fix(tradein/scraper-kit): клампить diff_percent перед INSERT offer_price_history (#2277)
All checks were successful
CI / changes (pull_request) Successful in 9s
CI Trade-In / changes (pull_request) Successful in 10s
CI Trade-In / frontend-checks (pull_request) Has been skipped
CI / backend-tests (pull_request) Has been skipped
CI / frontend-tests (pull_request) Has been skipped
CI / openapi-codegen-check (pull_request) Has been skipped
CI Trade-In / backend-tests (pull_request) Successful in 1m40s

Прямые app-level INSERT-пути (domclick/cian kit-провайдеры + legacy cian_detail)
читают diff_percent из скрапленного HTML/JSON и не проходят через БД-триггер
из миграции 131 — гарбадж/экстремальное значение с редизайна source-страницы
переполняет NUMERIC(8,2) и спамит "numeric field overflow" в проде.
This commit is contained in:
bot-backend 2026-07-04 13:30:41 +03:00
parent ab250a132b
commit f814f222a1
6 changed files with 210 additions and 3 deletions

View file

@ -34,6 +34,24 @@ if TYPE_CHECKING:
logger = logging.getLogger(__name__)
# offer_price_history.diff_percent — NUMERIC(8,2) (миграция
# 131_fix_diff_percent_overflow.sql). Прямой INSERT ниже читает diff_percent из
# скрапленного Cian JSON и не проходит через БД-триггер record_listing_price_change()
# (тот срабатывает только на UPDATE listings) — гарбадж/экстремальное значение
# с редизайна source-страницы иначе переполняет колонку.
_DIFF_PERCENT_BOUND = 999999.99
def _clamp_diff_percent(value: float | int | None) -> float | None:
"""Клампит diff_percent в границы NUMERIC(8,2) (зеркало migration 131 trigger-clamp)."""
if value is None or isinstance(value, bool):
return None
try:
numeric = float(value)
except (TypeError, ValueError):
return None
return min(max(numeric, -_DIFF_PERCENT_BOUND), _DIFF_PERCENT_BOUND)
@dataclass
class DetailEnrichment:
@ -390,7 +408,7 @@ def save_detail_enrichment(db: Session, listing_id: int, enrichment: DetailEnric
"lid": listing_id,
"ct": change["change_time"],
"price": change["price_rub"],
"diff": change.get("diff_percent"),
"diff": _clamp_diff_percent(change.get("diff_percent")),
},
)

