fix(sf-21): mv_layout_velocity weighted AVG (P0 data corruption) #290

Merged
lekss361 merged 1 commit from fix/sf-21-mv-weighted-avg into main 2026-05-17 14:02:46 +00:00

View file

@ -0,0 +1,88 @@
-- 100_fix_mv_layout_velocity_weighted_avg.sql
-- SF Bug #21 (P0) — mv_layout_velocity использовал unweighted AVG() по
-- deals_total_avg_area_m2 и deals_total_avg_price_thousand_rub_per_m2.
-- objective_corpus_room_month хранит строки за каждый (project, room, month)
-- включая месяцы с 0 сделок (count=0, area=0, price=0). Нули тянули
-- среднее вниз в 2-14×. Примеры (verified prod):
-- obj 65307, room=3: 114 месяцев, 106 нулевых → AVG=7.4 м² vs weighted=105.8 м²
-- obj 65903, room=4+: 36 мес, 33 нулевых → AVG=9.1 м² vs weighted=109.0 м²
--
-- Fix: заменяем AVG(x) на SUM(x * count) / NULLIF(SUM(count), 0) — weighted average.
--
-- Dependencies: mv_layout_velocity не имеет зависимых VIEW/MV (проверено 2026-05-17).
-- Existing indexes: mv_layout_velocity_pk (UNIQUE obj_id, room_bucket),
-- mv_layout_velocity_obj_idx (obj_id) — пересоздаются ниже.
--
-- ВАЖНО: этот файл выполняет DROP + CREATE + REFRESH (non-concurrent, блокировка ~30 с).
-- REFRESH CONCURRENTLY после первого populate — безопасен через layout_velocity_refresh.py.
--
-- Deploy order: SQL migration → app deploy (никаких backend изменений в этом PR).
-- Idempotency: DROP IF EXISTS → безопасен при повторном запуске.
BEGIN;
DROP MATERIALIZED VIEW IF EXISTS mv_layout_velocity CASCADE;
CREATE MATERIALIZED VIEW mv_layout_velocity AS
WITH last24mo AS (
SELECT
ocm.project_name,
CASE
WHEN ocm.room_bucket = 'студия' THEN 'studio'
ELSE ocm.room_bucket
END AS room_bucket,
ocm.deals_total_count,
ocm.deals_total_avg_area_m2,
ocm.deals_total_avg_price_thousand_rub_per_m2,
ocm.deals_total_vol_m2,
ocm.report_month
FROM objective_corpus_room_month ocm
WHERE ocm.report_month >= (NOW() - INTERVAL '24 months')::date
)
SELECT
cm.domrf_obj_id AS obj_id,
l.room_bucket,
SUM(l.deals_total_count)::int AS total_deals_24mo,
-- FIX SF #21: weighted average — нули из месяцев без сделок не тянут вниз.
-- Формула: SUM(avg_x * count) / SUM(count) эквивалентна SUM(total_x) / SUM(count),
-- т.к. avg_x = total_x / count за один месяц → произведение восстанавливает total_x.
(SUM(l.deals_total_avg_area_m2 * l.deals_total_count)
/ NULLIF(SUM(l.deals_total_count), 0))::numeric(10, 2) AS avg_area_m2,
(SUM(l.deals_total_avg_price_thousand_rub_per_m2 * l.deals_total_count)
/ NULLIF(SUM(l.deals_total_count), 0))::numeric(12, 2) AS avg_price_thousand_rub_per_m2,
SUM(l.deals_total_vol_m2)::numeric(12, 2) AS total_vol_m2,
MIN(l.report_month) AS window_start,
MAX(l.report_month) AS window_end,
COUNT(DISTINCT l.report_month)::int AS months_with_data
FROM last24mo l
JOIN objective_complex_mapping cm
ON cm.objective_complex_name = l.project_name
WHERE l.room_bucket IS NOT NULL
AND cm.domrf_obj_id IS NOT NULL
AND cm.objective_group = 'Екатеринбург'
GROUP BY cm.domrf_obj_id, l.room_bucket
WITH NO DATA;
-- UNIQUE index: required for REFRESH CONCURRENTLY (periodic via layout_velocity_refresh.py).
-- GROUP BY (obj_id, room_bucket) guarantees uniqueness.
CREATE UNIQUE INDEX mv_layout_velocity_pk
ON mv_layout_velocity (obj_id, room_bucket);
-- Lookup index for /best-layouts endpoint queries by obj_id.
CREATE INDEX mv_layout_velocity_obj_idx
ON mv_layout_velocity (obj_id);
-- Initial populate (non-concurrent — MV just created, CONCURRENTLY requires populated MV).
-- Блокировка ~30 с на production при ~19 738 source rows.
REFRESH MATERIALIZED VIEW mv_layout_velocity;
COMMENT ON MATERIALIZED VIEW mv_layout_velocity IS
'Per-(obj_id, room_bucket) deals aggregation за last 24 months. '
'WEIGHTED average площади и цены (SF Bug #21 fix, 2026-05-17). '
'Source: objective_corpus_room_month × objective_complex_mapping (EKB only). '
'Refresh via layout_velocity_refresh.py (CONCURRENTLY после initial populate).';
COMMIT;