fix parsing db and tests

This commit is contained in:
qananasikq
2026-04-08 14:23:50 +03:00
parent fd1d1d3c9f
commit 3edd706110
7 changed files with 112 additions and 13 deletions

View File

@@ -22,10 +22,10 @@ class NetworkCapture:
_seen_resp: set[str] = field(default_factory=set) _seen_resp: set[str] = field(default_factory=set)
_origin: str | None = None _origin: str | None = None
def attach(self, page: Page) -> None: def attach(self, page: Page, origin_url: str | None = None) -> None:
# Подписываемся на request/response события страницы. # Подписываемся на request/response события страницы.
try: try:
self._origin = urlparse(page.url).netloc.lower() or None self._origin = urlparse(origin_url or page.url).netloc.lower() or None
except Exception: except Exception:
self._origin = None self._origin = None
page.on("request", self._on_request) page.on("request", self._on_request)
@@ -47,7 +47,7 @@ class NetworkCapture:
if len(self.requests) >= self.settings.gentle.max_requests: if len(self.requests) >= self.settings.gentle.max_requests:
return return
key = f"{request.method}:{request.url}:{request.post_data or ''}" key = f"{request.method}:{request.url}:{request.post_data or ''}"
# Дедупликация одинаковых запросов. # Дедупликация одинаковых запросов.
if key in self._seen_req: if key in self._seen_req:
return return
self._seen_req.add(key) self._seen_req.add(key)
@@ -95,7 +95,7 @@ class NetworkCapture:
@staticmethod @staticmethod
def _categorize(url: str) -> str: def _categorize(url: str) -> str:
# Простая эвристика для разбивки ответов по смыслу. # Простая эвристика для разбивки ответов по смыслу.
low = url.lower() low = url.lower()
mapping = { mapping = {
"images": ["image", "media", "photos", "gallery"], "images": ["image", "media", "photos", "gallery"],

View File

@@ -130,7 +130,10 @@ class CarMapper:
"brand": brand, "model": model, "year": year, "price": price, "mileage": mileage, "brand": brand, "model": model, "year": year, "price": price, "mileage": mileage,
"color": color, "drive": drive, "gearbox": gearbox, "body_type": body_type, "color": color, "drive": drive, "gearbox": gearbox, "body_type": body_type,
"engine_volume": engine_volume, "is_damaged": is_damaged, "is_sold": is_sold, "engine_volume": engine_volume, "is_damaged": is_damaged, "is_sold": is_sold,
"image_count": len(images_records), "country": country, "selling_type": "AUCTION", "one_owner": one_owner,
"new_car": new_car, "evaluation": evaluation, "non_smoking": non_smoking,
"rental": rental, "repair_history": repair_history,
"images": [image.fullres_image for image in images_records],
}, sort_keys=True, default=str).encode()).hexdigest() }, sort_keys=True, default=str).encode()).hexdigest()
return CarRecord( return CarRecord(

