feat(scraper-kit): copy remaining sweep orchestrators (yandex/cian/domclick/full-load), strangler (#2135 F2) #2174

Merged
lekss361 merged 1 commit from feat/scraper-kit-pipeline-copy2 into main 2026-07-02 17:29:06 +00:00
4 changed files with 2508 additions and 7 deletions

View file

@ -28,9 +28,16 @@ from app.services.house_imv_backfill import (
) )
from app.services.matching import match_or_create_house as _match_or_create_house from app.services.matching import match_or_create_house as _match_or_create_house
from app.services.matching import upsert_listing_source as _upsert_listing_source from app.services.matching import upsert_listing_source as _upsert_listing_source
from app.services.yandex_address_backfill import (
_RE_HAS_HOUSE_NUMBER,
_extract_address_from_title,
)
from app.services.yandex_address_backfill import ( from app.services.yandex_address_backfill import (
backfill_yandex_addresses as _backfill_yandex_addresses, backfill_yandex_addresses as _backfill_yandex_addresses,
) )
from app.services.yandex_price_history import (
record_yandex_price_history as _record_yandex_price_history,
)
if TYPE_CHECKING: if TYPE_CHECKING:
from sqlalchemy.orm import Session from sqlalchemy.orm import Session
@ -261,6 +268,15 @@ class RealEnrichmentJobs:
heartbeat=heartbeat, heartbeat=heartbeat,
) )
def record_yandex_price_history(self, db: Session, lots: list[Any]) -> int:
return _record_yandex_price_history(db, lots)
def extract_address_from_title(self, html: str) -> str | None:
return _extract_address_from_title(html)
def address_has_house_number(self, address: str) -> bool:
return bool(_RE_HAS_HOUSE_NUMBER.search(address))
__all__ = [ __all__ = [
"RealEnrichmentJobs", "RealEnrichmentJobs",

View file

@ -0,0 +1,481 @@
"""Golden-parity (F2): kit sweep-оркестраторы ≡ `app.services.scrape_pipeline`.
Продолжение test_scraper_kit_pipeline_parity.py (F1 покрыл avito city sweep). Здесь
остальные 7 sweep'ов, скопированных в `scraper_kit.orchestration.pipeline` (#2135 F2):
City sweep'ы (приоритет — активны в проде):
- run_yandex_city_sweep combos SERP + save + price-history
- run_cian_city_sweep SERP + newbuilding_only-фильтр + save
- run_domclick_city_sweep BFF citywide + честный статус (done / failed)
Плюс:
- run_avito_newbuilding_sweep citywide novostroyka SERP + save
Full load'ы (smoke-parity — импорт + базовый прогон через on_bucket):
- run_avito_full_load / run_cian_full_load / run_yandex_full_load
Метод как в F1: гоняем ОБА orchestrator'а на идентичном мокнутом I/O, сравниваем
итоговый counters.to_dict() + последовательность (method, numeric-counters) вызовов
scrape_runs. Развязка `app.*` (config/matcher/enrichment/runs/shutdown инжектируются)
не должна менять ветвление/счётчики. Без сети, без БД.
"""
from __future__ import annotations
import os
from types import SimpleNamespace
from typing import Any
from unittest.mock import AsyncMock, MagicMock, patch
import pytest
os.environ.setdefault("DATABASE_URL", "postgresql+psycopg://test:test@localhost/test_db")
from scraper_kit.orchestration.pipeline import (
run_avito_full_load as new_avito_full_load,
)
from scraper_kit.orchestration.pipeline import (
run_avito_newbuilding_sweep as new_nb_sweep,
)
from scraper_kit.orchestration.pipeline import (
run_cian_city_sweep as new_cian_city_sweep,
)
from scraper_kit.orchestration.pipeline import (
run_cian_full_load as new_cian_full_load,
)
from scraper_kit.orchestration.pipeline import (
run_domclick_city_sweep as new_domclick_city_sweep,
)
from scraper_kit.orchestration.pipeline import (
run_yandex_city_sweep as new_yandex_city_sweep,
)
from scraper_kit.orchestration.pipeline import (
run_yandex_full_load as new_yandex_full_load,
)
from app.services.scrape_pipeline import (
run_avito_full_load as old_avito_full_load,
)
from app.services.scrape_pipeline import (
run_avito_newbuilding_sweep as old_nb_sweep,
)
from app.services.scrape_pipeline import (
run_cian_city_sweep as old_cian_city_sweep,
)
from app.services.scrape_pipeline import (
run_cian_full_load as old_cian_full_load,
)
from app.services.scrape_pipeline import (
run_domclick_city_sweep as old_domclick_city_sweep,
)
from app.services.scrape_pipeline import (
run_yandex_city_sweep as old_yandex_city_sweep,
)
from app.services.scrape_pipeline import (
run_yandex_full_load as old_yandex_full_load,
)
OLD_PFX = "app.services.scrape_pipeline"
NEW_PFX = "scraper_kit.orchestration.pipeline"
# ── recording scrape_runs ────────────────────────────────────────────────────
_NON_NUMERIC_KEYS = {"enrichment_abort_note", "done_buckets"}
class _RunsRecorder:
"""Записывает вызовы scrape_runs-финализаторов. is_cancelled всегда False."""
