gendesign/tradein-mvp/backend/app/services/scrapers/base.py
bot-backend b736d7b7e0 feat(estimator): exclude city-centroid listings from radius analogs (Refs #769 Part E)
Закрывает последнюю часть #769 (A1/A2/B/C/D в #798/#804). Finding #17:
bare-city адреса геокодились в city-центроид и сохранялись как точные lat/lon →
mislocated листинги участвовали в radius ST_DWithin-аналогах.

- 089: ALTER TABLE listings ADD COLUMN IF NOT EXISTS geo_precision text (idempot).
- ScrapedLot.geo_precision + проброс в save_listings.
- geocode_missing.py: geo_precision='city' через существующий _geocode_is_coarse.
- estimator _fetch_analogs Tier H+W: + AND (geo_precision IS DISTINCT FROM 'city').
  NULL проходит (консервативно, сдвига до backfill нет). Pricing не тронут.

7 тестов; 270 passed, ruff clean.
2026-05-30 21:22:21 +03:00

490 lines
22 KiB
Python
Raw Blame History

This file contains ambiguous Unicode characters

This file contains Unicode characters that might be confused with other characters. If you think that this is intentional, you can safely ignore this warning. Use the Escape button to reveal them.

"""Базовый framework для парсеров объявлений (Avito / Cian / DomKlik / ...).
Общие задачи:
- httpx.AsyncClient с realistic browser headers + UA rotation
- tenacity retry с exp backoff
- Sleep между запросами (anti-ban)
- ScrapedLot — единая Pydantic схема всех источников
- Запись в `listings` Postgres с дедупом по dedup_hash
- После INSERT/UPDATE — hook в matching service для регистрации в
`listing_sources` и привязки к каноническому `houses` через `house_sources`.
Каждый конкретный парсер (avito.py, cian.py, ...) наследуется от BaseScraper
и реализует `_fetch_page()` и `_parse_listings()`.
"""
from __future__ import annotations
import asyncio
import hashlib
import logging
import random
from abc import ABC, abstractmethod
from datetime import date
from typing import Any
import httpx
from pydantic import BaseModel, Field
from sqlalchemy import text
from sqlalchemy.orm import Session
from tenacity import retry, stop_after_attempt, wait_exponential
from app.services.matching import match_or_create_house, upsert_listing_source
from app.services.scrapers.snapshot_writer import upsert_listing_snapshot
logger = logging.getLogger(__name__)
# ── Realistic browser User-Agents (Apr 2026 versions) ──────────────────────
_USER_AGENTS: list[str] = [
# Chrome on macOS
"Mozilla/5.0 (Macintosh; Intel Mac OS X 14_5) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/132.0.0.0 Safari/537.36", # noqa: E501
# Chrome on Windows
"Mozilla/5.0 (Windows NT 10.0; Win64; x64) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/132.0.0.0 Safari/537.36", # noqa: E501
# Firefox on macOS
"Mozilla/5.0 (Macintosh; Intel Mac OS X 14.5; rv:127.0) Gecko/20100101 Firefox/127.0",
# Safari on macOS
"Mozilla/5.0 (Macintosh; Intel Mac OS X 14_5) AppleWebKit/605.1.15 (KHTML, like Gecko) Version/17.5 Safari/605.1.15", # noqa: E501
# Edge on Windows
"Mozilla/5.0 (Windows NT 10.0; Win64; x64) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/132.0.0.0 Safari/537.36 Edg/132.0.0.0", # noqa: E501
]
def random_user_agent() -> str:
return random.choice(_USER_AGENTS)
# ── Унифицированная схема результата ────────────────────────────────────────
class ScrapedLot(BaseModel):
"""Результат парсинга одного объявления.
Все источники приводятся к этому виду перед записью в Postgres.
Поля optional если источник их не отдаёт.
