T8a: POST /scrape/cian-price-history — fetches Cian detail pages (curl_cffi), extracts priceChanges from _cianConfig defaultState, upserts offer_price_history. Wraps existing cian_detail.fetch_detail + save_detail_enrichment. T10: POST /scrape/yandex-address-backfill — enriches yandex listings.address with house number from detail-page <title> regex; resets geocode_tried_at for re-geocoding. Targets active EKB listings without a house number in address. T7: POST /scrape/house-imv-backfill — bulk Avito IMV evaluation per house. Picks median lot-params from linked listings, calls evaluate_via_imv, persists house_imv_evaluations + house_placement_history + house_suggestions. Resumable via imv_status markers. Runs in foreground (BackgroundTasks for long batches can be added later if needed). All three: NOT wired to scheduler (anti-rate-limit, intentional manual trigger). All three: idempotent / graceful per-item error handling + counters response. 27 unit tests added in tests/test_backfill_wave2.py (all passing).
218 lines
7.4 KiB
Python
218 lines
7.4 KiB
Python
"""T10: Yandex listing address backfill from detail-page <title>.
|
|
|
|
Yandex SERP cards contain only street-level addresses (e.g. «улица Горького»).
|
|
The offer detail page <title> has the full address including house number:
|
|
«Екатеринбург, улица Горького, 36 — id 12345».
|
|
|
|
This service finds active yandex listings without a house number in `address`,
|
|
fetches each detail page via curl_cffi(chrome120), extracts the full address
|
|
from the <title> regex, and UPDATE listings.address.
|
|
|
|
Deliberately NOT wired into scheduler — manual trigger only.
|
|
Triggered via POST /api/v1/admin/scrape/yandex-address-backfill.
|
|
"""
|
|
|
|
from __future__ import annotations
|
|
|
|
import asyncio
|
|
import logging
|
|
import re
|
|
import time
|
|
from dataclasses import dataclass, field
|
|
|
|
from curl_cffi.requests import AsyncSession
|
|
from sqlalchemy import text
|
|
from sqlalchemy.orm import Session
|
|
|
|
logger = logging.getLogger(__name__)
|
|
|
|
# Extract 'Екатеринбург, <full address>' from <title> — matches:
|
|
# «Продажа квартиры — Екатеринбург, улица Горького, 36 — id 7654321 на Яндекс.Недвижимости»
|
|
# Group 1 = «Екатеринбург, улица Горького, 36» (stripped of trailing punctuation/spaces)
|
|
_RE_TITLE_ADDRESS = re.compile(
|
|
r"<title>[^<]*?—\s*"
|
|
r"(?:[^—,]*?,\s*)?" # optional ЖК/жилой комплекс prefix
|
|
r"(Екатеринбург,\s*[^<—]+?)"
|
|
r"\s*—\s*id\s*\d+[^<]*</title>", # allow trailing text after id NNN (e.g. 'на Яндекс...')
|
|
re.IGNORECASE,
|
|
)
|
|
|
|
# Listings without a house number: address contains no digit after comma-space
|
|
# (e.g. «улица Горького» but not «улица Горького, 36»).
|
|
# This regex matches a number right after «, » (house number pattern).
|
|
_RE_HAS_HOUSE_NUMBER = re.compile(r",\s*\d+")
|
|
|
|
|
|
@dataclass
|
|
class YandexAddressBackfillResult:
|
|
checked: int = 0
|
|
saved: int = 0
|
|
skipped: int = 0
|
|
errors: int = 0
|
|
duration_sec: float = field(default=0.0)
|
|
|
|
|
|
def _extract_address_from_title(html: str) -> str | None:
|
|
"""Extract full address from Yandex offer detail page <title>."""
|
|
html_clean = html.replace("\xa0", " ")
|
|
m = _RE_TITLE_ADDRESS.search(html_clean)
|
|
if not m:
|
|
return None
|
|
addr = m.group(1).strip().rstrip(",").strip()
|
|
return addr if addr else None
|
|
|
|
|
|
def _eligible_listings(db: Session, limit: int) -> list[dict]:
|
|
"""Active EKB yandex listings without house number in address."""
|
|
rows = (
|
|
db.execute(
|
|
text("""
|
|
SELECT id, source_id, address, source_url
|
|
FROM listings
|
|
WHERE source = 'yandex'
|
|
AND region_code = 66
|
|
AND is_active = TRUE
|
|
AND address IS NOT NULL
|
|
AND NOT (address ~ ',\\s*\\d+')
|
|
AND source_url IS NOT NULL
|
|
ORDER BY scraped_at DESC
|
|
LIMIT :lim
|
|
"""),
|
|
{"lim": limit},
|
|
)
|
|
.mappings()
|
|
.all()
|
|
)
|
|
return [dict(r) for r in rows]
|
|
|
|
|
|
async def backfill_yandex_addresses(
|
|
db: Session,
|
|
*,
|
|
limit: int = 200,
|
|
request_delay_sec: float = 3.0,
|
|
) -> YandexAddressBackfillResult:
|
|
"""Fetch Yandex detail pages and update address with full street+house number.
