Implement identity graph stitching and configurable profile storage

Issue persistent signed anonymous IDs and accept cookie/token-bound events. Resolve verified email, phone, OAuth, and namespaced third-party IDs into transactionally merged master identities with historical event aliases. Protect shared-device account switching, tenant isolation, and event retries. Store sparse canonical profiles as PostgreSQL JSONB or local SQLite JSON. Support an optional versioned Cassandra wide-column projection through a transactional outbox, retry worker, and immediate canonical read fallback. Document deployment settings and provide opt-in database integration tests. Validation: 43 local tests passed; 6 external database tests skipped. Ruff checks, formatting, and offline uv lock verification passed.
2026-09-15 23:49:43 +08:00
parent c30a136102
commit 0fc6c68a99
27 changed files with 2993 additions and 103 deletions
@@ -0,0 +1,87 @@
+"""Optional supplied Cassandra environment; never install/start a server here."""
+
+import os
+from uuid import uuid4
+
+import pytest
+
+from adcloud_server.database import Database
+from adcloud_server.identity_models import ProfilePatch, StitchRequest
+from adcloud_server.identity_service import IdentityService
+from adcloud_server.profile_store import CassandraProfileStore
+from adcloud_server.projection import ProfileProjector
+from adcloud_server.settings import Settings
+
+pytestmark = pytest.mark.skipif(
+    not os.getenv("ADCLOUD_TEST_CASSANDRA_HOSTS"),
+    reason="No Cassandra integration cluster configured",
+)
+
+
+@pytest.fixture
+def components(tmp_path):
+    settings = Settings(
+        database_url="sqlite:///" + str(tmp_path / "graph.db"),
+        profile_backend="cassandra",
+        tenant_id="test-" + str(uuid4()),
+        identity_secret="test-cassandra-" + "x" * 48,
+        cassandra_hosts=tuple(os.environ["ADCLOUD_TEST_CASSANDRA_HOSTS"].split(",")),
+        cassandra_port=int(os.getenv("ADCLOUD_TEST_CASSANDRA_PORT", "9042")),
+        cassandra_keyspace=os.environ["ADCLOUD_TEST_CASSANDRA_KEYSPACE"],
+        cassandra_username=os.getenv("ADCLOUD_TEST_CASSANDRA_USERNAME", ""),
+        cassandra_password=os.getenv("ADCLOUD_TEST_CASSANDRA_PASSWORD", ""),
+        cassandra_tls=os.getenv("ADCLOUD_TEST_CASSANDRA_TLS", "true").lower() == "true",
+        cassandra_ca_file=os.getenv("ADCLOUD_TEST_CASSANDRA_CA_FILE") or None,
+    )
+    db = Database(settings.database_url, settings.tenant_id)
+    db.initialize()
+    store = CassandraProfileStore(settings)
+    try:
+        yield IdentityService(settings, db, store), ProfileProjector(db, store), store
+    finally:
+        store.close()
+        db.close()
+
+
+def test_cassandra_sparse_snapshot_roundtrip(components):
+    service, _worker, store = components
+    master = str(uuid4())
+    document = {
+        "attributes": {"nested": {"interest": "running"}, "enabled": True, "score": 0},
+        "tags": ["runner"],
+    }
+    store.put(service.tenant, master, 1, document)
+    assert store.get(service.tenant, master, 1) == document
+    assert store.get(service.tenant, master, 2) is None
+
+
+def test_cassandra_immutable_revision_prevents_stale_attribute_resurrection(components):
+    service, _worker, store = components
+    master = str(uuid4())
+    old = {"attributes": {"remove": "old"}, "tags": []}
+    new = {"attributes": {}, "tags": []}
+    store.put(service.tenant, master, 2, new)
+    store.put(service.tenant, master, 1, old)
+    assert store.get(service.tenant, master, 2) == new
+
+
+def test_cassandra_worker_publishes_stitched_profile(components):
+    service, worker, _store = components
+    anon = service.anonymous()
+    stitched = service.stitch(
+        anon["anonymous_token"],
+        StitchRequest(
+            identifiers=[{"kind": "email", "value": "alice@example.com"}],
+            attributes={"interest": "running"},
+        ),
+    )
+    assert service.profile(stitched["master_id"])["projection_status"] == "pending"
+    worker.flush()
+    current = service.profile(stitched["master_id"])
+    assert current["projection_status"] == "current"
+    assert current["attributes"] == {"interest": "running"}
+    service.patch_profile(
+        stitched["master_id"], ProfilePatch(attributes={"interest": None})
+    )
+    worker.flush()
+    assert service.profile(stitched["master_id"])["attributes"] == {}