"""
source: str # 'avito' / 'cian' / 'domklik' / ...
source_url: str # ссылка на оригинал — кликается из UI
source_id: str | None = None # внутренний id источника (для update)
# Локация — должна быть хотя бы address ИЛИ (lat, lon)
address: str | None = None
lat: float | None = None
lon: float | None = None
# Параметры квартиры
rooms: int | None = None # 0 = студия
area_m2: float | None = None
floor: int | None = None
total_floors: int | None = None
year_built: int | None = None
house_type: str | None = None
repair_state: str | None = None
has_balcony: bool | None = None
kadastr_num: str | None = None
# Avito house linking (Stage 2a — search → houses)
house_source: str | None = None # 'avito' / 'cian'
house_ext_id: str | None = None # Avito's '3171365'
house_url: str | None = None # full URL of /catalog/houses/...
listing_segment: str | None = None # 'vtorichka' / 'novostroyki'
# Цена (обязательно)
price_rub: int = Field(gt=0)
price_per_m2: int | None = None
# Геокодинг
# None → точность неизвестна (координаты пришли из скрейпера напрямую).
# 'city' → геокодер вернул city-centroid (нет номера дома в адресе) —
# листинг исключается из radius-аналогов estimator'а.
geo_precision: str | None = None
# Метаданные
listing_date: date | None = None
days_on_market: int | None = None
photo_urls: list[str] = Field(default_factory=list)
raw_payload: dict[str, Any] | None = None
# Cian-specific extensions (Stage 2 of CianScraper v1)
living_area_m2: float | None = None
bedrooms_count: int | None = None
balconies_count: int | None = None
loggias_count: int | None = None
description_minhash: str | None = None
cadastral_number: str | None = None
building_cadastral_number: str | None = None
phones: list[dict] = Field(default_factory=list)
is_homeowner: bool | None = None
is_pro_seller: bool | None = None
bargain_allowed: bool | None = None
sale_type: str | None = None
metro_stations: list[dict] = Field(default_factory=list)
def compute_dedup_hash(self) -> str:
"""SHA256(source + stable_key) — стабильный uniqueness key.
stable_key = source_id (Cian offer_id 330200428, Avito data-item-id) если
есть; иначе source_url БЕЗ query-string. Cian/Avito URL несут волатильный
`?context=...` токен — без strip каждый ре-скрейп давал бы новый hash.
#753: price_rub в ключ НЕ входит. Смена цены того же объявления — это UPDATE
существующей строки (`ON CONFLICT (dedup_hash) DO UPDATE`), а не новый объект.
Раньше цена была частью ключа → каждое изменение цены порождало дубль
активного listing → искажение asking_to_sold_ratio / sales-vs-listings /
активного инвентаря. Смена формулы разово даёт новый hash существующим
строкам — следующий скрейп upsert'нет их по новому ключу (всплеск разовый).
"""
h = hashlib.sha256()
h.update(self.source.encode())
h.update(b"|")
key = self.source_id if self.source_id else (self.source_url or "").split("?")[0]
h.update((key or "").encode())
return h.hexdigest()
def compute_price_per_m2(self) -> int | None:
"""price_per_m2 = price_rub / area_m2 если area задана."""
if self.area_m2 and self.area_m2 > 0:
return int(self.price_rub / self.area_m2)
return None
# ── BaseScraper ─────────────────────────────────────────────────────────────
class BaseScraper(ABC):
"""Базовый класс — наследовать для каждого источника.
Subclass должен реализовать:
- `name` (class attr) — 'avito' / 'cian' / ...
