perf: SyncJob transfer matching does full-table scan on every family sync

[JSONbored]

Sentry: SURE-APP-WX (18,894 events, last seen today)

Family::Syncer#perform_post_sync calls auto_match_transfers! with no date filter, causing a full cartesian scan across all entries in a family's accounts after every sync.

For families with 10,000+ entries this is an O(N²) self-join:

SELECT ... FROM entries inflow_candidates
JOIN entries outflow_candidates ON (
  inflow_candidates.date BETWEEN outflow_candidates.date - 4 AND outflow_candidates.date + 4
  ...
)
JOIN accounts inflow_accounts ON ...
WHERE inflow_accounts.family_id = $1 AND outflow_accounts.family_id = $2
-- no date bound — scans every unmatched entry ever

Most entries added during an ongoing sync are recent (last few days). Scanning all historical entries is unnecessary.

Fix: add a since_date: parameter to transfer_match_candidates and auto_match_transfers!, and pass since_date: 90.days.ago.to_date from Family::Syncer#perform_post_sync. Account-level syncs keep since_date: nil for full-history matching on a specific account.

[dosubot]

There's already an open PR addressing this exact issue — PR #2444 adds the since_date: parameter to transfer_match_candidates and auto_match_transfers!, passing 90.days.ago.to_date from Family::Syncer#perform_post_sync while keeping since_date: nil for account-level syncs.

For context, the current call site in Family::Syncer#perform_post_sync is simply family.auto_match_transfers! with no arguments [1], which triggers the unbounded family-wide scan you're describing.

Note that PR #2230 (merged June 11) already addressed the related issue from #2227 by scoping Account::Syncer post-sync matching to only the synced account. PR #2444 layers the date bound on top of that for the remaining family-level path.

To reply, just mention @dosu.

---

Share context across your team and agents. Try Dosu.