def __init__(self) -> None:
self.calls: list[tuple[str, dict[str, Any]]] = []
def is_cancelled(self, db: Any, run_id: int) -> bool:
return False
def update_heartbeat(self, db: Any, run_id: int, counters: dict[str, Any]) -> None:
self.calls.append(("update_heartbeat", dict(counters)))
def mark_done(self, db: Any, run_id: int, counters: dict[str, Any]) -> None:
self.calls.append(("mark_done", dict(counters)))
def mark_banned(self, db: Any, run_id: int, error: str, counters: dict[str, Any]) -> None:
self.calls.append(("mark_banned", dict(counters)))
def mark_failed(self, db: Any, run_id: int, error: str, counters: dict[str, Any]) -> None:
self.calls.append(("mark_failed", dict(counters)))
def _normalize(calls: list[tuple[str, dict[str, Any]]]) -> list[tuple[str, dict[str, int]]]:
"""Оставить имя метода + числовые counters (убрать note/done_buckets)."""
out: list[tuple[str, dict[str, int]]] = []
for method, counters in calls:
numeric = {k: v for k, v in counters.items() if k not in _NON_NUMERIC_KEYS}
out.append((method, numeric))
return out
def _config() -> SimpleNamespace:
"""Инжектируемый config (kit) + тот же объект как patched settings (old)."""
return SimpleNamespace(
scraper_fetch_mode="curl_cffi",
browser_http_endpoint="http://browser.test/fetch",
scraper_proxy_url=None,
avito_proxy_rotate_url=None,
avito_proxy_max_rotations=0,
avito_serp_ok_not_banned=True,
avito_proxy_rotate_settle_s=0.0,
proxy_rotate_attempts=1,
proxy_rotate_attempt_timeout_s=1.0,
cian_proxy_rotate_url=None,
cian_proxy_max_rotations=0,
yandex_proxy_rotate_url=None,
yandex_proxy_max_rotations=0,
scraper_skip_seen_today=False,
cian_full_load_per_fetch_timeout_s=0.0,
)
def _ctx_scraper(**attrs: Any) -> MagicMock:
"""MagicMock, пригодный и как `Cls()`, и как `async with Cls() as x`.
__aenter__ возвращает сам объект, __aexit__ no-op. Доп. атрибуты/методы
задаются через kwargs (async-методы AsyncMock).
"""
m = MagicMock()
m.__aenter__ = AsyncMock(return_value=m)
m.__aexit__ = AsyncMock(return_value=None)
for k, v in attrs.items():
setattr(m, k, v)
return m
def _async_session_cm() -> MagicMock:
sess = MagicMock()
sess.__aenter__ = AsyncMock(return_value=sess)
sess.__aexit__ = AsyncMock(return_value=None)
sess.close = AsyncMock()
return sess
_DriveResult = tuple[dict[str, int], list[tuple[str, dict[str, int]]]]
async def _assert_parity(old_res: _DriveResult, new_res: _DriveResult) -> dict[str, int]:
old_counters, old_calls = old_res
new_counters, new_calls = new_res
assert (
new_counters == old_counters
), f"counters diverged:\n old={old_counters}\n new={new_counters}"
assert (
new_calls == old_calls
), f"scrape_runs calls diverged:\n old={old_calls}\n new={new_calls}"
return new_counters
# ── Yandex city sweep ─────────────────────────────────────────────────────────
def _yandex_scraper(combos: list[tuple[str, list[Any]]]) -> MagicMock:
"""Fake YandexRealtyScraper: fetch_around_multi_room вызывает on_combo по combos."""