- `_fetch_page(client, ...)` — async GET к источнику
- `_parse_listings(html_or_json, ...)` — html → list[ScrapedLot]
"""
name: str = "base"
base_url: str = ""
request_delay_sec: float = 5.0 # sleep между запросами (anti-ban)
def __init__(self) -> None:
self._client: httpx.AsyncClient | None = None
async def __aenter__(self) -> BaseScraper:
self._client = httpx.AsyncClient(
timeout=20.0,
follow_redirects=True,
headers={
"User-Agent": random_user_agent(),
"Accept": "text/html,application/xhtml+xml,application/xml;q=0.9,*/*;q=0.8",
"Accept-Language": "ru-RU,ru;q=0.9,en;q=0.8",
"Accept-Encoding": "gzip, deflate, br",
"Connection": "keep-alive",
"Upgrade-Insecure-Requests": "1",
},
)
return self
async def __aexit__(self, *args: Any) -> None:
if self._client is not None:
await self._client.aclose()
@retry(stop=stop_after_attempt(3), wait=wait_exponential(multiplier=2, min=2, max=30))
async def _http_get(self, url: str, **kwargs: Any) -> httpx.Response:
"""GET с retry. Subclass-ы используют это вместо raw httpx."""
assert self._client is not None, "Use scraper as async context manager"
response = await self._client.get(url, **kwargs)
# 403/429 — ретрай поможет (рандомный UA в новом контексте), 5xx тоже
if response.status_code in {403, 429, 502, 503, 504}:
response.raise_for_status()
return response
async def sleep_between_requests(self) -> None:
"""Случайный jitter в районе self.request_delay_sec — чтобы парсер
не выглядел как detstvo-bot."""
jitter = random.uniform(0.7, 1.5)
await asyncio.sleep(self.request_delay_sec * jitter)
@abstractmethod
async def fetch_around(self, lat: float, lon: float, radius_m: int = 1000) -> list[ScrapedLot]:
"""Главный метод — собрать объявления вокруг (lat, lon) в радиусе radius_m.
Returns:
Список ScrapedLot. Может быть пустым.
"""
...
# ── Запись пачки результатов в Postgres ─────────────────────────────────────
def save_listings(
db: Session,
lots: list[ScrapedLot],
*,
run_id: int | None = None,
) -> tuple[int, int]:
"""Пишем list[ScrapedLot] в `listings` с upsert по dedup_hash.
Дополнительно пишет point-in-time snapshot в listings_snapshots (ON CONFLICT DO UPDATE
— берётся последний за день). Ошибка записи snapshot НЕ прерывает основной INSERT.
Args:
db: SQLAlchemy session.
lots: список распарсенных объявлений.
run_id: FK scrape_runs.id текущего run'а (опционально; None при ad-hoc вызовах).
Returns:
(inserted, updated) — counters для логов.
"""
if not lots:
return 0, 0
inserted = 0
updated = 0
matched = 0
match_failures = 0
for lot in lots:
ppm2 = lot.price_per_m2 or lot.compute_price_per_m2()
dedup = lot.compute_dedup_hash()
result = db.execute(
text(
"""
INSERT INTO listings (
source, source_url, source_id, dedup_hash,
address, lat, lon, region_code,
rooms, area_m2, floor, total_floors, year_built,
house_type, repair_state, has_balcony, kadastr_num,
house_source, house_ext_id, house_url, listing_segment,
price_rub, price_per_m2,
listing_date, days_on_market, photo_urls, raw_payload,
living_area_m2, bedrooms_count, balconies_count, loggias_count,
description_minhash, cadastral_number, building_cadastral_number,
phones, is_homeowner, is_pro_seller,
bargain_allowed, sale_type, metro_stations,
geo_precision,
scraped_at, last_seen_at
) VALUES (
:source, :source_url, :source_id, :dedup,
:address, :lat, :lon, 66,
:rooms, :area_m2, :floor, :total_floors, :year_built,
:house_type, :repair_state, :has_balcony, :kadastr,