View file

@ -37,6 +37,7 @@ from unittest.mock import AsyncMock, MagicMock
import pytest
from scraper_kit.domclick_exceptions import DomClickBlockedError, DomClickParseError
from scraper_kit.offer_price_history import clamp_diff_percent
from scraper_kit.providers.domclick.detail import (
DomClickDetailEnrichment,
_extract_ssr_state,
@ -409,3 +410,94 @@ def test_save_detail_enrichment_none_passthrough() -> None:
assert params["living_area_m2"] is None
# пустой raw_extra → "{}" (merge no-op, не затирает)
assert params["raw_extra"] == "{}"
# ── save_detail_enrichment — diff_percent клампится перед INSERT (#2277) ──────
# offer_price_history.diff_percent — NUMERIC(8,2) (миграция 131). Прямой INSERT
# здесь не проходит через БД-триггер (тот срабатывает только на UPDATE listings) —
# garbage/экстремальный diff_percent из скрапленного HTML иначе переполняет колонку.
def _insert_diff_param(db: MagicMock) -> float | None:
insert_calls = [c for c in db.execute.call_args_list if "offer_price_history" in str(c[0][0])]
assert len(insert_calls) == 1
return insert_calls[0][0][1]["diff"]
def test_save_detail_enrichment_clamps_extreme_positive_diff() -> None:
db = MagicMock()
db.execute.return_value.rowcount = 1
ct = datetime(2026, 4, 1, tzinfo=UTC)
e = DomClickDetailEnrichment(
item_id="x",
source_url=_CARD_URL,
price_changes=[{"change_time": ct, "price_rub": 5000000, "diff_percent": 5000000}],
)
save_detail_enrichment(db, 1, e)
assert _insert_diff_param(db) == 999999.99
def test_save_detail_enrichment_clamps_extreme_negative_diff() -> None:
db = MagicMock()
db.execute.return_value.rowcount = 1
ct = datetime(2026, 4, 1, tzinfo=UTC)
e = DomClickDetailEnrichment(
item_id="x",
source_url=_CARD_URL,
price_changes=[{"change_time": ct, "price_rub": 5000000, "diff_percent": -5000000}],
)
save_detail_enrichment(db, 1, e)
assert _insert_diff_param(db) == -999999.99
def test_save_detail_enrichment_normal_diff_unchanged() -> None:
db = MagicMock()
db.execute.return_value.rowcount = 1
ct = datetime(2026, 4, 1, tzinfo=UTC)
e = DomClickDetailEnrichment(
item_id="x",
source_url=_CARD_URL,
price_changes=[{"change_time": ct, "price_rub": 5000000, "diff_percent": -2.5}],
)
save_detail_enrichment(db, 1, e)
assert _insert_diff_param(db) == -2.5
def test_save_detail_enrichment_none_diff_unchanged() -> None:
db = MagicMock()
db.execute.return_value.rowcount = 1
ct = datetime(2026, 4, 1, tzinfo=UTC)
e = DomClickDetailEnrichment(
item_id="x",
source_url=_CARD_URL,
price_changes=[{"change_time": ct, "price_rub": 5000000, "diff_percent": None}],
)
save_detail_enrichment(db, 1, e)
assert _insert_diff_param(db) is None
# ── clamp_diff_percent (scraper_kit.offer_price_history) — standalone unit ────
# Нет отдельной tests/ директории для scraper-kit-пакета (см. audit перед фиксом) —
# прямые unit-тесты хелпера живут здесь, рядом с save_detail_enrichment-тестами,
# которые его же и используют.
def test_clamp_diff_percent_extreme_positive() -> None:
assert clamp_diff_percent(5000000) == 999999.99
def test_clamp_diff_percent_extreme_negative() -> None:
assert clamp_diff_percent(-5000000) == -999999.99
def test_clamp_diff_percent_in_range_unchanged() -> None:
assert clamp_diff_percent(-2.5) == -2.5
def test_clamp_diff_percent_none() -> None:
assert clamp_diff_percent(None) is None
def test_clamp_diff_percent_bool_treated_as_none() -> None:
assert clamp_diff_percent(True) is None
assert clamp_diff_percent(False) is None

View file

@ -608,6 +608,71 @@ def test_save_detail_enrichment_stamps_detail_enriched_at():
assert "detail_enriched_at = NOW()" in update_sql
# ── save_detail_enrichment — diff_percent клампится перед INSERT (#2277) ──────
# offer_price_history.diff_percent — NUMERIC(8,2) (миграция 131). Этот прямой
# INSERT не проходит через БД-триггер record_listing_price_change() (тот
# срабатывает только на UPDATE listings) — garbage/экстремальный diff_percent
# из скрапленного Cian JSON иначе переполняет колонку.
def _insert_diff_param(db: MagicMock) -> float | None:
insert_calls = [c for c in db.execute.call_args_list if "offer_price_history" in str(c[0][0])]
assert len(insert_calls) == 1
return insert_calls[0][0][1]["diff"]
def test_save_detail_enrichment_clamps_extreme_positive_diff():
db = MagicMock()
db.execute.return_value.fetchone.return_value = None
enrichment = DetailEnrichment(
price_changes=[
{"change_time": "2026-04-01T00:00:00Z", "price_rub": 5000000, "diff_percent": 5000000}
]
)
save_detail_enrichment(db, listing_id=2001, enrichment=enrichment)
assert _insert_diff_param(db) == 999999.99
def test_save_detail_enrichment_clamps_extreme_negative_diff():
db = MagicMock()
db.execute.return_value.fetchone.return_value = None
enrichment = DetailEnrichment(
price_changes=[
{
"change_time": "2026-04-01T00:00:00Z",
"price_rub": 5000000,
"diff_percent": -5000000,
}
]
)
save_detail_enrichment(db, listing_id=2002, enrichment=enrichment)
assert _insert_diff_param(db) == -999999.99
def test_save_detail_enrichment_normal_diff_unchanged():
db = MagicMock()
db.execute.return_value.fetchone.return_value = None
enrichment = DetailEnrichment(
price_changes=[
{"change_time": "2026-04-01T00:00:00Z", "price_rub": 5000000, "diff_percent": -2.5}
]
)
save_detail_enrichment(db, listing_id=2003, enrichment=enrichment)
assert _insert_diff_param(db) == -2.5
def test_save_detail_enrichment_none_diff_unchanged():
db = MagicMock()
db.execute.return_value.fetchone.return_value = None
enrichment = DetailEnrichment(
price_changes=[
{"change_time": "2026-04-01T00:00:00Z", "price_rub": 5000000, "diff_percent": None}
]
)
save_detail_enrichment(db, listing_id=2004, enrichment=enrichment)
assert _insert_diff_param(db) is None
# ── fetch_detail browser_fetcher param ───────────────────────────────────────

