Introduce RocksDB-backed Merkle Tree #6
Conversation
| let mut tmp = [0u8; 32]; | ||
| value.to_big_endian(&mut tmp); | ||
|
|
||
| let key = U256::from_little_endian(&derived_key); |
There was a problem hiding this comment.
Do these endianness conversions work correctly on all platforms? i.e. is derived_key little-endian even on big-endian platform?
There was a problem hiding this comment.
I would assume so, though I can't guarantee. Do you have an idea on how to assert this?
There was a problem hiding this comment.
Maybe checking the LogQuery implementation is one way? I'm not sure if this is a big concern right now, since there are very little big endian platforms alive anymore, and probably none of them will run this code, but I got curious since there's both endianness conversion / assumption in our code base. 🙂
| pub struct TreeWrapper<'a> { | ||
| pub tree: MerkleTree<'a, RocksDBWrapper>, | ||
| // FIXME: How to save this for persistant storage? | ||
| pub index_to_key: Vec<U256>, |
There was a problem hiding this comment.
Could we just write it ~as is? i.e. <num values: 32bit unsigned>,<value0>, <value1>, <value2>... - all numbers big endian encoded?
There was a problem hiding this comment.
I'm not fully sure of what you want, would it still be a Vec, containing the length as the first element and the rest of the values (keys) in their non-parsed variant?
There was a problem hiding this comment.
Oh, sorry I expressed myself a bit poorly. I basically responded to your FIXME comment question from above. No need to change anything here, but my thinking was that one possibility for persistent storage solution here would be to just write the values out to file, prefixing them by number of elements.
No need to do anything here 🙂
| pub struct TreeWrapper<'a> { | ||
| pub tree: MerkleTree<'a, RocksDBWrapper>, | ||
| // FIXME: How to save this for persistant storage? | ||
| pub index_to_key: Vec<U256>, |
There was a problem hiding this comment.
Oh, sorry I expressed myself a bit poorly. I basically responded to your FIXME comment question from above. No need to change anything here, but my thinking was that one possibility for persistent storage solution here would be to just write the values out to file, prefixing them by number of elements.
No need to do anything here 🙂
| let mut tmp = [0u8; 32]; | ||
| value.to_big_endian(&mut tmp); | ||
|
|
||
| let key = U256::from_little_endian(&derived_key); |
There was a problem hiding this comment.
Maybe checking the LogQuery implementation is one way? I'm not sure if this is a big concern right now, since there are very little big endian platforms alive anymore, and probably none of them will run this code, but I got curious since there's both endianness conversion / assumption in our code base. 🙂
CommitBlockInfoV1and returning the new L2 block number.clippycomplaints.For now the database is purged before each run since we need to figure out a way to also persist the index to key mappings.