fix: enable dev SRO collection and handle real lookup responses

2026-09-15 12:47:16 +02:00
parent 18971d33ec
commit 3d7fb78678
7 changed files with 597 additions and 30 deletions
--- a/tests/apps/parsers/test_sro_upstream_empty.py
+++ b/tests/apps/parsers/test_sro_upstream_empty.py
@@ -0,0 +1,260 @@
+"""Recognize the observed lookup 404 template without publishing an outage as empty."""
+
+import json
+import zipfile
+from unittest.mock import Mock
+from urllib.parse import parse_qs, urlsplit
+
+import pytest
+from apps.parsers.models import ParserSourceArtifact, SroOrganizationLookup
+from apps.parsers.registry_snapshots import SnapshotValidationError
+from apps.parsers.sro_http import SroHttpClient, SroPage
+from apps.parsers.sro_membership import (
+    SRO_LOOKUP_URL,
+    parse_sro_lookup,
+    refresh_sro_membership,
+)
+from django.utils import timezone
+from organizations.models import Organization, OrganizationSourceRecord
+
+from tests.apps.parsers.test_sro_membership import FixtureSite
+
+LOOKUP_TITLE = (
+    "Проверка членства в реестре СРО, проверить допуск организации в СРО по ИНН"
+)
+NOT_FOUND_HEADING = "Запрашиваемая страница на сайте отсутствует."
+LOOKUP_EMPTY = (
+    f"<html><head><title>{LOOKUP_TITLE}</title></head>"
+    f"<body><h1>{NOT_FOUND_HEADING}</h1></body></html>"
+).encode()
+
+
+@pytest.fixture(autouse=True)
+def sro_settings(settings):
+    settings.SRO_DEV_COLLECTION_ENABLED = True
+
+
+def response_404(body=LOOKUP_EMPTY):
+    response = Mock(status_code=404, is_redirect=False)
+    response.iter_content.return_value = [body]
+    return response
+
+
+@pytest.mark.parametrize("query", ("1234567890", "1027700132195"))
+def test_http_returns_bounded_lookup_404_with_status(query):
+    response = response_404()
+    session = Mock()
+    session.get.return_value = response
+    url = f"{SRO_LOOKUP_URL}?q={query}&sro_name=&search=Find"
+    page = SroHttpClient(session=session).get(url)
+    assert (page.url, page.body, page.status_code) == (url, LOOKUP_EMPTY, 404)
+    session.get.assert_called_once()
+    response.close.assert_called_once()
+
+
+@pytest.mark.parametrize(
+    "suffix",
+    (
+        "/",
+        "/members/?q=1234567890",
+        "/proverka_dopuska/extra/?q=1234567890",
+        "/proverka_dopuska/",
+        "/proverka_dopuska/?q=12345678901",
+        "/proverka_dopuska/?q=1234567890&q=",
+        "/proverka_dopuska/?q=1234567890&q=1027700132195",
+        "/proverka_dopuska/?q=not-an-identifier",
+        "/proverka_dopuska/?q=１２３４５６７８９０",
+    ),
+)
+def test_other_404s_are_not_reclassified_or_read(suffix):
+    response = response_404()
+    session = Mock()
+    session.get.return_value = response
+    with pytest.raises(SnapshotValidationError, match="^sro_page_not_found$"):
+        SroHttpClient(session=session).get(f"https://www.reestr-sro.ru{suffix}")
+    response.iter_content.assert_not_called()
+    response.close.assert_called_once()
+
+
+def test_lookup_404_response_body_size_limit(settings):
+    settings.SRO_HTTP_MAX_RESPONSE_BYTES = 4
+    response = response_404(b"12345")
+    session = Mock()
+    session.get.return_value = response
+    with pytest.raises(SnapshotValidationError, match="^source_response_too_large$"):
+        SroHttpClient(session=session).get(f"{SRO_LOOKUP_URL}?q=1234567890")
+    response.close.assert_called_once()
+
+
+def test_parser_recognizes_only_observed_negative_404():
