thyeem / monotree Goto Github PK

At the moment, I lazy initialize a single sled db instance, then create multiple Monotree instances via open_tree method on the sled db instance. Would like to see direct support OOTB if it is not already available,

Hi, I was looking for a simple merklized database in Rust to experiment with and this came up. It looks nice and simple and fast.

I ran some benchmarks locally and read the examples, and I like the design in general. I am not sure if this is being maintained much now, but a few questions on proper usage:

• It exposes batches of all inserts or all removes or all gets. What about a batch of mixed inserts and removed? Is that just limited due to keeping the API simple? Or is there a reason it would be tricky to add?
• This stores key_hash -> value_hash with merkle proof potential, which is nice an small. To use as a real storage, I would need a sibling DB to store the raw key->value mapping (unmerklized) and keep this just for proofs, correct?
• RocksDB offers checkpoints, so I could provide read-only access to one version (for consistent long-reads/admin) while continuing to write. This is not exposed, but I assume this is mainly an issue of API clarity?

I don't want to ask you to implement these, but I would be interested in hearing if your thoughts on them and if they could be added without huge complexity overhead.

One issue there is keeping 2 databases in sync if I cannot commit atomically over them both. I could imagine sharing one eg. RocksDB under the hood, with all data stored with eg. 0 prefix and all data from MonoTree with 1. Then building a batch that contains all data updates and all MonoTree updates, so I can get the data and the proofs together in an atomic commit. This would require:

• Prepending a prefix to writes, stripping from reads
• Passing in a DB rather than creating it internally - for every call
• (Most challenging) Relying on "someone else" to manage the batches and access

Apart from the API challenges, do you think something like that would be feasible?

Thank you for your time reading this. And let me know if I can support the project somehow.

It says that you can generate non inclusion proofs, but there doesn't seem to be any method provided for doing that?

export , import all nodes from root
as json , bson

Much appreciated