View File

@@ -1,3 +1,4 @@
import json
import logging import logging
from contextlib import contextmanager from contextlib import contextmanager
from datetime import datetime, timezone from datetime import datetime, timezone
@@ -13,6 +14,41 @@ from .schemas import CarRecord
logger = logging.getLogger("iaai_scraper.db") logger = logging.getLogger("iaai_scraper.db")
CAR_DB_FIELDS = {
"parser_id",
"brand",
"model",
"year",
"price",
"currency",
"mileage",
"country",
"is_sold",
"color",
"drive",
"gearbox",
"steering_wheel",
"body_type",
"engine_volume",
"selling_type",
"one_owner",
"new_car",
"is_hidden",
"origin",
"origin_url",
"origin_id",
"is_damaged",
"evaluation",
"non_smoking",
"rental",
"repair_history",
"slug",
"last_seen_at",
"content_hash",
"raw_attributes",
}
class PersistenceService: class PersistenceService:
def __init__(self, settings: Settings) -> None: def __init__(self, settings: Settings) -> None:
@@ -64,15 +100,21 @@ class PersistenceService:
for image_payload in images: for image_payload in images:
session.add(Image(fullres_image=str(image_payload["fullres_image"]), preview_image=str(image_payload["preview_image"]), order_index=int(image_payload.get("order_index", 0)), car_id=car_id)) session.add(Image(fullres_image=str(image_payload["fullres_image"]), preview_image=str(image_payload["preview_image"]), order_index=int(image_payload.get("order_index", 0)), car_id=car_id))
@staticmethod
def _car_payload(record: CarRecord) -> dict[str, object]:
payload = record.model_dump(mode="python")
result = {key: value for key, value in payload.items() if key in CAR_DB_FIELDS}
# Serialize raw_attributes dict to JSON string for Text column.
if "raw_attributes" in result and isinstance(result["raw_attributes"], dict):
result["raw_attributes"] = json.dumps(result["raw_attributes"], ensure_ascii=False, default=str)
return result
def upsert_car(self, record: CarRecord): def upsert_car(self, record: CarRecord):
"""Insert/update/skip по content_hash.""" """Insert/update/skip по content_hash."""
# Готовим payload отдельно от вложенных изображений и служебных полей. # В БД отправляем только поля, реально существующие в финальной схеме cars.
payload = record.model_dump(mode="python") payload = self._car_payload(record)
images = payload.pop("images", []) images = [image.model_dump(mode="python") for image in record.images]
payload.pop("raw_attributes", None) content_hash = str(payload.get("content_hash") or "")
payload.pop("mapping_notes", None)
content_hash = payload.pop("content_hash", "")
payload["content_hash"] = content_hash
with self.session_scope() as session: with self.session_scope() as session:
# поиск по origin_id # поиск по origin_id
car = session.execute(select(Car).where(Car.origin_id == record.origin_id)).scalar_one_or_none() car = session.execute(select(Car).where(Car.origin_id == record.origin_id)).scalar_one_or_none()

View File

@@ -54,6 +54,7 @@ class Car(Base):
slug: Mapped[str] = mapped_column(String(), nullable=False) slug: Mapped[str] = mapped_column(String(), nullable=False)
last_seen_at: Mapped[datetime] = mapped_column(DateTime(timezone=True), nullable=False, default=func.now()) last_seen_at: Mapped[datetime] = mapped_column(DateTime(timezone=True), nullable=False, default=func.now())
content_hash: Mapped[str] = mapped_column(String(64), nullable=False, default="", index=True) content_hash: Mapped[str] = mapped_column(String(64), nullable=False, default="", index=True)
raw_attributes: Mapped[str | None] = mapped_column(Text, nullable=True)
images: Mapped[list["Image"]] = relationship("Image", back_populates="car", cascade="all, delete-orphan") images: Mapped[list["Image"]] = relationship("Image", back_populates="car", cascade="all, delete-orphan")

View File

@@ -104,6 +104,18 @@ class TestPersistenceServiceIntegration(unittest.TestCase):
self.assertEqual(result["action"], "skipped") self.assertEqual(result["action"], "skipped")
self.assertEqual(result["images_upserted"], 0) self.assertEqual(result["images_upserted"], 0)
def test_persistence_ignores_non_db_fields(self) -> None:
record = self._record("1000", content_hash="schema-test")
record.raw_attributes = {"vin": "123"}
record.mapping_notes = ["note"]
result = self.persistence.upsert_car(record)
self.assertEqual(result["action"], "inserted")
with self.persistence.session_scope() as session:
car = session.execute(select(Car).where(Car.origin_id == "1000")).scalar_one()
self.assertEqual(car.origin_id, "1000")
if __name__ == "__main__": if __name__ == "__main__":
unittest.main() unittest.main()

