Order statistic tree

In computer science, an order statistic tree is a variant of the binary search tree (or more generally, a B-tree) that supports two additional operation beyond insertion, lookup and deletion:

• Select(i) — find the i'th smallest element stored in the tree
• Rank(x) – find the rank of element x in the tree, i.e. its index in the sorted list of elements of the tree

Both operations can be performed in O(log n) time in the average case; when a self-balancing tree is used as the base data structure, this bound also applies in the worst case.

To turn a regular search tree into an order statistic tree, the nodes of the tree need to store one additional value, which is the size of the subtree rooted at that node (i.e., the number of nodes below it). All operations that modify the tree must adjust this information to preserve the invariant that

size[x] = size[left[x]] + size[right[x]] + 1

where size[nil] = 0 by definition. Select can then be implemented as

function Select(t, i)
    // Returns the i'th element (zero-indexed) of the elements in t
    r ← size[left[t]]
    if i = r
        return key[t]
    else if i < r
        return Select(left[t], i)
    else
        return Select(right[t], i - (r + 1))

External links

• Order statistic tree on PineWiki, Yale University.
• The Python package blist uses order statistic B-trees to implement lists with fast insertion at arbitrary positions.