diff --git a/tradein-mvp/backend/app/api/v1/admin.py b/tradein-mvp/backend/app/api/v1/admin.py
index 1982a79e..e771642c 100644
--- a/tradein-mvp/backend/app/api/v1/admin.py
+++ b/tradein-mvp/backend/app/api/v1/admin.py
@@ -26,6 +26,7 @@ from app.services.scrape_pipeline import (
run_avito_city_sweep,
run_cian_city_sweep,
run_n1_city_sweep,
+ run_yandex_city_sweep,
)
from app.services.scraper_settings import invalidate_cache
from app.services.scrapers.avito import AvitoScraper
@@ -852,6 +853,93 @@ def cancel_cian_city_sweep(
return {"ok": True, "run_id": run_id, "cancelled": cancelled}
+# ── Yandex city sweep (#861) — on-demand trigger + runs list ─────────────────
+
+
+class YandexCitySweepStartRequest(BaseModel):
+ pages_per_anchor: int = Field(default=2, ge=1, le=10)
+ request_delay_sec: float = Field(default=7.0, ge=3.0, le=30.0)
+ enrich_address: bool = True
+ radius_m: int = Field(default=1500, ge=500, le=5000)
+
+
+@router.post("/scrape/yandex-city-sweep", response_model=CitySweepStartResponse)
+async def start_yandex_city_sweep(
+ payload: YandexCitySweepStartRequest,
+ background_tasks: BackgroundTasks,
+ db: Annotated[Session, Depends(get_db)],
+) -> CitySweepStartResponse:
+ """Запустить Yandex.Недвижимость ЕКБ city sweep в background. Returns run_id.
+
+ Один прогон: SERP (fetch_around_multi_room, rooms × price-range combos)
+ → save_listings → address-enrich (detail → полный адрес с домом).
+ Прокси уже внутри YandexRealtyScraper + address-enrich сессии.
+
+ Coop cancel: POST /scrape/yandex-city-sweep/{run_id}/cancel.
+ НЕ добавляется в scrape_schedules — ручной/контролируемый запуск
+ (schedule yandex_city_sweep DORMANT, включается оператором вручную).
+ """
+ run_id = runs_mod.create_run(db, source="yandex_city_sweep", params=payload.model_dump())
+
+ async def _sweep_task() -> None:
+ sweep_db = SessionLocal()
+ try:
+ await run_yandex_city_sweep(
+ sweep_db,
+ run_id=run_id,
+ radius_m=payload.radius_m,
+ pages_per_anchor=payload.pages_per_anchor,
+ request_delay_sec=payload.request_delay_sec,
+ enrich_address=payload.enrich_address,
+ )
+ except Exception:
+ logger.exception("yandex-sweep background task run_id=%d crashed", run_id)
+ finally:
+ sweep_db.close()
+
+ background_tasks.add_task(_sweep_task)
+ logger.info("yandex-sweep queued run_id=%d params=%s", run_id, payload.model_dump())
+ return CitySweepStartResponse(
+ run_id=run_id,
+ status="running",
+ pages_per_anchor=payload.pages_per_anchor,
+ detail_top_n=0, # у Yandex sweep нет detail-шага (address-enrich отдельно)
+ )
+
+
+@router.get("/scrape/yandex-city-sweep/runs", response_model=list[ScrapeRunRow])
+def list_yandex_city_sweep_runs(
+ db: Annotated[Session, Depends(get_db)],
+ limit: int = 10,
+) -> list[ScrapeRunRow]:
+ """Список последних N Yandex city sweep runs (для UI polling). Default limit=10."""
+ rows = runs_mod.list_recent(db, source="yandex_city_sweep", limit=limit)
+ return [
+ ScrapeRunRow(
+ run_id=r["run_id"],
+ source=r["source"],
+ status=r["status"],
+ params=r.get("params"),
+ counters=r.get("counters"),
+ error=r.get("error"),
+ started_at=r["started_at"].isoformat() if r.get("started_at") else None,
+ finished_at=r["finished_at"].isoformat() if r.get("finished_at") else None,
+ heartbeat_at=r["heartbeat_at"].isoformat() if r.get("heartbeat_at") else None,
+ )
+ for r in rows
+ ]
+
+
+@router.post("/scrape/yandex-city-sweep/{run_id}/cancel")
+def cancel_yandex_city_sweep(
+ run_id: int,
+ db: Annotated[Session, Depends(get_db)],
+) -> dict[str, object]:
+ """Отменить running Yandex sweep. Cooperative: проверяется каждый anchor."""
+ cancelled = runs_mod.mark_cancelled(db, run_id)
+ return {"ok": True, "run_id": run_id, "cancelled": cancelled}
+
+
# ── Scraper settings: global + per-source delay management ───────────────────
diff --git a/tradein-mvp/backend/app/services/scheduler.py b/tradein-mvp/backend/app/services/scheduler.py
index 413232a7..1bbbadab 100644
--- a/tradein-mvp/backend/app/services/scheduler.py
+++ b/tradein-mvp/backend/app/services/scheduler.py
@@ -253,6 +253,7 @@ async def trigger_yandex_city_sweep_run(db: Session, schedule_row: dict[str, Any
pages_per_anchor=int(params.get("pages_per_anchor", 2)),
request_delay_sec=float(params.get("request_delay_sec", 9.0)),
radius_m=int(params.get("radius_m", 1500)),
+ enrich_address=bool(params.get("enrich_address", True)),
)
except Exception:
logger.exception("scheduler: run_yandex_city_sweep crashed run_id=%d", run_id)
diff --git a/tradein-mvp/backend/app/services/scrape_pipeline.py b/tradein-mvp/backend/app/services/scrape_pipeline.py
index bd33cfee..2a81101a 100644
--- a/tradein-mvp/backend/app/services/scrape_pipeline.py
+++ b/tradein-mvp/backend/app/services/scrape_pipeline.py
@@ -538,16 +538,18 @@ async def run_avito_city_sweep(
class YandexCitySweepCounters:
"""Aggregate counters для Yandex.Недвижимость city sweep run.
