alucha

alucha

Hash Array Map Trie (HAMT) implementation based on CHAMP paper and inspired by Clojure. Library provides a builder API inspired by clojure transients that can be utilized for bulk updates.

Name

I wanted to call this library maple, but that name seems to be taken so I chose to call it Alucha instead, which is a type of (most delicious) plum tree.

Status

Prototype implements hash-array map tries (HAMT) by synthesizing CHAMP algorithm with HAMTs used by IPFS UnixFS and IPLD HAMT. CHAMP algorithm utilized several optimizations that are incompatible with IPFS UnixFS and IPLD HAMT, this implementation attempts to utilize those optimizations at runtime and provide a way to map it to IPFS UnixFS compatible representation.

Optimizations

CHAMP algorithm exploits branching factor of 32 because it can be manipulated efficiently on 32-bit processors. It also allows storing HAMT bitmaps in a single integer.

In order to support IPFS UnixFS implementation (which uses 256 factor by default) branching factor is configurable, but relevant optimizations only apply when branching factor of 32 is used.

It is also worth noting that even with branching factor of 32 it can store large amount of entries in a relatively shallow tree e.g. tree 6 levels deep could store around 33,554,432 entries.

Credits

• Leveling up Clojure’s Hash Maps
• Hash Array Mapped Tries
• Optimizing Hash-Array Mapped Tries for Fast and Lean Immutable JVM Collections
• Lean Hash Array Mapped Trie (Lean Map)
• Ideal hash trees