Light1YT
af5e730438
All checks were successful
CI / frontend-tests (pull_request) Has been skipped
CI / changes (push) Successful in 7s
CI / frontend-tests (push) Has been skipped
CI / changes (pull_request) Successful in 7s
CI / openapi-codegen-check (push) Successful in 1m57s
CI / openapi-codegen-check (pull_request) Successful in 1m50s
CI / backend-tests (push) Successful in 8m53s
CI / backend-tests (pull_request) Successful in 8m51s
obj_pricing (OBJ-3 #1332) агрегировал objective_lots (2.1ГБ) по всем ~286 маппингам через GROUP BY domrf_obj_id, но LEFT JOIN потреблял только ~43 ближних (3км) obj_id — далёкие медианы считались и выбрасывались. Добавлен nearby_obj CTE (3км-множество считается раз), obj_pricing scoped через WHERE cm.domrf_obj_id IN (SELECT obj_id FROM nearby_obj), финальный SELECT идёт FROM nearby_obj → scope == потребляемое множество. Планировщик теперь bitmap index scan вместо full parallel seq scan. Прод EXPLAIN (центр ЕКБ): ~700мс→~290мс, objective_lots buffer reads 212658→20561 (~10x). Байт-идентично: per-obj_id агрегаты независимы, 0 расхождений на 43 ближних группах. Цена/fallback не тронуты (scope-only). Регрешн-гард против отката к full-table агрегации. Refs #70 |
||
|---|---|---|
| .. | ||
| __init__.py | ||
| admin_cadastre.py | ||
| admin_etl.py | ||
| admin_jobs.py | ||
| admin_leads.py | ||
| admin_scrape.py | ||
| admin_weight_profiles.py | ||
| analytics.py | ||
| chat.py | ||
| concepts.py | ||
| custom_pois.py | ||
| insights.py | ||
| landing.py | ||
| locations.py | ||
| market.py | ||
| me.py | ||
| own_projects.py | ||
| parcels.py | ||
| photos.py | ||
| pilot.py | ||
| ping.py | ||
| trade_in.py | ||
| users.py | ||