- Проще avito CitySweepCounters: у Yandex SERP нет house-catalog / detail-enrich
- шага (только search → save), поэтому houses/detail-полей нет.
+ Фазы: SERP (fetch_around_multi_room) → save_listings → address-enrich.
+ address_* — результат T10-обогащения (detail <title> → полный адрес с номером дома).
"""
anchors_total: int = 0
anchors_done: int = 0
- pages_fetched: int = 0
lots_fetched: int = 0
lots_inserted: int = 0
lots_updated: int = 0
+ address_attempted: int = 0
+ address_enriched: int = 0
+ address_failed: int = 0
errors_count: int = 0
def to_dict(self) -> dict[str, int]:
@@ -569,33 +571,50 @@ async def run_yandex_city_sweep(
radius_m: int = 1500,
pages_per_anchor: int = 2,
request_delay_sec: float | None = None,
+ enrich_address: bool = True,
) -> YandexCitySweepCounters:
- """Yandex.Недвижимость city sweep: iterate anchors × pages → save.
+ """Yandex.Недвижимость city sweep: SERP → save → address-enrich.
- Проще avito sweep — у Yandex SERP нет house-catalog / detail-enrich шага:
- per anchor: YandexRealtyScraper().fetch_around(lat, lon, radius_m, page=p)
- → save_listings(db, lots) → accumulate counters.
+ Фазы per anchor:
+ 1. SERP: YandexRealtyScraper.fetch_around_multi_room(lat, lon, radius_m,
+ max_pages=pages_per_anchor) — rooms × price-range combos, dedup внутри.
+ 2. SAVE: save_listings(db, lots, run_id=run_id).
+ 3. ADDRESS-ENRICH (если enrich_address=True): для yandex-листингов этого anchor'а
+ без номера дома в address (address IS NOT NULL AND NOT (address ~ ',\\s*\\d+'))
+ запрашиваем detail-страницу и извлекаем полный адрес из <title>.
+ Переиспользует приватные функции yandex_address_backfill:
+ _eligible_listings_for_ids → _fetch_and_update_one.
- Anti-bot / cancel семантика (зеркало avito sweep, см. #561):
- - Cooperative cancel: is_cancelled(db, run_id) проверяется перед каждым anchor;
- при cancel — update_heartbeat + return (mark_cancelled делает caller-flow по статусу).
- - YandexRealtyScraper НЕ определяет block/rate exception классов (fetch_around
- ловит исключения внутри и возвращает []). Поэтому per-anchor ловим broad
- Exception, логируем и continue; защитно прерываем sweep (early abort + mark_banned),
- если N anchor'ов ПОДРЯД упали с исключением — это сигнал системного блока IP.
- - request_delay_sec: YandexRealtyScraper.fetch_around НЕ принимает delay-override
- (он сам спит self.request_delay_sec между страницами через sleep_between_requests).
- Поэтому request_delay_sec применяется здесь как доп. asyncio.sleep МЕЖДУ anchor'ами.
+ Anti-bot / cancel семантика:
+ - Cooperative cancel: is_cancelled(db, run_id) проверяется перед каждым anchor.
+ - YandexRealtyScraper не propagate'ит block/rate exceptions (глотает → []).
+ Поэтому per-anchor ловим broad Exception; при N подряд — mark_banned + abort.
+ - request_delay_sec: доп. asyncio.sleep МЕЖДУ anchor'ами (поверх per-page sleep
+ внутри scraper'а). Также используется как inter-request delay для address-enrich.
Возвращает YandexCitySweepCounters (агрегат по всем anchor'ам).
+ mark_done вызывается ВСЕГДА (кроме cancel / ban / fatal).
"""
+ from sqlalchemy import text as _text
+
+ from app.services.yandex_address_backfill import (
+ _RE_HAS_HOUSE_NUMBER,
+ _extract_address_from_title,
+ )
+
_anchors = anchors if anchors is not None else EKB_ANCHORS
counters = YandexCitySweepCounters(anchors_total=len(_anchors))
inter_anchor_delay = request_delay_sec if request_delay_sec is not None else 7.0
+ enrich_delay = request_delay_sec if request_delay_sec is not None else 3.0
consecutive_failures = 0
+ # Прокси для address-enrich curl_cffi сессии (зеркало yandex_address_backfill).
+ _proxy_url = settings.scraper_proxy_url
+ _proxies = {"http": _proxy_url, "https": _proxy_url} if _proxy_url else None
+
try:
for idx, (lat, lon, name) in enumerate(_anchors, start=1):
+ # ── Cooperative cancel ───────────────────────────────────────────
if scrape_runs.is_cancelled(db, run_id):
logger.info(
"yandex-sweep run_id=%d: cancelled at anchor #%d/%d (%s)",
@@ -616,29 +635,34 @@ async def run_yandex_city_sweep(
lat,
lon,
)
- try:
- # Fresh scraper per anchor — uses its own httpx.AsyncClient via the
- # async-context-manager. fetch_around fetches ONE SERP page; loop pages.
- anchor_lots: list[ScrapedLot] = []
- async with YandexRealtyScraper() as scraper:
- for page in range(pages_per_anchor):
- page_lots = await scraper.fetch_around(lat, lon, radius_m, page=page)
- counters.pages_fetched += 1
- if not page_lots:
- # Пустая страница → дальше результатов нет, прекращаем пагинацию.
- break
- anchor_lots.extend(page_lots)
+ # ── Phase 1+2: SERP + save ───────────────────────────────────────
+ anchor_lots: list[ScrapedLot] = []
+ try:
+ async with YandexRealtyScraper() as scraper:
+ anchor_lots = await scraper.fetch_around_multi_room(
+ lat,
+ lon,
+ radius_m,
+ max_pages=pages_per_anchor,
+ )
counters.lots_fetched += len(anchor_lots)
if anchor_lots:
inserted, updated = save_listings(db, anchor_lots, run_id=run_id)
counters.lots_inserted += inserted
counters.lots_updated += updated
consecutive_failures = 0
+ logger.info(
+ "yandex-sweep run_id=%d anchor %s: SERP fetched=%d ins=%d upd=%d",
+ run_id,
+ name,
+ len(anchor_lots),
+ counters.lots_inserted,
+ counters.lots_updated,
+ )
except Exception as e:
- # YandexRealtyScraper не propagate'ит block/rate exceptions, поэтому здесь
- # ловим broad Exception per-anchor: логируем и continue (изоляция anchor'а).
