Merge pull request 'fix(tradein): hotfix миграции 108 — deploy red (typo + collision-safe репойнт)' (#1547) from fix/migration-108-collision-safe into main
Some checks failed
Deploy Trade-In / changes (push) Successful in 7s
Deploy Trade-In / build-frontend (push) Has been skipped
Deploy Trade-In / build-browser (push) Has been skipped
Deploy Trade-In / test (push) Failing after 35s
Deploy Trade-In / build-backend (push) Has been skipped
Deploy Trade-In / deploy (push) Has been skipped

Reviewed-on: #1547
This commit is contained in:
lekss361 2026-06-16 05:48:38 +00:00
commit 7d39aebbd4

View file

@ -23,7 +23,7 @@
-- house_imv_evaluations (house_id) — CASCADE — UNIQUE(house_id) → dedup -- house_imv_evaluations (house_id) — CASCADE — UNIQUE(house_id) → dedup
-- house_suggestions (house_id) — CASCADE — UNIQUE(house_id, ext_item_id) → dedup -- house_suggestions (house_id) — CASCADE — UNIQUE(house_id, ext_item_id) → dedup
-- external_valuations (house_id) — CASCADE — нет ограничивающего UNIQUE -- external_valuations (house_id) — CASCADE — нет ограничивающего UNIQUE
-- address_mismatch_audit (house_id) — CASCADE — нет ограничивающего UNIQUE -- address_mismatch_audit (house_id) — CASCADE — UNIQUE(house_id,audit_batch) → dedup
-- --
-- Идемпотентность: все UPDATE/DELETE ищут по mapping временной таблицы. При повторном -- Идемпотентность: все UPDATE/DELETE ищут по mapping временной таблицы. При повторном
-- запуске mapping будет пустым (дублей нет) → все UPDATE/DELETE затронут 0 строк → no-op. -- запуске mapping будет пустым (дублей нет) → все UPDATE/DELETE затронут 0 строк → no-op.
@ -80,7 +80,7 @@ clusters_with_count AS (
), ),
-- Только дома, входящие в кластер-дубль -- Только дома, входящие в кластер-дубль
dup_houses AS ( dup_houses AS (
SELECT h.id, c.cluster_key SELECT h.id, cl.cluster_key
FROM houses h FROM houses h
JOIN clustered cl ON cl.id = h.id JOIN clustered cl ON cl.id = h.id
JOIN clusters_with_count cw ON cw.cluster_key = cl.cluster_key JOIN clusters_with_count cw ON cw.cluster_key = cl.cluster_key
@ -203,47 +203,58 @@ FROM _108_dup_mapping m
WHERE haa.house_id = m.dup_id; WHERE haa.house_id = m.dup_id;
-- 3c. houses_price_dynamics UNIQUE(house_id, month_date, source) -- 3c. houses_price_dynamics UNIQUE(house_id, month_date, source)
-- Удалить строки дубля, у которых (month_date,source) уже есть у canonical. -- Collision-safe: дедуп по ЦЕЛЕВОМУ ключу (canonical, month_date, source) —
DELETE FROM houses_price_dynamics hpd -- ловит и dup-vs-canonical, и dup-vs-dup в кластерах >2. Выживает строка
USING _108_dup_mapping m -- canonical'а (m.canonical_id IS NULL), иначе минимальный id.
WHERE hpd.house_id = m.dup_id DELETE FROM houses_price_dynamics t
AND EXISTS ( USING (
SELECT 1 FROM houses_price_dynamics hpd2 SELECT t2.id,
WHERE hpd2.house_id = m.canonical_id ROW_NUMBER() OVER (
AND hpd2.month_date = hpd.month_date PARTITION BY COALESCE(m.canonical_id, t2.house_id), t2.month_date, t2.source
AND hpd2.source = hpd.source ORDER BY (m.canonical_id IS NULL) DESC, t2.id ASC
); ) AS rn
-- Репойнтить оставшиеся FROM houses_price_dynamics t2
LEFT JOIN _108_dup_mapping m ON m.dup_id = t2.house_id
) d
WHERE t.id = d.id AND d.rn > 1;
