You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
[NRT-771] Count AH-DB queries in the submit guard, not reader method names
Review caught the guard being weaker than it claimed. It spied on `note_data`
and `note_type_dict` by name, and the fix removed `note_data` from that path
entirely, so the observed pair was (0, 1) for both runs. Regressing to call the
*batched* `notes_data_for_anki_nids([nid])` once per note would issue thousands
of queries at the bulk cap and leave the test green — the earlier check only
proved it catches a return to the one implementation it was written against.
Count at the peewee layer instead, so a per-note read reintroduced through any
helper is caught. Verified against that scenario: it now fails 12 == 24.
Submit both new-note and change candidates too. The media step runs over each
list separately, so a change-suggestion-only regression was uncovered.
0 commit comments