- logger.exception("yandex-sweep run_id=%d: anchor %s failed", run_id, name)
+ # YandexRealtyScraper не propagate'ит block/rate exceptions — ловим broad.
+ logger.exception("yandex-sweep run_id=%d: anchor %s SERP failed", run_id, name)
counters.errors_count += 1
consecutive_failures += 1
if consecutive_failures >= YANDEX_SWEEP_MAX_CONSECUTIVE_FAILURES:
@@ -662,6 +686,128 @@ async def run_yandex_city_sweep(
counters.to_dict(),
)
return counters
+ counters.anchors_done = idx
+ scrape_runs.update_heartbeat(db, run_id, counters.to_dict())
+ # Пауза перед следующим anchor'ом даже при ошибке.
+ if idx < len(_anchors):
+ await asyncio.sleep(inter_anchor_delay)
+ continue
+
+ # ── Phase 3: address-enrich ─────────────────────────────────────
+ if enrich_address and anchor_lots:
+ # Собираем source_id из SERP-результатов этого anchor'а → ищем
+ # в DB только листинги, которые были только что сохранены/обновлены.
+ source_ids = [lot.source_id for lot in anchor_lots if lot.source_id]
+ if source_ids:
+ id_rows = (
+ db.execute(
+ _text("""
+ SELECT id, source_id, address, source_url
+ FROM listings
+ WHERE source = 'yandex'
+ AND source_id = ANY(CAST(:ids AS text[]))
+ AND address IS NOT NULL
+ AND NOT (address ~ ',\\s*\\d+')
+ AND source_url IS NOT NULL
+ """),
+ {"ids": source_ids},
+ )
+ .mappings()
+ .all()
+ )
+ items = [dict(r) for r in id_rows]
+ counters.address_attempted += len(items)
+
+ if items:
+ from curl_cffi.requests import AsyncSession as _AsyncSession
+
+ async with _AsyncSession(
+ impersonate="chrome120",
+ timeout=30.0,
+ proxies=_proxies,
+ headers={"Accept-Language": "ru-RU,ru;q=0.9,en;q=0.8"},
+ ) as enrich_session:
+ for eidx, item in enumerate(items):
+ lid: int = item["id"]
+ old_addr: str = item["address"]
+ url: str = item["source_url"]
+ try:
+ resp = await enrich_session.get(url, allow_redirects=True)
+ if resp.status_code != 200:
+ logger.warning(
+ "yandex-sweep run_id=%d address-enrich:"
+ " HTTP %d listing_id=%d",
+ run_id,
+ resp.status_code,
+ lid,
+ )
+ counters.address_failed += 1
+ else:
+ new_addr = _extract_address_from_title(resp.text)
+ if (
+ new_addr is None
+ or new_addr == old_addr
+ or not _RE_HAS_HOUSE_NUMBER.search(new_addr)
+ ):
+ # Не удалось улучшить адрес — пропускаем (не ошибка).
+ pass
+ else:
+ 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()
+ counters.address_enriched += 1
+ logger.info(
+ "yandex-sweep run_id=%d: "
+ "address updated listing_id=%d "
+ "%r -> %r",
+ run_id,
+ lid,
+ old_addr,
+ new_addr,
+ )
+ except Exception as upd_exc:
+ counters.address_failed += 1
+ counters.errors_count += 1
+ logger.warning(
+ "yandex-sweep run_id=%d: "
+ "address DB update failed "
+ "listing_id=%d: %s",
+ run_id,
+ lid,
+ upd_exc,
+ )
+ try:
+ db.rollback()
+ except Exception:
+ pass
+ except Exception as fetch_exc:
+ counters.address_failed += 1
+ logger.warning(
+ "yandex-sweep run_id=%d: address-enrich "
+ "fetch failed listing_id=%d: %s",
+ run_id,
+ lid,
+ fetch_exc,
+ )
+ if eidx < len(items) - 1:
+ await asyncio.sleep(enrich_delay)
+
+ logger.info(
+ "yandex-sweep run_id=%d anchor %s: " "address enrich=%d/%d failed=%d",
+ run_id,
+ name,
+ counters.address_enriched,
+ counters.address_attempted,
+ counters.address_failed,
+ )
counters.anchors_done = idx
scrape_runs.update_heartbeat(db, run_id, counters.to_dict())
@@ -672,14 +818,17 @@ async def run_yandex_city_sweep(
scrape_runs.mark_done(db, run_id, counters.to_dict())
logger.info(
- "yandex-sweep run_id=%d done: anchors=%d/%d pages=%d lots=%d (ins=%d/upd=%d) errors=%d",
+ "yandex-sweep run_id=%d done: anchors=%d/%d lots=%d (ins=%d/upd=%d) "
+ "address=%d/%d failed=%d errors=%d",
run_id,
counters.anchors_done,
counters.anchors_total,
- counters.pages_fetched,
counters.lots_fetched,
counters.lots_inserted,
counters.lots_updated,
+ counters.address_enriched,
+ counters.address_attempted,
+ counters.address_failed,
counters.errors_count,
)
return counters
diff --git a/tradein-mvp/backend/tests/test_yandex_city_sweep.py b/tradein-mvp/backend/tests/test_yandex_city_sweep.py
index 6aae9ae0..a8102b71 100644
--- a/tradein-mvp/backend/tests/test_yandex_city_sweep.py
+++ b/tradein-mvp/backend/tests/test_yandex_city_sweep.py
@@ -1,20 +1,31 @@
-"""Offline unit tests for run_yandex_city_sweep (#561, shipped DORMANT).
+"""Offline unit tests for run_yandex_city_sweep (SERP+address-enrich, #861).
Pure & fast: NO live network, NO DB. We monkeypatch:
- YandexRealtyScraper.__init__ / __aenter__ / __aexit__ — avoid get_scraper_delay
- DB read + real httpx.AsyncClient creation.
