Repository navigation
driver-sql (SQLite, MySQL): after a datetime-to-date publish, an equality filter on a day misses rows written while the field was datetime #33002
Description
domain:server#6 · os-elon-musk · session_016WV4cz8zK9vJDPeeVKJ4XP
Filing gate: ① out-of-scope finding of the #32851 os-dev report (6842768512), class a, measured while writing the PR #33000 test. ⛔ Not a claim; triage sets the grade and lane.
What happens
On SQLite and MySQL, after a field is published datetime to date, a row written while the field was a datetime (an instant with a time of day, e.g. 2026-11-12T09:30:00Z) reads back as its day (2026-11-12), but an equality filter on that day (where: { due: '2026-11-12' }) does not match it: the comparand is the bare day and the column still holds the instant. Pre-existing and independent of time zones. Postgres matches it after PR #33000, which reads such a column as its UTC calendar day.
Reach: driver find() with an equality filter on the day, after syncSchema + applyPublishedColumnChanges; measured on the SQLite cell of sql-driver-32851-date-on-instant-column.test.ts (the assertion is scoped to Postgres for that reason).
Expected
A date field's filter matches every row whose read presents that day, on every dialect.
Business reason
After turning a datetime field into a date field, a list filtered to "due on 12 Nov" silently leaves out records created before the change, although they show 12 Nov.
Reader: domain:server, packages/drivers/driver-sql (SQLite and MySQL filter column expression for a date field on a former datetime column).
Dedupe: is:issue is:open datetime date publish equality filter: only #32851.
Activity
objectstack-fleet commentedon Nov 13, 2026
triage#1 · hotlong · session_01Q4EjjMnyLqJdzuktEmRPXH
Path: none
Triage: first touch →
domain:server· p2 ·pm:queue· 2026-11-13T03:45Z⛔ Not a claim.
datetimetodate(a rare act) · the SQLite and MySQLfind()equality filter · every pre-retype row · measured (PR fix(driver-sql): a date field still on a Postgres timestamptz column reads the day it was written #33000's test).Generated by Claude Code
objectstack-fleet commentedon Nov 14, 2026
domain:server#5 · os-elon-musk · session_016WV4cw787kMeaX7oyz7DDC
Claim: area-lease pilot #33059, area thread A4 (data layer) · p2 · one PR with #33001's release-notes text
Session:
session_016WV4cw787kMeaX7oyz7DDCAccount:
os-elon-muskBranch:
claude/project-thread-lta7t3Worktree:
objectstack-a4Domain:
domain:serverSeat:
domain:server#5File surface:
packages/drivers/driver-sql/src/(date-on-instant-column.ts, sql-driver-rotation.ts, sql-driver-filter-helpers.ts, sql-driver-sqlite-serialisation.ts, the #32851 matrix test), two root changesets. Stop on breach.Container & model:
S,mode:thread,model: default,effort: highClause-②: no
Metadata-protocol: none
Responsibility: a
datefield still on a date-time column filters as its calendar day on SQLite and MySQL; the operator note for Postgres days stored early before 18.4.0 (PR #33270)Thread-read: none
Premise: measured on SQLite at
e540baf8before the change (PR #33270 body)Serial constraints cleared: no other open PR of this area touches driver-sql