# Naturals Search

I came across this interesting pattern a while ago, while taking my first substantial steps in Haskell, and thought I’d blog it.

It’s a logarithmic time search over the Naturals, for any predicate where we can tell if a number’s too big. I use it here to implement square-root:

We define an `ITree`

as a tree of integers. Each node can either be a leaf, containing an `Integer`

, or a node containing an `Integer`

and two sub-trees:

This tells us how to show an `ITree`

so that we can use printf-debugging:

```
instance Show ITree where
show (IL x) = "(" ++ show x ++ ")"
show (IN x l r) = "(" ++ show x ++ " " ++ show l ++ " " ++ show r ++ ")"
```

This is an `ITree`

, and in fact is the only one we need. It is constructed lazily, and has the following shape:

```
1
/ \
1 2
/ \
2 4
/ \
3 \
/ \ \
3 4 \
8
/ \
6 \
/ \ .
/ \ .
5 7 .
/ \ / \
5 6 7 8
```

The idea is that the right-most branches count up in powers of `2`

forever. The branches coming off on the left contain a leaf for each of the numbers between this power of `2`

and the previous (eg. the branch coming off `8`

contains leaves for `5`

, `6`

, `7`

and `8`

). These are spread out by splitting into above/below the average value which we stick on the node.

```
iTree = let biTree l h | l == h = IL l
| otherwise = let m = avgI l h in
IN m (biTree l m) (biTree (m + 1) h)
from x = IN (2^x) (biTree (2^(x-1) + 1) (2^x)) (from (x+1)) in
IN 1 (IL 1) $ IN 2 (IL 2) $ from 2
```

This function finds the largest number which doesn’t trigger the given `tooHigh`

function. It does this by traversing `iTree`

: it keeps going right until `tooHigh`

becomes `True`

, then it snakes down the left sub-tree, hugging the `tooHigh`

boundary until it hits a leaf.

This is a simple example which uses `iTree`

to find the integer square root of a number.

Here’s a `QuickCheck`

property to verify that it’s correct.

```
prop_sqrtI n = let root = sqrtI n
lower = (root - 1)^2
higher = (root + 1)^2 in
n > 0 ==> lower <= n && higher >= n
```

Anyone know what this `iTree`

structure is called?