|
|
|
|
Args:
|
|
db: SQLAlchemy session.
|
|
limit: max listings to process.
|
|
request_delay_sec: sleep between requests (anti-bot; default 3.0s).
|
|
|
|
Returns:
|
|
YandexAddressBackfillResult with checked/saved/skipped/errors counters.
|
|
"""
|
|
from app.core.config import settings
|
|
|
|
result = YandexAddressBackfillResult()
|
|
t0 = time.time()
|
|
|
|
items = _eligible_listings(db, limit)
|
|
result.checked = len(items)
|
|
|
|
if not items:
|
|
logger.info("yandex_address_backfill: no eligible listings")
|
|
result.duration_sec = time.time() - t0
|
|
return result
|
|
|
|
logger.info(
|
|
"yandex_address_backfill: %d listings, delay=%.1fs",
|
|
result.checked,
|
|
request_delay_sec,
|
|
)
|
|
|
|
_proxy_url = settings.scraper_proxy_url
|
|
_proxies = {"http": _proxy_url, "https": _proxy_url} if _proxy_url else None
|
|
|
|
async with AsyncSession(
|
|
impersonate="chrome120",
|
|
timeout=30.0,
|
|
proxies=_proxies,
|
|
headers={
|
|
"Accept-Language": "ru-RU,ru;q=0.9,en;q=0.8",
|
|
},
|
|
) as session:
|
|
for i, item in enumerate(items):
|
|
lid: int = item["id"]
|
|
old_addr: str = item["address"]
|
|
url: str = item["source_url"]
|
|
|
|
try:
|
|
resp = await session.get(url, allow_redirects=True)
|
|
except Exception as exc:
|
|
logger.warning("yandex_address_backfill: fetch failed listing_id=%s: %s", lid, exc)
|
|
result.errors += 1
|
|
await asyncio.sleep(request_delay_sec)
|
|
continue
|
|
|
|
if resp.status_code != 200:
|
|
logger.warning(
|
|
"yandex_address_backfill: HTTP %d listing_id=%s", resp.status_code, lid
|
|
)
|
|
result.errors += 1
|
|
await asyncio.sleep(request_delay_sec)
|
|
continue
|
|
|
|
new_addr = _extract_address_from_title(resp.text)
|
|
|
|
if new_addr is None:
|
|
logger.debug("yandex_address_backfill: no address extracted listing_id=%s", lid)
|
|
result.skipped += 1
|
|
elif new_addr == old_addr:
|
|
logger.debug(
|
|
"yandex_address_backfill: unchanged listing_id=%s addr=%r", lid, old_addr
|
|
)
|
|
result.skipped += 1
|
|
elif not _RE_HAS_HOUSE_NUMBER.search(new_addr):
|
|
# Extracted address still has no house number — skip to avoid overwrite
|
|
# with another street-only value.
|
|
logger.debug(
|
|
"yandex_address_backfill: still no house number listing_id=%s addr=%r",
|
|
lid,
|
|
new_addr,
|
|
)
|
|
result.skipped += 1
|
|
else:
|
|
logger.info(
|
|
"yandex_address_backfill: updating listing_id=%s %r -> %r",
|
|
lid,
|
|
old_addr,
|
|
new_addr,
|
|
)
|
|
try:
|
|
db.execute(
|
|
text("""
|
|
UPDATE listings
|
|
SET address = :addr,
|
|
geocode_tried_at = NULL
|
|
WHERE id = CAST(:id AS bigint)
|
|
"""),
|
|
{"addr": new_addr, "id": lid},
|
|
)
|
|
db.commit()
|
|
result.saved += 1
|
|
except Exception as exc:
|
|
logger.warning(
|
|
"yandex_address_backfill: DB update failed listing_id=%s: %s",
|
|
lid,
|
|
exc,
|
|
)
|
|
result.errors += 1
|
|
try:
|
|
db.rollback()
|
|
except Exception:
|
|
pass
|
|
|
|
if i < len(items) - 1:
|
|
await asyncio.sleep(request_delay_sec)
|
|
|
|
result.duration_sec = time.time() - t0
|
|
logger.info(
|
|
"yandex_address_backfill done: checked=%d saved=%d skipped=%d errors=%d %.1fs",
|
|
result.checked,
|
|
result.saved,
|
|
result.skipped,
|
|
result.errors,
|
|
result.duration_sec,
|
|
)
|
|
return result
|