Refactor to new ingestion pipeline architecture with discovery, fetch, enrichment services

This commit is contained in:
qananasikq
2026-04-20 16:27:42 +03:00
parent 55203d33ea
commit 65d5f8e1eb
23 changed files with 3436 additions and 518 deletions

View File

@@ -0,0 +1,217 @@
from __future__ import annotations
import unittest
from unittest.mock import patch
from iaai_scraper.core.config import Settings
from iaai_scraper.discovery.sitemap import (
SitemapBlockedError,
SitemapDiscoveryError,
SitemapFetchResult,
_filter_vehicle_urls,
discover_vehicle_urls_from_sitemap,
discover_vehicle_urls_from_sitemap_with_stats,
)
class TestSitemapDiscovery(unittest.TestCase):
@staticmethod
def _fetch_result(url: str, payload: bytes, *, source: str = "curl_cffi") -> SitemapFetchResult:
return SitemapFetchResult(url=url, payload=payload, source=source)
def test_filter_vehicle_urls_dedupes_and_normalizes(self) -> None:
urls = _filter_vehicle_urls([
"https://www.iaai.com/VehicleDetail/111~US?foo=1",
"https://www.iaai.com/VehicleDetail/111~US?bar=2",
"https://www.iaai.com/VehicleDetail/222~US",
"https://www.iaai.com/about",
])
self.assertEqual(
urls,
[
"https://www.iaai.com/VehicleDetail/111~US",
"https://www.iaai.com/VehicleDetail/222~US",
],
)
def test_discover_vehicle_urls_from_sitemap(self) -> None:
index_xml = b"""
<sitemapindex xmlns=\"http://www.sitemaps.org/schemas/sitemap/0.9\">
<sitemap><loc>https://www.iaai.com/sitemap-a.xml</loc></sitemap>
<sitemap><loc>https://www.iaai.com/sitemap-b.xml</loc></sitemap>
<sitemap><loc>https://www.iaai.com/sitemapauctions1.xml</loc></sitemap>
</sitemapindex>
"""
sitemap_a = b"""
<urlset xmlns=\"http://www.sitemaps.org/schemas/sitemap/0.9\">
<url><loc>https://www.iaai.com/VehicleDetail/111~US</loc></url>
<url><loc>https://www.iaai.com/VehicleDetail/222~US?x=1</loc></url>
</urlset>
"""
sitemap_b = b"""
<urlset xmlns=\"http://www.sitemaps.org/schemas/sitemap/0.9\">
<url><loc>https://www.iaai.com/VehicleDetail/222~US?y=2</loc></url>
<url><loc>https://www.iaai.com/VehicleDetail/333~US</loc></url>
</urlset>
"""
def _fake_fetch(url: str) -> SitemapFetchResult:
if url.endswith("sitemap_index.xml"):
return self._fetch_result(url, index_xml)
if url.endswith("sitemap-a.xml"):
return self._fetch_result(url, sitemap_a)
if url.endswith("sitemap-b.xml"):
return self._fetch_result(url, sitemap_b)
raise AssertionError(f"unexpected url: {url}")
with patch("iaai_scraper.discovery.sitemap._SitemapDownloader.fetch", side_effect=_fake_fetch):
result = discover_vehicle_urls_from_sitemap("https://www.iaai.com/sitemap_index.xml")
self.assertEqual(
result,
[
"https://www.iaai.com/VehicleDetail/111~US",
"https://www.iaai.com/VehicleDetail/222~US",
"https://www.iaai.com/VehicleDetail/333~US",
],
)
def test_discover_vehicle_urls_raises_on_empty_index(self) -> None:
empty_index = b"<sitemapindex xmlns=\"http://www.sitemaps.org/schemas/sitemap/0.9\"></sitemapindex>"
with patch(
"iaai_scraper.discovery.sitemap._SitemapDownloader.fetch",
return_value=self._fetch_result("https://www.iaai.com/sitemap_index.xml", empty_index),
):
with self.assertRaises(SitemapDiscoveryError):
discover_vehicle_urls_from_sitemap("https://www.iaai.com/sitemap_index.xml")
def test_discover_vehicle_urls_skips_malformed_and_non_vehicle_sitemaps(self) -> None:
sitemap_vehicle = b"""
<urlset xmlns=\"http://www.sitemaps.org/schemas/sitemap/0.9\">
<url><loc>https://www.iaai.com/VehicleDetail/111~US</loc></url>
</urlset>
"""
sitemap_non_vehicle = b"""
<urlset xmlns=\"http://www.sitemaps.org/schemas/sitemap/0.9\">
<url><loc>https://www.iaai.com/SalesList/111~US/04222026</loc></url>
</urlset>
"""
index_xml = b"""
<sitemapindex xmlns=\"http://www.sitemaps.org/schemas/sitemap/0.9\">
<sitemap><loc>https://www.iaai.com/sitemap1.xml</loc></sitemap>
<sitemap><loc>https://www.iaai.com/sitemapbranches1.xml</loc></sitemap>
<sitemap><loc>https://www.iaai.com/sitemap2.xml</loc></sitemap>
<sitemap><loc>https://www.iaai.com/sitemap3.xml</loc></sitemap>
</sitemapindex>
"""
def _fake_fetch(url: str) -> SitemapFetchResult:
if url.endswith("sitemap_index.xml"):
