mirror of
https://github.com/MBombeck/HealthLog.git
synced 2026-11-08 20:42:43 +00:00
Marc-André Bombeck
f8f460bb66
Migration 0263 added three columns beside the Apple Health `stats:*` authority ladder. Only `aggregation_provenance` is ever read back: the external-measurement reconciler selects it and refuses a lower-authority write against a higher-authority canonical row, and the cumulative drain reads it too. That column stays. `aggregation_contributor_count` and `aggregation_selected_source_hash` are written by three call sites and read by none. No select, no raw SQL, no export, no backup, no API contract, no repair path. The import-time selection keeps its deterministic tiebreak, which compares an in-memory map key computed per flush rather than a value read back from the table. Both are plain nullable scalars with no index, unique, check or foreign key, so the drop is a catalog-only operation that touches no measured value.
37 lines
2.0 KiB
SQL
37 lines
2.0 KiB
SQL
-- v1.32.35 — drop the two write-only aggregation columns on `measurements`.
|
|
--
|
|
-- Migration 0263 added three columns alongside the Apple Health `stats:*`
|
|
-- authority ladder. Only one of them was ever read back:
|
|
--
|
|
-- * `aggregation_provenance` is load-bearing and STAYS. The external
|
|
-- measurement reconciler selects it and refuses a lower-authority write
|
|
-- against a higher-authority canonical row, which is what keeps an
|
|
-- export.xml source-day estimate from clobbering a native HealthKit
|
|
-- statistic. The cumulative drain reads it too.
|
|
-- * `aggregation_contributor_count` and `aggregation_selected_source_hash`
|
|
-- are written and never read. Three call sites write them (the batch
|
|
-- ingest and the cumulative drain write explicit NULLs, the export.xml
|
|
-- import writes real values); nothing selects them, no raw SQL names
|
|
-- them, they reach no export, no backup, no API contract and no repair
|
|
-- path. Verified by a repo-wide search over every tracked file for both
|
|
-- the Prisma field names and the column names.
|
|
--
|
|
-- The import-time selection algorithm keeps its deterministic tiebreak: the
|
|
-- source hash it compares is an in-memory map key computed per flush, not a
|
|
-- value read back from this table. Only the persisted copies go.
|
|
--
|
|
-- DATA BOUNDARY — read before editing.
|
|
--
|
|
-- Both columns are plain nullable scalars added by 0263 with no index, no
|
|
-- unique, no check and no foreign key, so `DROP COLUMN` rewrites the tuples of
|
|
-- `measurements` and stops there. `DROP COLUMN` never cascades into another
|
|
-- table on its own, and there is no CASCADE keyword here. No measured value,
|
|
-- timestamp, unit, note or identity field is touched, and the documented
|
|
-- LEGACY_UNKNOWN repair path operates on `aggregation_provenance`, which
|
|
-- survives. In Postgres this is a catalog-only operation, so it is cheap even
|
|
-- on the largest table in the schema.
|
|
|
|
ALTER TABLE "measurements"
|
|
DROP COLUMN IF EXISTS "aggregation_contributor_count",
|
|
DROP COLUMN IF EXISTS "aggregation_selected_source_hash";
|