node package manager


A Merkle Tree skeleton, written in IcedCoffeeScript


A JS Merkle Tree implementation, as used by keybase.


npm install merkle-tree

And then

var Base = require('merkle-tree').Base;


make test

All tests should pass.


This module is just a library, and for it to do anything useful, you'll have to subclass the Base class required above. As an example, we provide a subclass of a Merkle-Tree that lives in memory, which can be accessed as follows:

var merkle_mod = require('merkle-tree');
// M = the number of children per interior node. 
// N = the maximum number of leaves before a resplit. 
var config = new merkle_mod.Config({ N : 4, M : 16 });
var myTree = new merkle_mod.MemTree(config);
// Keys are hashes expressed as hex strings. 
var key = "961b6dd3ede3cb8ecbaacbd68de040cd78eb2ed5889130cceb4c49268ea4d506";
var value = { "foo" : 10 };
// We're just inserting one, but you can insert as many as you'd like. 
myTree.upsert({'key' : key, 'value' : value}, function(err, new_root_hash) {
    // Finding by default checks the hashes on all interior nodes down the tree. 
    // If you want to speed up your 'finds', then you can pass `skip_verify : true` 
    // to your find. 
    myTree.find({'key' : key, 'skip_verify' : false}, function(err, val2) {
        assert.equal(value, val2);
// You can either build a tree one key/value pair at a time, as above, or 
// you can build the whole thing at once. 
var data = new merkle_mod.SortedMap({
  "list": [
     ["aabbcc", "dog" ],
     ["ddccee", "cat" ],
     ["00aa33", "bird" ]
});{"sorted_map" : data }, function (err) {

To review, the Merkle Tree module provides the following classes:

  • Config -- A configuration object that controls the shape of the tree.
  • SortedMap -- A sorted map of key/value pairs that used for inputting a whole bunch of data at a time, and is also used internally.
  • Base -- A base, abstract tree implementation that needs to specialized.
  • MemTree -- A speciailization of Base; all data lives in memory and disappears when the process ends.

The Base class has the following method calls:

  • build({sortedMap}, cb) --- Build a tree from scratch using the given sorted map of data, and callback when done.
  • upsert({key,value,[txinfo]}, cb) --- Update or insert the given value at the given key. Provide optional txinfo that is passed to the storage engine.
  • find({key}, cb) --- Find the given key in the Merkle tree, starting from the root and going down.

How to Make an On-Disk Tree

The keybase server stores its Merkle tree on disk. It implements the following methods of the Base class to do so:

  • hash_fn(s) -- A function to hash an interior node into a key. Return the hex-string hash of the given string. I'd just use SHA512: require('crypto').createHash('SHA512').update(s).digest('hex').
  • store_node({key, obj, obj_s}, cb) --- Store the node value obj under the key key. For convenience, you are also passed obj_s, the stringification of the object.
  • lookup_node({key},cb) --- Read from disk the node whose key is key. Callback with the parsed (not stringified) object
  • lookup_root(cb) --- Should callback with the hash of the most recent tree root.
  • commit_root({key,txinfo}, cb) --- Store the root hash to disk, optionally with the txinfo transaction info annotation.

For an example of how to do this, see the simple MemTree class.