joelvim / merkle

A simple merkle tree implementation

Merkle tree library

This project implements a simple merkle tree library in ReasonML.

How to build an run the tests?

This project is managed by esy build tool.

To build the project, run the following commands

esy install
esy pesy
esy test

Choices made

Hashing function

I chose not to rely on a given hash function. As the purpose of the test was not to evaluate cryptography knowledge and method, I tried to remain generic against hashing. It allows switching hashing function, and testing with any dumping hashing function. In my case, the identity function :)

Tree initilization

I consider initializing a merkle tree with an empty dataset is a non sense. So I prevent it by forcing the API user to provide at least one data as input. I would have loved to have an API with a single non empty list parameter, but I don't know OCaml ecosystem well enough to use the right type. So I did it simply : two params, the head of the list, and the tail of the list that can be empty.

Empty chunks

Same as for the dataset, I consider an empty chunk of bytes is not a valid input. However, I do not know if it is possible by type to enforce non emptyness of a bytes value. So I return an error if an empty chunk is received.

Error handling

OCaml provides various ways to handle errors (exceptions, result type, even option if you want). I like strongly typed APIs, and I like knowing that a method can fail by reading it's type. My Java/Scala experience taught me that exceptions are error prone and often leads to porrly handled or forgotten exceptions. So I prefer using the result type way of doing things, event if it can have a performance impact., but in a first time,

Performance

Regarding the tree algorithm, I am not good at writing high performance program. I prefer writing a correct program, easy to reason about and maintain than a perfectly optimized one that could fail in some cases that where not anticipated.

I consider that an algorithm can be optimized if needed, but optimization should not be blind, and must be done in the context of a real use case scenario, to properly identify the performance bottlenecks.