-- Репойнтить оставшиеся (целевой ключ теперь свободен)
UPDATE houses_price_dynamics hpd UPDATE houses_price_dynamics hpd
SET house_id = m.canonical_id SET house_id = m.canonical_id
FROM _108_dup_mapping m FROM _108_dup_mapping m
WHERE hpd.house_id = m.dup_id; WHERE hpd.house_id = m.dup_id;
-- 3d. house_imv_evaluations UNIQUE(house_id) — у canonical уже может быть запись -- 3d. house_imv_evaluations UNIQUE(house_id)
-- Удалить строки дубля если canonical уже имеет evaluation. -- Collision-safe: одна evaluation на canonical. Выживает canonical'а, иначе min id.
DELETE FROM house_imv_evaluations hie DELETE FROM house_imv_evaluations t
USING _108_dup_mapping m USING (
WHERE hie.house_id = m.dup_id SELECT t2.id,
AND EXISTS ( ROW_NUMBER() OVER (
SELECT 1 FROM house_imv_evaluations hie2 PARTITION BY COALESCE(m.canonical_id, t2.house_id)
WHERE hie2.house_id = m.canonical_id ORDER BY (m.canonical_id IS NULL) DESC, t2.id ASC
); ) AS rn
-- Репойнтить оставшиеся (canonical не имеет evaluation — забираем от дубля) FROM house_imv_evaluations t2
LEFT JOIN _108_dup_mapping m ON m.dup_id = t2.house_id
) d
WHERE t.id = d.id AND d.rn > 1;
-- Репойнтить оставшиеся
UPDATE house_imv_evaluations hie UPDATE house_imv_evaluations hie
SET house_id = m.canonical_id SET house_id = m.canonical_id
FROM _108_dup_mapping m FROM _108_dup_mapping m
WHERE hie.house_id = m.dup_id; WHERE hie.house_id = m.dup_id;
-- 3e. house_suggestions UNIQUE(house_id, ext_item_id) -- 3e. house_suggestions UNIQUE(house_id, ext_item_id)
-- Удалить строки дубля, у которых (ext_item_id) уже есть у canonical. -- Collision-safe: дедуп по (canonical, ext_item_id) — ловит dup-vs-dup.
DELETE FROM house_suggestions hs DELETE FROM house_suggestions t
USING _108_dup_mapping m USING (
WHERE hs.house_id = m.dup_id SELECT t2.id,
AND EXISTS ( ROW_NUMBER() OVER (
SELECT 1 FROM house_suggestions hs2 PARTITION BY COALESCE(m.canonical_id, t2.house_id), t2.ext_item_id
WHERE hs2.house_id = m.canonical_id ORDER BY (m.canonical_id IS NULL) DESC, t2.id ASC
AND hs2.ext_item_id = hs.ext_item_id ) AS rn
); FROM house_suggestions t2
LEFT JOIN _108_dup_mapping m ON m.dup_id = t2.house_id
) d
WHERE t.id = d.id AND d.rn > 1;
-- Репойнтить оставшиеся -- Репойнтить оставшиеся
UPDATE house_suggestions hs UPDATE house_suggestions hs
SET house_id = m.canonical_id SET house_id = m.canonical_id
@ -268,7 +279,20 @@ SET house_id = m.canonical_id
FROM _108_dup_mapping m FROM _108_dup_mapping m
WHERE ev.house_id = m.dup_id; WHERE ev.house_id = m.dup_id;
-- 3i. address_mismatch_audit (CASCADE, нет блокирующего UNIQUE) -- 3i. address_mismatch_audit UNIQUE(house_id, audit_batch) — collision-safe
-- (заголовок ранее ошибочно помечал «нет блокирующего UNIQUE»).
DELETE FROM address_mismatch_audit t
USING (
SELECT t2.id,
ROW_NUMBER() OVER (
PARTITION BY COALESCE(m.canonical_id, t2.house_id), t2.audit_batch
ORDER BY (m.canonical_id IS NULL) DESC, t2.id ASC
) AS rn
FROM address_mismatch_audit t2
LEFT JOIN _108_dup_mapping m ON m.dup_id = t2.house_id
) d
WHERE t.id = d.id AND d.rn > 1;
-- Репойнтить оставшиеся
UPDATE address_mismatch_audit ama UPDATE address_mismatch_audit ama
SET house_id = m.canonical_id SET house_id = m.canonical_id
FROM _108_dup_mapping m FROM _108_dup_mapping m