Jul 18, 2026vector, functional-programming, trees, data-structures

RRB Vector

An RRB vector is a persistent, immutable sequence implemented using a Relaxed Radix Balanced tree¹.

It's essentially a persistent vector redesigned to make concatenation and slicing fast, while retaining efficient indexed access²

            root
     ┌───────┼────────┐
     ▼       ▼        ▼
  subtree  subtree  subtree
   64 el.   41 el.   28 el.
     │        │         │
[leaves containing blocks of values]

RRB

Radix: an index is divided into bit groups used to navigate the tree. Implementations commonly use a branching factor of 32.
Balanced: the leaves remain at approximately the same tree depth. i.e. you do not want one branch to become arbitrarily deep like a linked list. Keeping the tree shallow preserves logarithmic lookup and update behavior.
Relaxed: subtrees do not have to be completely full or uniformly sized. Nodes can store cumulative subtree sizes to route index lookups through these uneven branches.

Performance

Operation	Approximate complexity
Indexed lookup	`O(log₃₂ n)`
Immutable update	`O(log₃₂ n)`
Append	`O(log₃₂ n)`, effectively near constant
Concatenate vectors	`O(log n)`
Materialized slice	`O(log n)`
Iteration	`O(n)`

Use Cases

Unlike an array, it supports efficient immutable changes.
Unlike a linked list, it provides fast indexed access.
Unlike a basic persistent vector, it supports efficient concatenation and real, non-view slicing.
Unlike a rope, it remains suitable as a general-purpose indexed sequence.

RRB

Performance

Use Cases

See Also

References