View File

@@ -25,6 +25,30 @@ class TestCarMapper(unittest.TestCase):
# Длина hex-представления SHA-256 # Длина hex-представления SHA-256
self.assertEqual(len(record.content_hash), 64) self.assertEqual(len(record.content_hash), 64)
def test_content_hash_changes_when_image_set_changes(self) -> None:
record_one = self.mapper.map_to_car_record(
vehicle_url="https://www.iaai.com/VehicleDetail/45089484~US",
vehicle_summary={
"make": "Toyota",
"model": "Camry",
"year": "2014",
"image_urls": ["https://vis.iaai.com/resizer?imageKeys=1&width=845&height=633"],
},
payload_insights={"vehicle_core": {}, "pricing": {}, "damage": {}, "auction": {}, "images": {}},
)
record_two = self.mapper.map_to_car_record(
vehicle_url="https://www.iaai.com/VehicleDetail/45089484~US",
vehicle_summary={
"make": "Toyota",
"model": "Camry",
"year": "2014",
"image_urls": ["https://vis.iaai.com/resizer?imageKeys=2&width=845&height=633"],
},
payload_insights={"vehicle_core": {}, "pricing": {}, "damage": {}, "auction": {}, "images": {}},
)
self.assertNotEqual(record_one.content_hash, record_two.content_hash)
def test_deduplicates_images_by_image_key(self) -> None: def test_deduplicates_images_by_image_key(self) -> None:
urls = [ urls = [
"https://vis.iaai.com/resizer?imageKeys=1&width=200&height=150", "https://vis.iaai.com/resizer?imageKeys=1&width=200&height=150",

View File

@@ -52,8 +52,9 @@ class TestScraperSync(unittest.TestCase):
scraper.persistence.upsert_car = MagicMock(return_value={"action": "inserted", "images_upserted": 1}) scraper.persistence.upsert_car = MagicMock(return_value={"action": "inserted", "images_upserted": 1})
scraper.collect_listing = MagicMock(return_value={"vehicle_urls": ["https://www.iaai.com/VehicleDetail/222~US"]}) scraper.collect_listing = MagicMock(return_value={"vehicle_urls": ["https://www.iaai.com/VehicleDetail/222~US"]})
page = MagicMock()
scraper._scrape_on_page = MagicMock(return_value={"db_record": make_db_record("222")}) scraper._scrape_on_page = MagicMock(return_value={"db_record": make_db_record("222")})
scraper._get_page = MagicMock(return_value=MagicMock()) scraper._get_page = MagicMock(return_value=page)
scraper.car_mapper.map_to_car_record = MagicMock(side_effect=AssertionError("should not be called")) scraper.car_mapper.map_to_car_record = MagicMock(side_effect=AssertionError("should not be called"))
result = scraper.sync_listing() result = scraper.sync_listing()
@@ -63,6 +64,7 @@ class TestScraperSync(unittest.TestCase):
self.assertIn("trace_id", result) self.assertIn("trace_id", result)
self.assertIn("elapsed_seconds", result) self.assertIn("elapsed_seconds", result)
self.assertEqual(scraper.persistence.upsert_car.call_count, 1) self.assertEqual(scraper.persistence.upsert_car.call_count, 1)
page.close.assert_called_once()
def test_sync_listing_respects_limit(self) -> None: def test_sync_listing_respects_limit(self) -> None:
scraper = self._make_scraper() scraper = self._make_scraper()
@@ -84,6 +86,21 @@ class TestScraperSync(unittest.TestCase):
self.assertEqual(scraper._scrape_on_page.call_count, 1) self.assertEqual(scraper._scrape_on_page.call_count, 1)
def test_sync_listing_does_not_count_skipped_as_upserted(self) -> None:
scraper = self._make_scraper()
scraper.persistence.create_tables = MagicMock()
scraper.persistence.start_sync_run = MagicMock(return_value=4)
scraper.persistence.finish_sync_run = MagicMock()
scraper.persistence.upsert_car = MagicMock(return_value={"action": "skipped", "images_upserted": 0})
scraper.collect_listing = MagicMock(return_value={"vehicle_urls": ["https://www.iaai.com/VehicleDetail/444~US"]})
scraper._scrape_on_page = MagicMock(return_value={"db_record": make_db_record("444")})
scraper._get_page = MagicMock(return_value=MagicMock())
result = scraper.sync_listing()
self.assertEqual(result["cars_upserted"], 0)
def test_close_resets_browser_state(self) -> None: def test_close_resets_browser_state(self) -> None:
scraper = self._make_scraper() scraper = self._make_scraper()
scraper.context = MagicMock() scraper.context = MagicMock()