fix docker scraping

improve batch sync

add postgres upsert

fix sync locking

improve listing sync

speed up scraper

clean up project

prepare for github

update docker setup
This commit is contained in:
qananasikq
2026-04-13 16:35:08 +03:00
parent b9557922ed
commit a4c93a1df1
30 changed files with 2239 additions and 624 deletions

View File

@@ -51,15 +51,12 @@ class TestScraperSync(unittest.TestCase):
scraper.persistence.create_tables = MagicMock()
scraper.persistence.start_sync_run = MagicMock(return_value=2)
scraper.persistence.finish_sync_run = MagicMock()
scraper.persistence.upsert_car = MagicMock(return_value={"action": "inserted", "images_upserted": 1})
scraper.persistence.get_existing_origin_urls = MagicMock(return_value=set())
scraper.persistence.get_existing_origin_ids = MagicMock(return_value=set())
scraper.persistence.get_existing_urls_and_ids = MagicMock(return_value=(set(), set()))
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._get_page = MagicMock(return_value=page)
scraper.car_mapper.map_to_car_record = MagicMock(side_effect=AssertionError("should not be called"))
scraper.sync_batch = MagicMock(return_value={
"cars_upserted": 1, "cars_failed": 0, "images_upserted": 1, "failures": [],
})
result = scraper.sync_listing()
@@ -67,8 +64,7 @@ class TestScraperSync(unittest.TestCase):
self.assertEqual(result["cars_failed"], 0)
self.assertIn("trace_id", result)
self.assertIn("elapsed_seconds", result)
self.assertEqual(scraper.persistence.upsert_car.call_count, 1)
page.close.assert_called_once()
scraper.sync_batch.assert_called_once()
def test_sync_listing_respects_limit(self) -> None:
scraper = self._make_scraper()
@@ -76,38 +72,25 @@ class TestScraperSync(unittest.TestCase):
scraper.persistence.create_tables = MagicMock()
scraper.persistence.start_sync_run = MagicMock(return_value=3)
scraper.persistence.finish_sync_run = MagicMock()
scraper.persistence.upsert_car = MagicMock(return_value={"action": "inserted", "images_upserted": 1})
scraper.persistence.get_existing_origin_urls = MagicMock(return_value=set())
scraper.persistence.get_existing_origin_ids = MagicMock(return_value=set())
scraper.collect_listing = MagicMock(return_value={
"vehicle_urls": [
"https://www.iaai.com/VehicleDetail/222~US",
"https://www.iaai.com/VehicleDetail/333~US",
]
# Проверка пути only_new с limit.
scraper._collect_listing_iterative = MagicMock(return_value=(
["https://www.iaai.com/VehicleDetail/222~US"],
["https://www.iaai.com/VehicleDetail/222~US",
"https://www.iaai.com/VehicleDetail/333~US"],
{"vehicle_urls": [], "pages_collected": 1, "early_stopped": False},
0,
))
scraper.sync_batch = MagicMock(return_value={
"cars_upserted": 1, "cars_failed": 0, "images_upserted": 1, "failures": [],
})
scraper._scrape_on_page = MagicMock(return_value={"db_record": make_db_record("222")})
scraper._get_page = MagicMock(return_value=MagicMock())
scraper.sync_listing(limit=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.persistence.get_existing_origin_urls = MagicMock(return_value=set())
scraper.persistence.get_existing_origin_ids = MagicMock(return_value=set())
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)
# Должен уйти только один URL.
scraper.sync_batch.assert_called_once()
batch_urls = scraper.sync_batch.call_args[0][0]
self.assertEqual(len(batch_urls), 1)
def test_sync_listing_only_new_filters_existing_by_url_and_origin_id(self) -> None:
scraper = self._make_scraper()
@@ -115,9 +98,10 @@ class TestScraperSync(unittest.TestCase):
scraper.persistence.create_tables = MagicMock()
scraper.persistence.start_sync_run = MagicMock(return_value=5)
scraper.persistence.finish_sync_run = MagicMock()
scraper.persistence.upsert_car = MagicMock(return_value={"action": "inserted", "images_upserted": 0})
scraper.persistence.get_existing_origin_urls = MagicMock(return_value={"https://www.iaai.com/VehicleDetail/111~US"})
scraper.persistence.get_existing_origin_ids = MagicMock(return_value={"222"})
scraper.persistence.get_existing_urls_and_ids = MagicMock(return_value=(
{"https://www.iaai.com/VehicleDetail/111~US"},
{"iaai:222"},
))
scraper.collect_listing = MagicMock(return_value={
"vehicle_urls": [
@@ -126,26 +110,34 @@ class TestScraperSync(unittest.TestCase):
"https://www.iaai.com/VehicleDetail/333~US", # new
]
})
page = MagicMock()
scraper._get_page = MagicMock(return_value=page)
scraper._scrape_on_page = MagicMock(return_value={"db_record": make_db_record("333")})
# Возвращаем результат для одного нового авто.
scraper.sync_batch = MagicMock(return_value={
"cars_upserted": 1, "cars_failed": 0, "images_upserted": 0, "failures": [],
})
result = scraper.sync_listing(only_new=True)
self.assertEqual(result["skipped_existing"], 2)
self.assertEqual(result["cars_upserted"], 1)
self.assertEqual(scraper._scrape_on_page.call_count, 1)
scraper.persistence.get_existing_origin_urls.assert_called_once()
scraper.persistence.get_existing_origin_ids.assert_called_once()
# В batch должен попасть только новый URL.
scraper.sync_batch.assert_called_once()
batch_urls = scraper.sync_batch.call_args[0][0]
self.assertEqual(len(batch_urls), 1)
self.assertIn("333", batch_urls[0])
scraper.persistence.get_existing_urls_and_ids.assert_called_once()
def test_close_resets_browser_state(self) -> None:
scraper = self._make_scraper()
http_pool = MagicMock()
scraper._http_pool = http_pool
scraper.context = MagicMock()
scraper.browser = MagicMock()
scraper.playwright = MagicMock()
scraper.close()
http_pool.clear.assert_called_once()
self.assertIsNone(scraper._http_pool)
self.assertIsNone(scraper.context)
self.assertIsNone(scraper.browser)
self.assertIsNone(scraper.playwright)
@@ -172,6 +164,11 @@ class TestScraperSync(unittest.TestCase):
"https://www.iaai.com/VehicleDetail/999~US",
)
def test_is_protection_or_network_error_detects_known_signals(self) -> None:
self.assertTrue(IAAIScraper._is_protection_or_network_error(RuntimeError("NS_ERROR_NET_INTERRUPT")))
self.assertTrue(IAAIScraper._is_protection_or_network_error(RuntimeError("captcha challenge")))
self.assertFalse(IAAIScraper._is_protection_or_network_error(RuntimeError("plain validation error")))
if __name__ == "__main__":
unittest.main()