- - YandexRealtyScraper.fetch_around — return a fixed list of fake ScrapedLot.
+ DB read + real curl_cffi session creation.
+ - YandexRealtyScraper.fetch_around_multi_room — return a fixed list of fake ScrapedLot.
- scrape_pipeline.save_listings — counter/no-op (no real listings table).
- scrape_runs.* (is_cancelled / update_heartbeat / mark_*) — in-memory fakes.
+ - curl_cffi.requests.AsyncSession — fake for address-enrich HTTP calls.
-These tests assert orchestration semantics only; live Yandex behavior is intentionally
-unverified (sweep is dormant until egress-IP rotation #623).
+Asserts orchestration semantics:
+ - SERP → save (all anchors, counters aggregated)
+ - address-enrich фаза вызывается если enrich_address=True и есть source_ids
+ - enrich_address=False → адрес-фаза пропускается
+ - cancel → останавливает sweep
+ - N consecutive SERP failures → mark_banned + abort
+ - per-item address-enrich error не валит sweep
+ - mark_done вызывается при нормальном завершении
"""
+
+from __future__ import annotations
+
import os
os.environ.setdefault("DATABASE_URL", "postgresql+psycopg://test:test@localhost:5432/test")
from typing import Any
+from unittest.mock import AsyncMock, MagicMock
import pytest
@@ -22,7 +33,7 @@ from app.services import scrape_pipeline
from app.services.scrapers.base import ScrapedLot
from app.services.scrapers.yandex_realty import YandexRealtyScraper
-# 3 anchors keep tests fast and let us probe "anchor k cancels/fails, others proceed".
+# 3 anchors keep tests fast and let us probe cancel/fail semantics.
TEST_ANCHORS: list[tuple[float, float, str]] = [
(56.8400, 60.6050, "A1"),
(56.7950, 60.5300, "A2"),
@@ -36,7 +47,7 @@ def _fake_lot(offer_id: str) -> ScrapedLot:
source="yandex",
source_url=f"https://realty.yandex.ru/offer/{offer_id}/",
source_id=offer_id,
- address="Екатеринбург, улица Тестовая, 1",
+ address="Екатеринбург, улица Тестовая",
price_rub=5_000_000,
area_m2=50.0,
rooms=2,
@@ -77,6 +88,7 @@ class _FakeRuns:
@pytest.fixture(autouse=True)
def _no_sleep(monkeypatch: pytest.MonkeyPatch) -> None:
"""Kill inter-anchor asyncio.sleep so tests are instant."""
+
async def _instant(_secs: float) -> None:
return None
@@ -85,12 +97,13 @@ def _no_sleep(monkeypatch: pytest.MonkeyPatch) -> None:
@pytest.fixture(autouse=True)
def _stub_scraper_lifecycle(monkeypatch: pytest.MonkeyPatch) -> None:
- """Neutralize __init__/__aenter__/__aexit__ — no get_scraper_delay DB, no httpx."""
+ """Neutralize __init__/__aenter__/__aexit__ — no get_scraper_delay DB, no curl_cffi."""
def _init(self: YandexRealtyScraper, city: str = "ekaterinburg") -> None:
self.city = city
self.request_delay_sec = 0.0
- self._client = None
+ self._cffi_session = None
+ self._cookies = {}
async def _aenter(self: YandexRealtyScraper) -> YandexRealtyScraper:
return self
@@ -111,114 +124,262 @@ def _install_runs(monkeypatch: pytest.MonkeyPatch, fake: _FakeRuns) -> None:
monkeypatch.setattr(scrape_pipeline.scrape_runs, "mark_banned", fake.mark_banned)
+def _make_db_mock(source_ids: list[str]) -> MagicMock:
+ """DB mock для address-enrich фазы: возвращает строки без номера дома."""
+ mock_db = MagicMock()
+ rows = [
+ {
+ "id": i + 1,
+ "source_id": sid,
+ "address": "Екатеринбург, улица Тестовая",
+ "source_url": f"https://realty.yandex.ru/offer/{sid}/",
+ }
+ for i, sid in enumerate(source_ids)
+ ]
+ mock_db.execute.return_value.mappings.return_value.all.return_value = rows
+ return mock_db
+
+
+def _fake_save(_db: Any, lots: list, *, run_id: int | None = None) -> tuple[int, int]:
+ return len(lots), 0
+
+
+# ── YandexCitySweepCounters ─────────────────────────────────────────────────
+
+
+def test_counters_defaults() -> None:
+ from app.services.scrape_pipeline import YandexCitySweepCounters
+
+ c = YandexCitySweepCounters()
+ assert c.anchors_total == 0
+ assert c.lots_fetched == 0
+ assert c.address_attempted == 0
+ assert c.address_enriched == 0
+ assert c.address_failed == 0
+ assert c.errors_count == 0
+
+
+def test_counters_to_dict_all_keys() -> None:
+ from dataclasses import fields
+
+ from app.services.scrape_pipeline import YandexCitySweepCounters
+
+ c = YandexCitySweepCounters()
+ d = c.to_dict()
+ expected = {f.name for f in fields(c)}
+ assert set(d.keys()) == expected
+
+
# ── Happy path: all anchors iterated, counters aggregated ───────────────────
async def test_sweep_iterates_all_anchors_and_aggregates(
monkeypatch: pytest.MonkeyPatch,
) -> None:
- fetch_calls: list[tuple[float, float, int]] = []
+ """fetch_around_multi_room вызывается для каждого anchor, counters агрегируются."""
+ serp_calls: list[str] = []
save_calls: list[int] = []
- async def fake_fetch(
+ async def fake_fetch_multi(
self: YandexRealtyScraper,
lat: float,
lon: float,
radius_m: int = 1000,
- page: int = 0,
+ max_pages: int = 2,
+ **_: Any,
) -> list[ScrapedLot]:
- fetch_calls.append((lat, lon, page))
- # 2 lots per page; let pagination run the full pages_per_anchor.