:house_source, :house_ext_id, :house_url, :listing_segment,
:price_rub, :ppm2,
:listing_date, :days_on_market,
CAST(:photos AS jsonb),
CAST(:raw AS jsonb),
:living_area_m2, :bedrooms_count, :balconies_count, :loggias_count,
:description_minhash, :cadastral_number, :building_cadastral_number,
CAST(:phones AS jsonb), :is_homeowner, :is_pro_seller,
:bargain_allowed, :sale_type, CAST(:metro_stations AS jsonb),
:geo_precision,
NOW(), NOW()
)
ON CONFLICT (dedup_hash) DO UPDATE
SET last_seen_at = NOW(),
is_active = true,
-- если цена изменилась — обновляем
price_rub = EXCLUDED.price_rub,
price_per_m2 = EXCLUDED.price_per_m2,
-- Cian-specific: обновляем при каждом re-scrape
living_area_m2 = EXCLUDED.living_area_m2,
bedrooms_count = EXCLUDED.bedrooms_count,
balconies_count = EXCLUDED.balconies_count,
loggias_count = EXCLUDED.loggias_count,
description_minhash = EXCLUDED.description_minhash,
cadastral_number = EXCLUDED.cadastral_number,
building_cadastral_number = EXCLUDED.building_cadastral_number,
phones = EXCLUDED.phones,
is_homeowner = EXCLUDED.is_homeowner,
is_pro_seller = EXCLUDED.is_pro_seller,
bargain_allowed = EXCLUDED.bargain_allowed,
sale_type = EXCLUDED.sale_type,
metro_stations = EXCLUDED.metro_stations
RETURNING id, (xmax = 0) AS inserted
"""
),
{
"source": lot.source,
"source_url": lot.source_url,
"source_id": lot.source_id,
"dedup": dedup,
"address": lot.address,
"lat": lot.lat,
"lon": lot.lon,
"rooms": lot.rooms,
"area_m2": lot.area_m2,
"floor": lot.floor,
"total_floors": lot.total_floors,
"year_built": lot.year_built,
"house_type": lot.house_type,
"repair_state": lot.repair_state,
"has_balcony": lot.has_balcony,
"kadastr": lot.kadastr_num,
"house_source": lot.house_source,
"house_ext_id": lot.house_ext_id,
"house_url": lot.house_url,
"listing_segment": lot.listing_segment,
"price_rub": lot.price_rub,
"ppm2": ppm2,
"listing_date": lot.listing_date,
"days_on_market": lot.days_on_market,
"photos": _to_json(lot.photo_urls),
"raw": _to_json(lot.raw_payload) if lot.raw_payload else None,
# Cian-specific columns — None for non-Cian scrapers (→ SQL NULL)
"living_area_m2": lot.living_area_m2,
"bedrooms_count": lot.bedrooms_count,
"balconies_count": lot.balconies_count,
"loggias_count": lot.loggias_count,
"description_minhash": lot.description_minhash,
"cadastral_number": lot.cadastral_number,
"building_cadastral_number": lot.building_cadastral_number,
"phones": _to_json(lot.phones) if lot.phones else None,
"is_homeowner": lot.is_homeowner,
"is_pro_seller": lot.is_pro_seller,
"bargain_allowed": lot.bargain_allowed,
"sale_type": lot.sale_type,
"metro_stations": _to_json(lot.metro_stations) if lot.metro_stations else None,
"geo_precision": lot.geo_precision,
},
).fetchone()
listing_id: int | None = None
if result is not None:
listing_id = int(result.id)
if result.inserted:
inserted += 1
else:
updated += 1
# ── Snapshot: point-in-time observation in listings_snapshots ───
# Fault-tolerant: failure here MUST NOT abort the listings batch.
# ON CONFLICT DO UPDATE — самый последний snapshot за день.
if listing_id is not None:
try:
with db.begin_nested():
upsert_listing_snapshot(
db,
listing_id=listing_id,
price_rub=lot.price_rub,
price_per_m2=ppm2,
run_id=run_id,
status="active",
)
except Exception as e:
logger.warning(
"save_listings:snapshot_failed source=%s listing_id=%s: %s",
lot.source,
listing_id,
e,
)