async def _fetch(*_a: Any, on_combo: Any = None, **_k: Any) -> None:
for label, lots in combos:
on_combo(label, lots)
return _ctx_scraper(fetch_around_multi_room=_fetch)
async def _drive_yandex_city(*, new: bool) -> _DriveResult:
recorder = _RunsRecorder()
db = MagicMock()
combos = [
("2к-price0", [MagicMock(source_id="y1"), MagicMock(source_id="y2")]),
("1к-price0", [MagicMock(source_id="y3")]),
]
scraper = _yandex_scraper(combos)
save_mock = MagicMock(side_effect=[(2, 0), (1, 0)])
cfg = _config()
kwargs = dict(
run_id=1,
anchors=None,
pages_per_anchor=1,
request_delay_sec=0.0,
enrich_address=False,
)
if new:
enrichment = MagicMock()
enrichment.record_yandex_price_history = MagicMock(return_value=5)
with (
patch(f"{NEW_PFX}.YandexRealtyScraper", return_value=scraper),
patch(f"{NEW_PFX}.save_listings", save_mock),
patch(f"{NEW_PFX}.runs", recorder),
):
counters = await new_yandex_city_sweep(
db, config=cfg, matcher=MagicMock(), enrichment=enrichment, **kwargs
)
else:
with (
patch(f"{OLD_PFX}.YandexRealtyScraper", return_value=scraper),
patch(f"{OLD_PFX}.save_listings", save_mock),
patch(f"{OLD_PFX}.record_yandex_price_history", MagicMock(return_value=5)),
patch(f"{OLD_PFX}.scrape_runs", recorder),
patch(f"{OLD_PFX}.settings", cfg),
patch(f"{OLD_PFX}.shutdown_requested", return_value=False),
):
counters = await old_yandex_city_sweep(db, **kwargs)
return counters.to_dict(), _normalize(recorder.calls)
@pytest.mark.asyncio
async def test_yandex_city_sweep_parity() -> None:
"""1 центральный anchor, 2 combos, save + price-history, enrich_address=False."""
counters = await _assert_parity(
await _drive_yandex_city(new=False), await _drive_yandex_city(new=True)
)
assert counters["lots_fetched"] == 3
assert counters["lots_inserted"] == 3
assert counters["price_history_rows"] == 5
assert counters["anchors_done"] == 1
# ── Cian city sweep ───────────────────────────────────────────────────────────
def _cian_lot(segment: str) -> MagicMock:
return MagicMock(listing_segment=segment, house_source=None, house_ext_id=None)
async def _drive_cian_city(*, new: bool) -> _DriveResult:
recorder = _RunsRecorder()
db = MagicMock()
# 3 novostroyki + 2 secondary → newbuilding_only оставит 3.
lots = [_cian_lot("novostroyki")] * 3 + [_cian_lot("vtorichnaya")] * 2
scraper = _ctx_scraper(fetch_around_multi_room=AsyncMock(return_value=lots))
save_mock = MagicMock(side_effect=[(3, 0)])
cfg = _config()
kwargs = dict(
run_id=1,
anchors=[(56.84, 60.60, "A1")],
radius_m=1000,
pages_per_anchor=1,
request_delay_sec=0.0,
detail_top_n=0,
enrich_houses=False,
newbuilding_only=True,
)
if new:
with (
patch(f"{NEW_PFX}.CianScraper", return_value=scraper),
patch(f"{NEW_PFX}.save_listings", save_mock),
patch(f"{NEW_PFX}.runs", recorder),
):
counters = await new_cian_city_sweep(db, config=cfg, matcher=MagicMock(), **kwargs)
else:
with (
patch("app.services.scrapers.cian.CianScraper", return_value=scraper),
patch(f"{OLD_PFX}.save_listings", save_mock),
patch(f"{OLD_PFX}.scrape_runs", recorder),
patch(f"{OLD_PFX}.settings", cfg),
patch(f"{OLD_PFX}.shutdown_requested", return_value=False),
):
counters = await old_cian_city_sweep(db, **kwargs)
return counters.to_dict(), _normalize(recorder.calls)
@pytest.mark.asyncio
async def test_cian_city_sweep_parity() -> None:
"""1 anchor, newbuilding_only отбрасывает вторичку, save новостроек, mark_done."""