- return [_fake_lot(f"{lat}-{page}-{i}") for i in range(2)]
+ serp_calls.append(f"{lat:.4f}")
+ return [_fake_lot(f"{lat}-{i}") for i in range(3)]
def fake_save(
_db: Any, lots: list[ScrapedLot], *, run_id: int | None = None
) -> tuple[int, int]:
save_calls.append(len(lots))
- # Pretend everything is a fresh insert.
return len(lots), 0
- monkeypatch.setattr(YandexRealtyScraper, "fetch_around", fake_fetch)
+ monkeypatch.setattr(YandexRealtyScraper, "fetch_around_multi_room", fake_fetch_multi)
monkeypatch.setattr(scrape_pipeline, "save_listings", fake_save)
fake = _FakeRuns()
_install_runs(monkeypatch, fake)
counters = await scrape_pipeline.run_yandex_city_sweep(
- db=object(),
+ db=MagicMock(),
run_id=1,
anchors=TEST_ANCHORS,
pages_per_anchor=2,
request_delay_sec=0.0,
+ enrich_address=False,
)
- # fetch_around called pages_per_anchor times per anchor (no empty-page short-circuit).
- assert len(fetch_calls) == len(TEST_ANCHORS) * 2
- # save_listings called once per anchor (anchor-level batch).
+ assert len(serp_calls) == len(TEST_ANCHORS)
assert len(save_calls) == len(TEST_ANCHORS)
- # Counters: 3 anchors × 2 pages × 2 lots = 12.
assert counters.anchors_total == len(TEST_ANCHORS)
assert counters.anchors_done == len(TEST_ANCHORS)
- assert counters.pages_fetched == len(TEST_ANCHORS) * 2
- assert counters.lots_fetched == 12
- assert counters.lots_inserted == 12
+ assert counters.lots_fetched == len(TEST_ANCHORS) * 3
+ assert counters.lots_inserted == len(TEST_ANCHORS) * 3
assert counters.lots_updated == 0
assert counters.errors_count == 0
- # mark_done called with final counters; no failure/ban.
+ assert counters.address_attempted == 0 # enrich_address=False
assert fake.done is not None
- assert fake.done["lots_fetched"] == 12
+ assert fake.done["lots_fetched"] == len(TEST_ANCHORS) * 3
assert fake.failed is None
assert fake.banned is None
-async def test_save_listings_receives_run_id(monkeypatch: pytest.MonkeyPatch) -> None:
- """run_id must be threaded into save_listings for snapshot provenance."""
- seen_run_ids: list[int | None] = []
-
- async def fake_fetch(self: YandexRealtyScraper, lat, lon, radius_m=1000, page=0):
- return [_fake_lot(f"{lat}-{page}")]
-
- def fake_save(_db, lots, *, run_id=None):
- seen_run_ids.append(run_id)
- return len(lots), 0
-
- monkeypatch.setattr(YandexRealtyScraper, "fetch_around", fake_fetch)
- monkeypatch.setattr(scrape_pipeline, "save_listings", fake_save)
- _install_runs(monkeypatch, _FakeRuns())
-
- await scrape_pipeline.run_yandex_city_sweep(
- db=object(), run_id=777, anchors=TEST_ANCHORS, pages_per_anchor=1,
- )
- assert seen_run_ids == [777, 777, 777]
+# ── enrich_address=False пропускает address-фазу ────────────────────────────
-# ── Empty page short-circuits pagination ────────────────────────────────────
+async def test_enrich_address_false_skips_address_phase(
+ monkeypatch: pytest.MonkeyPatch,
+) -> None:
+ """enrich_address=False → address-enrich фаза не выполняется."""
+ address_http_calls: list[str] = []
+ async def fake_fetch_multi(
+ self: YandexRealtyScraper,
+ lat: float,
+ lon: float,
+ radius_m: int = 1000,
+ max_pages: int = 2,
+ **_: Any,
+ ) -> list[ScrapedLot]:
+ return [_fake_lot(f"{lat}-0")]
-async def test_empty_page_stops_pagination(monkeypatch: pytest.MonkeyPatch) -> None:
- """An empty SERP page ends pagination for that anchor (no further pages)."""
- pages_seen: list[int] = []
+ async def fake_cffi_get(url: str, **_: Any) -> Any:
+ address_http_calls.append(url)
+ return MagicMock(status_code=200, text="<title>test")
- async def fake_fetch(self: YandexRealtyScraper, lat, lon, radius_m=1000, page=0):
- pages_seen.append(page)
- if page == 0:
- return [_fake_lot(f"{lat}-0")]
- return [] # page 1 empty → stop
-
- monkeypatch.setattr(YandexRealtyScraper, "fetch_around", fake_fetch)
- monkeypatch.setattr(
- scrape_pipeline, "save_listings", lambda _db, lots, *, run_id=None: (len(lots), 0)
- )
+ monkeypatch.setattr(YandexRealtyScraper, "fetch_around_multi_room", fake_fetch_multi)
+ monkeypatch.setattr(scrape_pipeline, "save_listings", _fake_save)
fake = _FakeRuns()
_install_runs(monkeypatch, fake)
counters = await scrape_pipeline.run_yandex_city_sweep(
- db=object(), run_id=2, anchors=TEST_ANCHORS, pages_per_anchor=5,
+ db=MagicMock(),
+ run_id=2,
+ anchors=TEST_ANCHORS[:1],
+ enrich_address=False,
+ request_delay_sec=0.0,
)
- # Per anchor: page 0 (1 lot) then page 1 (empty, break) → 2 pages each.
- assert pages_seen == [0, 1, 0, 1, 0, 1]
- assert counters.pages_fetched == 6 # 2 per anchor × 3
- assert counters.lots_fetched == 3 # 1 per anchor
- assert counters.anchors_done == 3
+
+ assert len(address_http_calls) == 0
+ assert counters.address_attempted == 0
+ assert counters.address_enriched == 0
+ assert fake.done is not None
+
+
+# ── address-enrich фаза: счётчики заполняются ───────────────────────────────
+
+
+async def test_address_enrich_fills_counters(
+ monkeypatch: pytest.MonkeyPatch,
+) -> None:
+ """address-enrich вызывается для листингов без номера дома; counters заполняются."""