# ── Hook: link listing → house via matching service ─────────────
# Idempotent (UPSERT on listing_sources UNIQUE(ext_source, ext_id))
# and fault-tolerant: failure here MUST NOT abort the listings batch.
# Wrapped in SAVEPOINT so a failed match only rolls back its own work,
# not the surrounding INSERT (see .claude/rules/backend.md SAVEPOINT).
if listing_id is not None:
try:
with db.begin_nested():
_link_listing_to_house(db, listing_id, lot)
matched += 1
except Exception as e:
# Best-effort hook: log and continue so the listings batch isn't aborted.
match_failures += 1
logger.warning(
"save_listings:match_failed source=%s source_id=%s listing_id=%s: %s",
lot.source,
lot.source_id,
listing_id,
e,
)
db.commit()
logger.info(
"save_listings: source=%s inserted=%d updated=%d matched=%d "
"match_failures=%d (total %d)",
lots[0].source if lots else "?",
inserted,
updated,
matched,
match_failures,
len(lots),
)
return inserted, updated
def _to_json(value: Any) -> str:
"""JSON serialization helper — для jsonb колонок."""
import json
return json.dumps(value, ensure_ascii=False, default=str)
def _link_listing_to_house(db: Session, listing_id: int, lot: ScrapedLot) -> None:
"""Hook scraped listing into matching service: resolve house, upsert listing_sources.
Steps:
1. match_or_create_house() — find or create canonical houses row for this
listing's address/coords. Uses Tier 0-3 matching (cadastr → source_exact →
fingerprint → geo_proximity → new).
2. upsert_listing_source() — register (ext_source, ext_id) → listing_id in
listing_sources. Idempotent via UNIQUE (ext_source, ext_id).
The matching method recorded is 'source_link': scraper already deduped by
`dedup_hash` (INSERT … ON CONFLICT DO UPDATE), so this is a direct registration
rather than a fuzzy listing-level match.
ext_id source: lot.source_id if present, else dedup_hash (Yandex without
stable source_id falls back to URL-based dedup_hash — same hash on re-scrape).
Skips silently if:
- lot has no source_id AND no address/lat/lon (cannot match house anyway)
Raises on DB errors — caller wraps in try/except + SAVEPOINT.
"""
ext_id = lot.source_id or lot.compute_dedup_hash()
# House resolution: needs at least address or (lat, lon). Skip otherwise —
# listing_sources requires listing_id but not house linkage, so still upsert.
house_id: int | None = None
if lot.address or (lot.lat is not None and lot.lon is not None):
# Use house_source/house_ext_id when scraper extracted them (Avito Houses
# catalog link, Cian newbuilding). Falls back to per-listing identity
# so each unique listing creates its own row if no canonical house exists.
h_src = lot.house_source or lot.source
h_ext = lot.house_ext_id or ext_id
try:
with db.begin_nested():
house_id, _conf, _method = match_or_create_house(
db,
ext_source=h_src,
ext_id=h_ext,
address=lot.address,
lat=lot.lat,
lon=lot.lon,
year_built=lot.year_built,
building_cadastral_number=lot.building_cadastral_number,
cadastral_number=lot.cadastral_number or lot.kadastr_num,
source_url=lot.house_url or lot.source_url,
)
except Exception as e:
# Fall through to listing-only upsert: listing_sources row still useful
# even without house linkage (e.g. for later backfill).
logger.warning(
"save_listings:house_match_failed source=%s ext_id=%s: %s",
lot.source,
ext_id,
e,
)
house_id = None
upsert_listing_source(
db,
listing_id=listing_id,
ext_source=lot.source,
ext_id=str(ext_id),
method="source_link",
confidence=1.0,
price_rub=lot.price_rub,
area_m2=lot.area_m2,
floor=lot.floor,
rooms_count=lot.rooms,
source_url=lot.source_url,
source_data={"house_id": house_id} if house_id else None,
)