Conversation
vicsn
left a comment
vicsn
left a comment
There was a problem hiding this comment.
Thank you for your contribution
| /// Reads the deployment from a buffer. | ||
| fn read_le<R: Read>(mut reader: R) -> IoResult<Self> { | ||
| // Read the id variant. | ||
| let id_variant = u8::read_le(&mut reader)?; |
There was a problem hiding this comment.
Can this be made backwards compatible?
| let mut deployment = | ||
| serializer.serialize_struct("Deployment", len + self.id.get().is_some() as usize)?; | ||
| if let Some(id) = self.id.get() { | ||
| deployment.serialize_field("id", id)?; |
| mod string; | ||
|
|
||
| use crate::{Transaction, Transition}; | ||
| use std::sync::OnceLock; |
There was a problem hiding this comment.
please move down to line 26 so external dependencies are in the same place
| #[derive(Clone)] | ||
| pub struct Deployment<N: Network> { | ||
| /// The deployment id | ||
| id: OnceLock<DeploymentID<N>>, |
There was a problem hiding this comment.
Would a design be possible where this is just an Option<...> like we did for the previous similar caching specified in the issue?
There was a problem hiding this comment.
That would require mutable constructs to deployment all over the code base as when we call to_deployment_id it's probably best to cache the id if not computed yet once, this also implies a redundant recursive chain of callers that will have to take an &mut. We could probably do a RefCell though?
2 participants
Motivation
closes #2706
Tx root derivation takes some time during block generation, but these can be cached when computed for the first time.
Test Plan
No critical or logic changes. CI should catch any regressions