambda (Lambda (Leaf "A"))

---

To get your randomly generated tree, you need to look at the System.Random module.

For example building on the previous, a tree of random height between a given upper and lower bound:

buildRandom :: (Integer, Integer) -> IO (Tree String)
buildRandom bounds = randomRIO bounds >>= return . build

This can be extended and modified to produce the behaviour you want (a fully random generated tree), but requires knowledge of monads etc., which might take some extra reading.