counters = await _assert_parity(
await _drive_cian_city(new=False), await _drive_cian_city(new=True)
)
assert counters["lots_fetched"] == 5
assert counters["lots_dropped_secondary"] == 2
assert counters["lots_inserted"] == 3
assert counters["anchors_done"] == 1
# ── DomClick city sweep ───────────────────────────────────────────────────────
async def _drive_domclick(*, new: bool, lots_n: int, blocked: bool) -> _DriveResult:
recorder = _RunsRecorder()
db = MagicMock()
lots = [MagicMock() for _ in range(lots_n)]
scraper = _ctx_scraper(
fetch_city=AsyncMock(return_value=lots),
blocked=blocked,
geo_filtered=0,
fetch_errors=0,
)
save_mock = MagicMock(side_effect=[(lots_n, 0)] if lots_n else [])
cfg = _config()
kwargs = dict(run_id=1, city_id=4, pages=1, request_delay_sec=0.0)
if new:
with (
patch(f"{NEW_PFX}.DomClickScraper", return_value=scraper),
patch(f"{NEW_PFX}.save_listings", save_mock),
patch(f"{NEW_PFX}.runs", recorder),
):
counters = await new_domclick_city_sweep(db, config=cfg, matcher=MagicMock(), **kwargs)
else:
with (
patch("app.services.scrapers.domclick.DomClickScraper", return_value=scraper),
patch(f"{OLD_PFX}.save_listings", save_mock),
patch(f"{OLD_PFX}.scrape_runs", recorder),
patch(f"{OLD_PFX}.settings", cfg),
patch(f"{OLD_PFX}.shutdown_requested", return_value=False),
):
counters = await old_domclick_city_sweep(db, **kwargs)
return counters.to_dict(), _normalize(recorder.calls)
@pytest.mark.asyncio
async def test_domclick_city_sweep_parity_done() -> None:
"""Citywide fetch_city → 4 lots saved → mark_done."""
counters = await _assert_parity(
await _drive_domclick(new=False, lots_n=4, blocked=False),
await _drive_domclick(new=True, lots_n=4, blocked=False),
)
assert counters["lots_fetched"] == 4
assert counters["lots_inserted"] == 4
@pytest.mark.asyncio
async def test_domclick_city_sweep_parity_blocked_failed() -> None:
"""QRATOR-блок + 0 lots → mark_failed (честный статус #1968)."""
counters = await _assert_parity(
await _drive_domclick(new=False, lots_n=0, blocked=True),
await _drive_domclick(new=True, lots_n=0, blocked=True),
)
assert counters["lots_fetched"] == 0
assert counters["blocked"] == 1
# ── Avito newbuilding sweep ───────────────────────────────────────────────────
async def _drive_nb_sweep(*, new: bool) -> _DriveResult:
recorder = _RunsRecorder()
db = MagicMock()
lots = [MagicMock() for _ in range(6)]
scraper = MagicMock()
scraper._cffi = None
scraper._browser = None
scraper.fetch_newbuildings = AsyncMock(return_value=lots)
save_mock = MagicMock(side_effect=[(5, 1)])
cfg = _config()
kwargs = dict(run_id=1, pages=2, request_delay_sec=0.0)
if new:
with (
patch(f"{NEW_PFX}.AvitoScraper", return_value=scraper),
patch(f"{NEW_PFX}.save_listings", save_mock),
patch(f"{NEW_PFX}.runs", recorder),
patch(f"{NEW_PFX}.AsyncSession", return_value=_async_session_cm()),
):
counters = await new_nb_sweep(db, config=cfg, matcher=MagicMock(), **kwargs)
else:
with (
patch(f"{OLD_PFX}.AvitoScraper", return_value=scraper),
patch(f"{OLD_PFX}.save_listings", save_mock),
patch(f"{OLD_PFX}.scrape_runs", recorder),
patch(f"{OLD_PFX}.settings", cfg),
patch(f"{OLD_PFX}.shutdown_requested", return_value=False),
patch(f"{OLD_PFX}.AsyncSession", return_value=_async_session_cm()),
):
counters = await old_nb_sweep(db, **kwargs)
return counters.to_dict(), _normalize(recorder.calls)
@pytest.mark.asyncio
async def test_avito_newbuilding_sweep_parity() -> None:
"""Citywide novostroyka SERP → save → heartbeat → mark_done."""
counters = await _assert_parity(
await _drive_nb_sweep(new=False), await _drive_nb_sweep(new=True)
)
assert counters["lots_fetched"] == 6
assert counters["lots_inserted"] == 5
assert counters["lots_updated"] == 1
# ── Full loads (smoke-parity через on_bucket) ─────────────────────────────────
def _full_load_scraper(buckets: list[tuple[str, list[Any]]]) -> MagicMock:
"""Fake scraper: fetch_all_secondary вызывает on_bucket по каждому бакету."""