+ # Fake SERP возвращает 2 листинга с source_id
+ lots = [_fake_lot("id1"), _fake_lot("id2")]
+
+ async def fake_fetch_multi(
+ self: YandexRealtyScraper,
+ lat: float,
+ lon: float,
+ radius_m: int = 1000,
+ max_pages: int = 2,
+ **_: Any,
+ ) -> list[ScrapedLot]:
+ return lots
+
+ monkeypatch.setattr(YandexRealtyScraper, "fetch_around_multi_room", fake_fetch_multi)
+ monkeypatch.setattr(scrape_pipeline, "save_listings", _fake_save)
+
+ # DB mock: возвращает 2 строки для address-enrich запроса
+ mock_db = _make_db_mock(["id1", "id2"])
+
+ # Title с полным адресом — должен быть извлечён через _extract_address_from_title
+ fake_response = MagicMock()
+ fake_response.status_code = 200
+ fake_response.text = (
+ "Продажа квартиры — Екатеринбург, улица Тестовая, 36 — id 123 "
+ "на Яндекс.Недвижимости"
+ )
+
+ # Мок curl_cffi AsyncSession как context manager
+ mock_session = AsyncMock()
+ mock_session.__aenter__ = AsyncMock(return_value=mock_session)
+ mock_session.__aexit__ = AsyncMock(return_value=None)
+ mock_session.get = AsyncMock(return_value=fake_response)
+
+ import curl_cffi.requests as cffi_mod
+
+ monkeypatch.setattr(cffi_mod, "AsyncSession", lambda **_kw: mock_session)
+
+ fake = _FakeRuns()
+ _install_runs(monkeypatch, fake)
+
+ counters = await scrape_pipeline.run_yandex_city_sweep(
+ db=mock_db,
+ run_id=3,
+ anchors=TEST_ANCHORS[:1],
+ enrich_address=True,
+ request_delay_sec=0.0,
+ )
+
+ assert counters.address_attempted == 2
+ assert counters.address_enriched == 2
+ assert counters.address_failed == 0
+ assert fake.done is not None
+
+
+# ── address-enrich: per-item HTTP error не валит sweep ──────────────────────
+
+
+async def test_address_enrich_http_error_does_not_abort(
+ monkeypatch: pytest.MonkeyPatch,
+) -> None:
+ """HTTP 403 при address-enrich → address_failed++, sweep продолжается."""
+
+ async def fake_fetch_multi(
+ self: YandexRealtyScraper,
+ lat: float,
+ lon: float,
+ radius_m: int = 1000,
+ max_pages: int = 2,
+ **_: Any,
+ ) -> list[ScrapedLot]:
+ return [_fake_lot("id-err")]
+
+ monkeypatch.setattr(YandexRealtyScraper, "fetch_around_multi_room", fake_fetch_multi)
+ monkeypatch.setattr(scrape_pipeline, "save_listings", _fake_save)
+
+ mock_db = _make_db_mock(["id-err"])
+
+ fake_response = MagicMock()
+ fake_response.status_code = 403
+ fake_response.text = ""
+
+ mock_session = AsyncMock()
+ mock_session.__aenter__ = AsyncMock(return_value=mock_session)
+ mock_session.__aexit__ = AsyncMock(return_value=None)
+ mock_session.get = AsyncMock(return_value=fake_response)
+
+ import curl_cffi.requests as cffi_mod
+
+ monkeypatch.setattr(cffi_mod, "AsyncSession", lambda **_kw: mock_session)
+
+ fake = _FakeRuns()
+ _install_runs(monkeypatch, fake)
+
+ counters = await scrape_pipeline.run_yandex_city_sweep(
+ db=mock_db,
+ run_id=4,
+ anchors=TEST_ANCHORS[:1],
+ enrich_address=True,
+ request_delay_sec=0.0,
+ )
+
+ assert counters.address_attempted == 1
+ assert counters.address_failed == 1
+ assert counters.address_enriched == 0
+ assert fake.done is not None # sweep завершился несмотря на ошибку
# ── Mid-sweep cancel stops further anchors ──────────────────────────────────
@@ -227,25 +388,35 @@ async def test_empty_page_stops_pagination(monkeypatch: pytest.MonkeyPatch) -> N
async def test_cancel_midway_stops_remaining_anchors(
monkeypatch: pytest.MonkeyPatch,
) -> None:
- fetch_calls: list[str] = []
+ serp_calls: list[str] = []
- async def fake_fetch(self: YandexRealtyScraper, lat, lon, radius_m=1000, page=0):
- fetch_calls.append(f"{lat}")
- return [_fake_lot(f"{lat}-{page}")]
+ async def fake_fetch_multi(
+ self: YandexRealtyScraper,
+ lat: float,
+ lon: float,
+ radius_m: int = 1000,
+ max_pages: int = 2,
+ **_: Any,
+ ) -> list[ScrapedLot]:
+ serp_calls.append(f"{lat}")
+ return [_fake_lot(f"{lat}-0")]
- monkeypatch.setattr(YandexRealtyScraper, "fetch_around", fake_fetch)
- monkeypatch.setattr(
- scrape_pipeline, "save_listings", lambda _db, lots, *, run_id=None: (len(lots), 0)
- )
+ monkeypatch.setattr(YandexRealtyScraper, "fetch_around_multi_room", fake_fetch_multi)
+ monkeypatch.setattr(scrape_pipeline, "save_listings", _fake_save)
# is_cancelled flips True on the 2nd check (before anchor #2) → only anchor #1 runs.
fake = _FakeRuns(cancel_at_anchor=2)
_install_runs(monkeypatch, fake)
counters = await scrape_pipeline.run_yandex_city_sweep(
- db=object(), run_id=3, anchors=TEST_ANCHORS, pages_per_anchor=1,
+ db=MagicMock(),
+ run_id=5,
+ anchors=TEST_ANCHORS,
+ pages_per_anchor=1,
+ enrich_address=False,
+ request_delay_sec=0.0,
)
# Only the first anchor fetched before cancel.
- assert len(fetch_calls) == 1
+ assert len(serp_calls) == 1
assert counters.anchors_done == 1
# Cancel path returns early: no mark_done, no ban/fail.
assert fake.done is None
@@ -253,48 +424,7 @@ async def test_cancel_midway_stops_remaining_anchors(
assert fake.banned is None
-# ── Per-anchor exception is isolated (single failure → continue) ────────────
-
-
-async def test_single_anchor_failure_is_isolated(
- monkeypatch: pytest.MonkeyPatch,
-) -> None:
- """Anchor #2 raises → anchors #1 and #3 still processed; sweep completes."""