+    page = SroPage(f"{SRO_LOOKUP_URL}?q=1234567890", LOOKUP_EMPTY, status_code=404)
+    assert parse_sro_lookup(page) == ([], None)
+    assert SroPage(page.url, b"fixture").status_code == 200
+
+
+@pytest.mark.parametrize(
+    "body",
+    (
+        b"<html><h1>404 Not Found</h1></html>",
+        b"<html><title>Login</title><h1>Sign in</h1></html>",
+        "Ничего не найдено".encode(),
+        LOOKUP_EMPTY.replace(LOOKUP_TITLE.encode(), b"Other source"),
+        LOOKUP_EMPTY.replace(NOT_FOUND_HEADING.encode(), b"Captcha"),
+        LOOKUP_EMPTY.replace(b"</body>", b"<h1>Other</h1></body>"),
+        LOOKUP_EMPTY.replace(b"</body>", b'<table class="sro-members"></table></body>'),
+    ),
+)
+def test_parser_rejects_unrecognized_404_even_with_empty_markers(body):
+    page = SroPage(f"{SRO_LOOKUP_URL}?q=1234567890", body, status_code=404)
+    with pytest.raises(
+        SnapshotValidationError, match="^sro_lookup_response_unrecognized$"
+    ):
+        parse_sro_lookup(page)
+
+
+@pytest.mark.parametrize(
+    "url,status",
+    (
+        (f"{SRO_LOOKUP_URL}?q=1234567890", 200),
+        (f"{SRO_LOOKUP_URL}?q=1234567890", 503),
+        ("https://www.reestr-sro.ru/members/?q=1234567890", 404),
+        ("https://other.invalid/proverka_dopuska/?q=1234567890", 404),
+    ),
+)
+def test_template_is_not_empty_for_other_status_or_url(url, status):
+    with pytest.raises(SnapshotValidationError):
+        parse_sro_lookup(SroPage(url, LOOKUP_EMPTY, status_code=status))
+
+
+class MixedFixtureSite(FixtureSite):
+    def __init__(self, organizations, negative_ids):
+        super().__init__(organizations)
+        self.negative_ids = negative_ids
+
+    def get(self, url):
+        parts = urlsplit(url)
+        query = parse_qs(parts.query).get("q", [None])[0]
+        if parts.path == "/proverka_dopuska/" and query in self.negative_ids:
+            self.urls.append(url)
+            return SroPage(url, LOOKUP_EMPTY, status_code=404)
+        return super().get(url)
+
+
+@pytest.fixture
+def organizations(db, settings, tmp_path):
+    settings.MEDIA_ROOT = str(tmp_path / "media")
+    settings.PARSER_PRIVATE_ARTIFACT_ROOT = str(tmp_path / "private")
+    return [
+        Organization.objects.create(
+            name=f"АО Фикстура {index}",
+            inn=f"{index}234567890",
+            ogrn=f"{index}027700132195",
+            okpo=f"0012345{index}",
+            directory_imported_at=timezone.now(),
+        )
+        for index in (1, 2)
+    ]
+
+
+def test_mixed_positive_and_known_404_publishes_with_status_provenance(organizations):
+    positive, negative = organizations
+    refresh_sro_membership(load_batch=1, mode="full", client=FixtureSite(organizations))
+    artifact, result = refresh_sro_membership(
+        load_batch=2,
+        mode="full",
+        client=MixedFixtureSite(organizations, {negative.ogrn}),
+    )
+    assert result.published == 1
+    assert (
+        OrganizationSourceRecord.objects.get().extension.organization_id == positive.uid
+    )
+    assert artifact.metadata["recognized_lookup_404_count"] == 1
+    assert artifact.metadata["successful_lookup_200_count"] == 1
+    assert artifact.metadata["not_found_organizations_count"] == 1
+    assert (
+        SroOrganizationLookup.objects.get(organization=negative).artifact_id
+        == artifact.uid
+    )
+    with zipfile.ZipFile(artifact.file.path) as raw:
+        assert raw.testzip() is None
+        manifest = json.loads(raw.read("manifest.json"))