async def _fetch(*_a: Any, on_bucket: Any = None, on_progress: Any = None, **_k: Any) -> None:
for key, lots in buckets:
on_bucket(key, lots)
scraper = _ctx_scraper(fetch_all_secondary=_fetch)
scraper._browser = None
return scraper
async def _drive_full_load(*, new: bool, source: str) -> _DriveResult:
recorder = _RunsRecorder()
db = MagicMock()
buckets = [
("2к:0-5m", [MagicMock(source_id=f"{source}1"), MagicMock(source_id=f"{source}2")]),
("1к:0-5m", [MagicMock(source_id=f"{source}3")]),
]
scraper = _full_load_scraper(buckets)
save_mock = MagicMock(side_effect=[(2, 0), (1, 0)])
cfg = _config()
old_map = {
"avito": (old_avito_full_load, new_avito_full_load, f"{OLD_PFX}.AvitoScraper"),
"cian": (old_cian_full_load, new_cian_full_load, "app.services.scrapers.cian.CianScraper"),
"yandex": (
old_yandex_full_load,
new_yandex_full_load,
"app.services.scrapers.yandex_realty.YandexRealtyScraper",
),
}
old_fn, new_fn, old_scraper_target = old_map[source]
if new:
new_scraper_target = {
"avito": f"{NEW_PFX}.AvitoScraper",
"cian": f"{NEW_PFX}.CianScraper",
"yandex": f"{NEW_PFX}.YandexRealtyScraper",
}[source]
extra: dict[str, Any] = {}
if source == "yandex":
enrichment = MagicMock()
enrichment.record_yandex_price_history = MagicMock(return_value=0)
extra["enrichment"] = enrichment
with (
patch(new_scraper_target, return_value=scraper),
patch(f"{NEW_PFX}.save_listings", save_mock),
patch(f"{NEW_PFX}.runs", recorder),
):
counters = await new_fn(db, run_id=1, config=cfg, matcher=MagicMock(), **extra)
else:
import contextlib
ctx = [
patch(old_scraper_target, return_value=scraper),
patch(f"{OLD_PFX}.save_listings", save_mock),
patch(f"{OLD_PFX}.scrape_runs", recorder),
patch(f"{OLD_PFX}.settings", cfg),
patch(f"{OLD_PFX}.shutdown_requested", return_value=False),
]
if source == "yandex":
ctx.append(patch(f"{OLD_PFX}.record_yandex_price_history", MagicMock(return_value=0)))
with contextlib.ExitStack() as stack:
for cm in ctx:
stack.enter_context(cm)
counters = await old_fn(db, run_id=1)
return counters.to_dict(), _normalize(recorder.calls)
@pytest.mark.asyncio
@pytest.mark.parametrize("source", ["avito", "cian", "yandex"])
async def test_full_load_smoke_parity(source: str) -> None:
"""Full load: 2 бакета через on_bucket → save + heartbeat → mark_done. Counters ≡."""
counters = await _assert_parity(
await _drive_full_load(new=False, source=source),
await _drive_full_load(new=True, source=source),
)
assert counters["unique_fetched"] == 3
assert counters["saved_inserted"] == 3
assert counters["saved_updated"] == 0

View file

@ -208,6 +208,36 @@ class EnrichmentJobs(Protocol):
""" """
... ...
# ── Yandex sweep-хелперы (#2135 F2) ──────────────────────────────────────
# Продуктовая логика, дёргаемая yandex city/full sweep-оркестраторами напрямую
# (не через отдельные backfill-джобы). Инжектируются вместо прямых импортов
# app.services.yandex_price_history / yandex_address_backfill.
def record_yandex_price_history(self, db: Session, lots: list[Any]) -> int:
"""Записать price-history из gate price.previous/trend. Проксирует
record_yandex_price_history. Returns число вставленных строк истории.
`lots` список kit-`ScrapedLot` (тип оставлен `Any`, чтобы не тянуть
доменный тип в контракт). Читается duck-typing'ом (source_id/price/…).
"""
...
def extract_address_from_title(self, html: str) -> str | None:
"""Извлечь полный адрес из <title> detail-страницы Yandex.
Проксирует `_extract_address_from_title` (yandex_address_backfill).
Returns адрес с номером дома или None если не распарсилось.
"""
...
def address_has_house_number(self, address: str) -> bool:
"""True если адрес содержит номер дома (`,\\s*\\d+`).
Проксирует regex `_RE_HAS_HOUSE_NUMBER` из yandex_address_backfill
gate для решения, обновлять ли address на обогащённый.
"""
...
__all__ = [ __all__ = [
"EnrichmentJobs", "EnrichmentJobs",