feat: discard txs by tx_count of sender #4520
Conversation
Codecov Report
Additional details and impacted files
... and 442 files with indirect coverage changes
Flags with carried forward coverage won't be shown. Click here to find out more.
|
chirag-bgh
changed the title
github-actions
bot
added
the
S-stale
|
reviewing this and will probably be touching some of it up soon! |
Rjected
added
M-prevent-stale
rkrasiuk
added
C-enhancement
| @@ -243,6 +246,23 @@ impl<T: TransactionOrdering> PendingPool<T> { | |||
| removed | |||
| } | |||
|
|
|||
| /// Updates the independent transaction and independent descendants set, assuming the given | |||
There was a problem hiding this comment.
needs reference "highest nonce" instead
| let mut removed = Vec::new(); | ||
|
|
||
| // penalize non-local txs if limit is still exceeded | ||
| while self.size() > limit.max_size || self.len() > limit.max_txs { |
There was a problem hiding this comment.
this is the inverse of L360 and L380?
I think we can rewrite this as loop {} and only keep a break
| let mut any_non_local = true; | ||
|
|
||
| // use descendants, so we pop a single tx per sender | ||
| for tx in self.highest_nonces.clone().iter() { |
There was a problem hiding this comment.
even though this only includes unique senders and we expect that the outer loop is really a single iterator.
I think we get rid of the clone if we instead push all txs ids that need to be removed to a tmp vec, keep track of the amount and sizes and only remove them once we have collected enough, then do the same again without locals check if necessary
There was a problem hiding this comment.
tried to implement this with transactions_to_remove
|
|
||
| // index of first tx to check | ||
| let mut removed_idx = 0; | ||
| 'outer: loop { |
There was a problem hiding this comment.
not immediately obvious why we need this and this looks a bit complex.
I guess we need this because removing 1 tx per unique sender might not be sufficient? But the highest nonce is only updated once we remove the transactions?
I think this can be simplified by putting L372-L397 in a loop that actually removes the recorded ids of removed so we don't need the loop below?
If the function takes a &mut Vec then the transactions can be pushed directly to this, for both (locals =true|false) calls
Under real circumstances this will only evict 1 to a few transactions so having a tmp buffer for the removed ids isn't too bad.
| total_size += tx.transaction.size(); | ||
| removed.push(*tx.transaction.id()); | ||
| } | ||
| // we can re-use the temp array |
| /// If the `remove_locals` flag is unset, then only non-local transactions will be removed. | ||
| pub(crate) fn transactions_to_remove( | ||
| &self, | ||
| pub(crate) fn limit_pool( |
There was a problem hiding this comment.
name should convey that this removes
| removed.push(tx); | ||
| } | ||
| } | ||
| self.limit_pool(&limit, true, &mut removed); | ||
|
|
||
| removed.shrink_to_fit(); |
There was a problem hiding this comment.
do we need to shrink anything here?
There was a problem hiding this comment.
nope, was left over from when the map was created with_capacity, removed
| .into_iter() | ||
| // sort by submission id | ||
| .collect::<BinaryHeap<_>>(); |
There was a problem hiding this comment.
not exactly sure how this works internally but I assume this into_iter collect is optimized so there are no allocs,
but I think this could be written as a loop over an iterator that pushes unique senders directly into a binaryheap
Closes #4269