- processed: list[str] = []
-
- async def fake_fetch(self: YandexRealtyScraper, lat, lon, radius_m=1000, page=0):
- if name_of(lat) == "A2":
- raise RuntimeError("boom on anchor 2")
- processed.append(name_of(lat))
- return [_fake_lot(f"{lat}-{page}")]
-
- def name_of(lat: float) -> str:
- for a_lat, _lon, nm in TEST_ANCHORS:
- if a_lat == lat:
- return nm
- return "?"
-
- monkeypatch.setattr(YandexRealtyScraper, "fetch_around", fake_fetch)
- monkeypatch.setattr(
- scrape_pipeline, "save_listings", lambda _db, lots, *, run_id=None: (len(lots), 0)
- )
- fake = _FakeRuns()
- _install_runs(monkeypatch, fake)
-
- counters = await scrape_pipeline.run_yandex_city_sweep(
- db=object(), run_id=4, anchors=TEST_ANCHORS, pages_per_anchor=1,
- )
- # A1 and A3 processed; A2 raised and was skipped.
- assert processed == ["A1", "A3"]
- assert counters.errors_count == 1
- assert counters.anchors_done == 3 # loop advanced past all 3 anchors
- assert counters.lots_fetched == 2 # only A1 + A3 contributed lots
- # Not an abort — sweep marked done despite the isolated failure.
- assert fake.done is not None
- assert fake.banned is None
-
-
-# ── Defensive abort: N consecutive failures → mark_banned, stop early ───────
+# ── Consecutive failures → mark_banned + abort ──────────────────────────────
async def test_consecutive_failures_abort_sweep(
@@ -303,14 +433,19 @@ async def test_consecutive_failures_abort_sweep(
"""All anchors raise → after MAX_CONSECUTIVE_FAILURES the sweep aborts + bans."""
attempts: list[str] = []
- async def fake_fetch(self: YandexRealtyScraper, lat, lon, radius_m=1000, page=0):
+ async def fake_fetch_fail(
+ self: YandexRealtyScraper,
+ lat: float,
+ lon: float,
+ radius_m: int = 1000,
+ max_pages: int = 2,
+ **_: Any,
+ ) -> list[ScrapedLot]:
attempts.append(f"{lat}")
raise RuntimeError("yandex blocked")
- monkeypatch.setattr(YandexRealtyScraper, "fetch_around", fake_fetch)
- monkeypatch.setattr(
- scrape_pipeline, "save_listings", lambda _db, lots, *, run_id=None: (len(lots), 0)
- )
+ monkeypatch.setattr(YandexRealtyScraper, "fetch_around_multi_room", fake_fetch_fail)
+ monkeypatch.setattr(scrape_pipeline, "save_listings", _fake_save)
fake = _FakeRuns()
_install_runs(monkeypatch, fake)
@@ -321,9 +456,13 @@ async def test_consecutive_failures_abort_sweep(
(56.8650, 60.6200, "A5"),
]
counters = await scrape_pipeline.run_yandex_city_sweep(
- db=object(), run_id=5, anchors=anchors, pages_per_anchor=1,
+ db=MagicMock(),
+ run_id=6,
+ anchors=anchors,
+ enrich_address=False,
+ request_delay_sec=0.0,
)
- # Abort exactly at the threshold (3 consecutive failures), not all 5 anchors.
+ # Abort exactly at the threshold (3 consecutive failures).
assert len(attempts) == scrape_pipeline.YANDEX_SWEEP_MAX_CONSECUTIVE_FAILURES
assert counters.errors_count == scrape_pipeline.YANDEX_SWEEP_MAX_CONSECUTIVE_FAILURES
assert counters.anchors_done == scrape_pipeline.YANDEX_SWEEP_MAX_CONSECUTIVE_FAILURES
@@ -332,11 +471,15 @@ async def test_consecutive_failures_abort_sweep(
assert fake.done is None # aborted, never marked done
+# ── Non-consecutive failures do not abort ───────────────────────────────────
+
+
async def test_non_consecutive_failures_do_not_abort(
monkeypatch: pytest.MonkeyPatch,
) -> None:
"""A success between two failures resets the consecutive counter → no abort."""
- def name_of(lat: float, anchors: list[tuple[float, float, str]]) -> str:
+
+ def _name_of(lat: float, anchors: list[tuple[float, float, str]]) -> str:
for a_lat, _lon, nm in anchors:
if a_lat == lat:
return nm
@@ -347,26 +490,178 @@ async def test_non_consecutive_failures_do_not_abort(
(56.7700, 60.5500, "A4"),
(56.8650, 60.6200, "A5"),
]
- # Fail A1, succeed A2, fail A3, succeed A4, fail A5 → never 3-in-a-row.
