chore(ntx-builder): store sponsorship fee notes in the database - #2494
chore(ntx-builder): store sponsorship fee notes in the database#2494SantiagoPittella wants to merge 4 commits into
Conversation
2b72f23 to
7196ac1
Compare
| pub(crate) async fn insert_sponsorship_notes( | ||
| &self, | ||
| notes: Vec<SponsorshipNote>, | ||
| ) -> Result<(), DatabaseError> { | ||
| self.writer | ||
| .write("insert_sponsorship_notes", move |tx| { | ||
| queries::insert_sponsorship_notes(tx, ¬es) | ||
| }) | ||
| .await | ||
| } | ||
|
|
||
| pub(crate) async fn mark_sponsorships_consumed( | ||
| &self, | ||
| nullifiers: Vec<Nullifier>, | ||
| block_num: BlockNumber, | ||
| ) -> Result<(), DatabaseError> { | ||
| self.writer | ||
| .write("mark_sponsorships_consumed", move |tx| { | ||
| queries::mark_sponsorships_consumed(tx, &nullifiers, block_num) | ||
| }) | ||
| .await | ||
| } |
There was a problem hiding this comment.
Would these not need to be part of the same db tx as the entire block update?
There was a problem hiding this comment.
In the apply_committed_block function it is part of the same tx, this are under cfg(test) and used only for testing purposes.
There was a problem hiding this comment.
I think ideally we would use production APIs to setup tests where possible i.e. actually get proper blocks in to inject and consume these. But that's part of a larger problem I think, so I guess its fine.
| -- Inserts a FEE_SPONSORSHIP note from a committed block. Uses `INSERT OR IGNORE` so re-applying | ||
| -- the same block (e.g. on a redelivery from the subscription stream) is a no-op rather than a |
There was a problem hiding this comment.
Would redelivery not be a pretty major bug?
There was a problem hiding this comment.
yes, that's a bug and shouldn't happen. This is defensive, to avoid breaking the DB in that case. If you think that we should error out I can change it (and also change insert_network_note.sql because that does the same)
There was a problem hiding this comment.
Up to you; I don't mind that much. The comment just made me concerned.
| -- are kept around (not deleted), mirroring the `notes` table lifecycle. | ||
| UPDATE sponsorship_notes | ||
| SET committed_at = ?2 | ||
| WHERE nullifier IN (SELECT value FROM rarray(?1)) AND committed_at IS NULL |
There was a problem hiding this comment.
We could technically avoid the IS NULL check or perhaps it speeds things up with a smaller set?
| /// Returns the unconsumed `FEE_SPONSORSHIP` notes bound to the given account's unconsumed feature | ||
| /// notes, grouped by feature note id. | ||
| pub fn pending_sponsorships( | ||
| tx: &ReadTx<'_>, |
There was a problem hiding this comment.
Name feels a bit weird; but perhaps I need to see how it gets used in the next PR :)
2 participants
Summary
Part 1 of 3 for #2327 (network account fees).
With the new fee sponsorship model, the fee for a network note (the "feature note") travels in a separate
FeeSponsorshipNotebound to it by note id. Sponsorship notes carry no attachments, so today the ntx-builder silently drops them during block ingestion.This PR makes the ntx-builder detect them (by script root) and persist them in a new
sponsorship_notestable, indexed by the feature note they pay for. Consumption is tracked by nullifier, and a sponsorship may arrive before or after its feature note since the binding is resolved at read time with a join. The index rebuilds automatically on restart via the existingblock replay.
No behavior change: transaction selection does not use the index yet. Follow-up PRs:
Changelog