return self._fetch_result(url, index_xml)
if url.endswith("sitemap1.xml"):
return self._fetch_result(url, sitemap_vehicle)
if url.endswith("sitemap2.xml"):
return self._fetch_result(url, sitemap_non_vehicle)
if url.endswith("sitemap3.xml"):
raise SitemapDiscoveryError("broken sitemap")
raise AssertionError(f"unexpected url: {url}")
with patch("iaai_scraper.discovery.sitemap._SitemapDownloader.fetch", side_effect=_fake_fetch):
result = discover_vehicle_urls_from_sitemap("https://www.iaai.com/sitemap_index.xml")
self.assertEqual(result, ["https://www.iaai.com/VehicleDetail/111~US"])
def test_discover_vehicle_urls_falls_back_to_regex_loc_extraction(self) -> None:
malformed_index = (
b"<sitemapindex>"
b"<sitemap><loc>https://www.iaai.com/sitemap1.xml</loc></sitemap>"
b"<broken"
)
sitemap_vehicle = b"""
<urlset xmlns=\"http://www.sitemaps.org/schemas/sitemap/0.9\">
<url><loc>https://www.iaai.com/VehicleDetail/111~US</loc></url>
</urlset>
"""
def _fake_fetch(url: str) -> SitemapFetchResult:
if url.endswith("sitemap_index.xml"):
return self._fetch_result(url, malformed_index)
if url.endswith("sitemap1.xml"):
return self._fetch_result(url, sitemap_vehicle)
raise AssertionError(f"unexpected url: {url}")
with patch("iaai_scraper.discovery.sitemap._SitemapDownloader.fetch", side_effect=_fake_fetch):
result = discover_vehicle_urls_from_sitemap("https://www.iaai.com/sitemap_index.xml")
self.assertEqual(result, ["https://www.iaai.com/VehicleDetail/111~US"])
def test_discovery_uses_direct_probe_when_index_has_no_urls(self) -> None:
index_xml = b"<sitemapindex xmlns=\"http://www.sitemaps.org/schemas/sitemap/0.9\"></sitemapindex>"
sitemap_one = b"""
<urlset xmlns=\"http://www.sitemaps.org/schemas/sitemap/0.9\">
<url><loc>https://www.iaai.com/VehicleDetail/111~US</loc></url>
</urlset>
"""
settings = Settings()
settings.discovery.sitemap_direct_probe_limit = 2
settings.discovery.sitemap_direct_probe_stop_after_misses = 1
def _fake_fetch(url: str) -> SitemapFetchResult:
if url.endswith("sitemap_index.xml"):
return self._fetch_result(url, index_xml)
if url.endswith("sitemap1.xml"):
return self._fetch_result(url, sitemap_one)
raise SitemapDiscoveryError("not found")
with patch("iaai_scraper.discovery.sitemap._SitemapDownloader.fetch", side_effect=_fake_fetch):
result = discover_vehicle_urls_from_sitemap_with_stats(
"https://www.iaai.com/sitemap_index.xml",
settings=settings,
)
self.assertEqual(result.vehicle_urls, ["https://www.iaai.com/VehicleDetail/111~US"])
self.assertEqual(result.stats.direct_probe_hits, 1)
def test_discovery_stats_report_direct_probe_hits(self) -> None:
index_xml = b"<sitemapindex xmlns=\"http://www.sitemaps.org/schemas/sitemap/0.9\"></sitemapindex>"
sitemap1 = b"""
<urlset xmlns=\"http://www.sitemaps.org/schemas/sitemap/0.9\">
<url><loc>https://www.iaai.com/VehicleDetail/111~US</loc></url>
</urlset>
"""
def _fake_fetch(url: str):
if url.endswith("sitemap_index.xml"):
return self._fetch_result(url, index_xml)
if url.endswith("sitemap1.xml"):
return self._fetch_result(url, sitemap1)
raise SitemapDiscoveryError("not found")
with patch("iaai_scraper.discovery.sitemap._SitemapDownloader.fetch", side_effect=_fake_fetch):
result = discover_vehicle_urls_from_sitemap_with_stats("https://www.iaai.com/sitemap_index.xml")
self.assertEqual(result.vehicle_urls, ["https://www.iaai.com/VehicleDetail/111~US"])
self.assertEqual(result.stats.transport, "curl_cffi")
self.assertEqual(result.stats.direct_probe_hits, 1)
self.assertGreaterEqual(result.stats.direct_probe_misses, 1)
def test_block_page_raises_discovery_error(self) -> None:
settings = Settings()
settings.discovery.sitemap_direct_probe_limit = 1
settings.discovery.sitemap_direct_probe_stop_after_misses = 1
with patch(
"iaai_scraper.discovery.sitemap._SitemapDownloader.fetch",
side_effect=SitemapBlockedError("Anti-bot page returned for https://www.iaai.com/sitemap_index.xml"),
):
with self.assertRaises(SitemapDiscoveryError):
discover_vehicle_urls_from_sitemap(
"https://www.iaai.com/sitemap_index.xml",
settings=settings,
)
if __name__ == "__main__":
unittest.main()