failing = {"A1", "A3", "A5"}
- async def fake_fetch(self: YandexRealtyScraper, lat, lon, radius_m=1000, page=0):
- if name_of(lat, anchors) in failing:
+ async def fake_fetch(
+ self: YandexRealtyScraper,
+ lat: float,
+ lon: float,
+ radius_m: int = 1000,
+ max_pages: int = 2,
+ **_: Any,
+ ) -> list[ScrapedLot]:
+ if _name_of(lat, anchors) in failing:
raise RuntimeError("transient")
- return [_fake_lot(f"{lat}-{page}")]
+ return [_fake_lot(f"{lat}-0")]
- monkeypatch.setattr(YandexRealtyScraper, "fetch_around", fake_fetch)
- monkeypatch.setattr(
- scrape_pipeline, "save_listings", lambda _db, lots, *, run_id=None: (len(lots), 0)
- )
+ monkeypatch.setattr(YandexRealtyScraper, "fetch_around_multi_room", fake_fetch)
+ monkeypatch.setattr(scrape_pipeline, "save_listings", _fake_save)
fake = _FakeRuns()
_install_runs(monkeypatch, fake)
counters = await scrape_pipeline.run_yandex_city_sweep(
- db=object(), run_id=6, anchors=anchors, pages_per_anchor=1,
+ db=MagicMock(),
+ run_id=7,
+ anchors=anchors,
+ enrich_address=False,
+ request_delay_sec=0.0,
)
assert counters.errors_count == 3 # A1, A3, A5
assert counters.anchors_done == 5 # walked the whole list, no early abort
assert counters.lots_fetched == 2 # A2 + A4 contributed
assert fake.banned is None
assert fake.done is not None
+
+
+# ── mark_done вызывается даже при пустом SERP ───────────────────────────────
+
+
+async def test_mark_done_always_called(
+ monkeypatch: pytest.MonkeyPatch,
+) -> None:
+ async def fake_fetch_empty(
+ self: YandexRealtyScraper,
+ lat: float,
+ lon: float,
+ radius_m: int = 1000,
+ max_pages: int = 2,
+ **_: Any,
+ ) -> list[ScrapedLot]:
+ return []
+
+ monkeypatch.setattr(YandexRealtyScraper, "fetch_around_multi_room", fake_fetch_empty)
+ monkeypatch.setattr(scrape_pipeline, "save_listings", _fake_save)
+ fake = _FakeRuns()
+ _install_runs(monkeypatch, fake)
+
+ counters = await scrape_pipeline.run_yandex_city_sweep(
+ db=MagicMock(),
+ run_id=8,
+ anchors=TEST_ANCHORS,
+ enrich_address=False,
+ request_delay_sec=0.0,
+ )
+
+ assert fake.done is not None
+ assert counters.lots_fetched == 0
+ assert counters.anchors_done == len(TEST_ANCHORS)
+
+
+# ── Admin endpoint tests (offline) ──────────────────────────────────────────
+
+
+@pytest.fixture
+def app_with_admin():
+ from fastapi import FastAPI
+ from fastapi.testclient import TestClient
+
+ from app.api.v1 import admin as admin_module
+ from app.core.db import get_db
+
+ app = FastAPI()
+ app.include_router(admin_module.router, prefix="/api/v1/admin")
+
+ def fake_db():
+ yield MagicMock()
+
+ app.dependency_overrides[get_db] = fake_db
+ return TestClient(app)
+
+
+def test_yandex_start_endpoint_validates_pages_per_anchor(app_with_admin) -> None:
+ """pages_per_anchor=99 превышает max=10 → 422."""
+ r = app_with_admin.post(
+ "/api/v1/admin/scrape/yandex-city-sweep",
+ json={"pages_per_anchor": 99},
+ )
+ assert r.status_code == 422
+
+
+def test_yandex_start_endpoint_validates_delay_too_low(app_with_admin) -> None:
+ """request_delay_sec=1.0 меньше min=3.0 → 422."""
+ r = app_with_admin.post(
+ "/api/v1/admin/scrape/yandex-city-sweep",
+ json={"request_delay_sec": 1.0},
+ )
+ assert r.status_code == 422
+
+
+def test_yandex_start_endpoint_ok(app_with_admin) -> None:
+ """Valid request → 200, run_id в ответе."""
+ from unittest.mock import patch
+
+ with (
+ patch("app.services.scrape_runs.create_run", return_value=55),
+ patch("app.services.scrape_pipeline.run_yandex_city_sweep", new_callable=AsyncMock),
+ ):
+ r = app_with_admin.post(
+ "/api/v1/admin/scrape/yandex-city-sweep",
+ json={"pages_per_anchor": 2, "enrich_address": True},
+ )
+ assert r.status_code == 200
+ body = r.json()
+ assert body["run_id"] == 55
+ assert body["status"] == "running"
+ assert body["pages_per_anchor"] == 2
+ assert body["detail_top_n"] == 0
+
+
+def test_yandex_cancel_endpoint(app_with_admin) -> None:
+ """Cancel endpoint возвращает run_id + cancelled=True."""
+ from unittest.mock import patch
+
+ with patch("app.services.scrape_runs.mark_cancelled", return_value=True):
+ r = app_with_admin.post("/api/v1/admin/scrape/yandex-city-sweep/99/cancel")
+ assert r.status_code == 200
+ body = r.json()
+ assert body["run_id"] == 99
+ assert body["cancelled"] is True
+ assert body["ok"] is True
+
+
+def test_yandex_cancel_endpoint_not_running(app_with_admin) -> None:
+ """Cancel на не-running sweep → cancelled=False."""
+ from unittest.mock import patch
+
+ with patch("app.services.scrape_runs.mark_cancelled", return_value=False):
+ r = app_with_admin.post("/api/v1/admin/scrape/yandex-city-sweep/100/cancel")
+ assert r.status_code == 200
+ assert r.json()["cancelled"] is False
+
+
+def test_yandex_list_runs_endpoint(app_with_admin) -> None:
+ """List runs endpoint возвращает список."""
+ from datetime import UTC, datetime
+ from unittest.mock import patch
+
+ fake_rows = [
+ {
+ "run_id": 20,
+ "source": "yandex_city_sweep",
+ "status": "done",
+ "params": {"pages_per_anchor": 2, "enrich_address": True},
+ "counters": {"lots_fetched": 200, "address_enriched": 80},
+ "error": None,
+ "started_at": datetime(2025, 5, 1, tzinfo=UTC),
+ "finished_at": datetime(2025, 5, 1, 1, tzinfo=UTC),
+ "heartbeat_at": datetime(2025, 5, 1, 1, tzinfo=UTC),
+ }
+ ]
+ with patch("app.services.scrape_runs.list_recent", return_value=fake_rows):
+ r = app_with_admin.get("/api/v1/admin/scrape/yandex-city-sweep/runs")
+ assert r.status_code == 200
+ body = r.json()
+ assert len(body) == 1
+ assert body[0]["run_id"] == 20
+ assert body[0]["source"] == "yandex_city_sweep"
+ assert "2025-05-01" in body[0]["started_at"]