+        assert sorted(item["status_code"] for item in manifest) == [200, 200, 404]
+        negative_entry = next(item for item in manifest if item["status_code"] == 404)
+        assert raw.read(negative_entry["file"]) == LOOKUP_EMPTY
+
+
+@pytest.mark.parametrize("mode", ("full", "incremental"))
+def test_all_known_404_rejects_without_changing_previous_snapshot_or_checkpoints(
+    organizations, mode
+):
+    previous, _ = refresh_sro_membership(
+        load_batch=1, mode="full", client=FixtureSite(organizations)
+    )
+    records = list(OrganizationSourceRecord.objects.order_by("uid").values())
+    checkpoints = list(
+        SroOrganizationLookup.objects.order_by("organization_id").values()
+    )
+    for organization in organizations:
+        organization.name += " изменённая"
+        organization.save(update_fields=["name"])
+    on_publish = Mock()
+    with pytest.raises(SnapshotValidationError, match="^sro_ambiguous_empty_scan$"):
+        refresh_sro_membership(
+            load_batch=2,
+            mode=mode,
+            client=MixedFixtureSite(organizations, {org.ogrn for org in organizations}),
+            on_publish=on_publish,
+        )
+    on_publish.assert_not_called()
+    assert list(OrganizationSourceRecord.objects.order_by("uid").values()) == records
+    assert (
+        list(SroOrganizationLookup.objects.order_by("organization_id").values())
+        == checkpoints
+    )
+    previous.refresh_from_db()
+    assert previous.status == ParserSourceArtifact.Status.PUBLISHED
+    rejected = ParserSourceArtifact.objects.get(load_batch=2)
+    assert rejected.status == ParserSourceArtifact.Status.REJECTED
+    assert rejected.metadata["recognized_lookup_404_count"] == 2
+    assert rejected.metadata["successful_lookup_200_count"] == 0
+    with zipfile.ZipFile(rejected.file.path) as raw:
+        assert raw.testzip() is None
+        assert [
+            item["status_code"] for item in json.loads(raw.read("manifest.json"))
+        ] == [404, 404]
+
+
+def test_negative_redirect_to_other_identifier_preserves_snapshot(organizations):
+    positive, negative = organizations
+    refresh_sro_membership(load_batch=1, mode="full", client=FixtureSite(organizations))
+    records = list(OrganizationSourceRecord.objects.order_by("uid").values())
+    checkpoints = list(
+        SroOrganizationLookup.objects.order_by("organization_id").values()
+    )
+
+    class RedirectedSite(MixedFixtureSite):
+        def get(self, url):
+            page = super().get(url)
+            if page.status_code == 404:
+                return SroPage(
+                    f"{SRO_LOOKUP_URL}?q={positive.ogrn}", page.body, status_code=404
+                )
+            return page
+
+    with pytest.raises(
+        SnapshotValidationError, match="^sro_lookup_identifier_mismatch$"
+    ):
+        refresh_sro_membership(
+            load_batch=2,
+            mode="full",
+            client=RedirectedSite(organizations, {negative.ogrn}),
+        )
+    assert list(OrganizationSourceRecord.objects.order_by("uid").values()) == records
+    assert (
+        list(SroOrganizationLookup.objects.order_by("organization_id").values())
+        == checkpoints
+    )
+    rejected = ParserSourceArtifact.objects.get(load_batch=2)
+    assert rejected.status == ParserSourceArtifact.Status.REJECTED
+    assert rejected.metadata["recognized_lookup_404_count"] == 0
+    assert rejected.metadata["parse_errors_count"] == 1