View file

@ -0,0 +1,30 @@
"""Клампинг diff_percent перед прямым INSERT в offer_price_history.
offer_price_history.diff_percent NUMERIC(8,2) (±999999.99, см. миграцию
131_fix_diff_percent_overflow.sql). Та миграция клампит diff_percent, который
СЧИТАЕТ БД-триггер record_listing_price_change() на UPDATE listings, но триггер
не срабатывает на прямые INSERT из provider-модулей (domclick/cian detail),
которые пишут diff_percent, взятый напрямую из скрапленного HTML/JSON внешнего
источника. Garbage/экстремальное значение с редизайна source-страницы иначе
переполняет колонку и спамит "numeric field overflow" в проде.
"""
from __future__ import annotations
_DIFF_PERCENT_BOUND = 999999.99
def clamp_diff_percent(value: float | int | None) -> float | None:
"""Clamp raw scraped diff_percent into offer_price_history's NUMERIC(8,2) bound.
Mirrors the DB-trigger clamp in migration 131_fix_diff_percent_overflow.sql
(LEAST(GREATEST(x, -999999.99), 999999.99)) so direct application-level INSERTs
(bypassing that trigger) cannot overflow the column either.
"""
if value is None or isinstance(value, bool):
return None
try:
numeric = float(value)
except (TypeError, ValueError):
return None
return min(max(numeric, -_DIFF_PERCENT_BOUND), _DIFF_PERCENT_BOUND)

View file

@ -21,6 +21,7 @@ from sqlalchemy import text
from sqlalchemy.orm import Session
from scraper_kit.cian_state_parser import extract_all_states, extract_state
from scraper_kit.offer_price_history import clamp_diff_percent
from scraper_kit.providers._base import build_curl_cffi_session
from scraper_kit.providers._proxy import curl_proxy_url
from scraper_kit.repair_state_normalizer import (
@ -400,7 +401,7 @@ def save_detail_enrichment(db: Session, listing_id: int, enrichment: DetailEnric
"lid": listing_id,
"ct": change["change_time"],
"price": change["price_rub"],
"diff": change.get("diff_percent"),
"diff": clamp_diff_percent(change.get("diff_percent")),
},
)

View file

@ -56,6 +56,7 @@ from sqlalchemy import text
from sqlalchemy.orm import Session
from scraper_kit.domclick_exceptions import DomClickBlockedError, DomClickParseError
from scraper_kit.offer_price_history import clamp_diff_percent
from scraper_kit.repair_state_normalizer import (
infer_repair_state_from_text,
normalize_repair_state,
@ -572,7 +573,7 @@ def save_detail_enrichment(
"lid": listing_id,
"ct": ct,
"price": price,
"diff": change.get("diff_percent"),
"diff": clamp_diff_percent(change.get("diff_percent")),
},
)
except Exception as exc: