rusackas commented on code in PR #42211:
URL: https://github.com/apache/superset/pull/42211#discussion_r3679356860
##########
tests/unit_tests/models/core_test.py:
##########
@@ -1375,6 +1375,85 @@ def test_purge_oauth2_tokens(session: Session) -> None:
assert database.name == "my_oauth2_db"
+def test_purge_oauth2_tokens_scoped_by_database_id(session: Session) -> None:
+ """
+ Ensure `purge_oauth2_tokens` filters by ``database_id``, not by the
+ token-table primary key.
+
+ The existing ``test_purge_oauth2_tokens`` case inserts a single token per
+ database in an empty schema, so token PKs and database PKs align 1:1 by
+ coincidence. This regression test inserts several tokens on ``database1``
+ first so token PKs advance past 1, then creates ``database2`` and purges
+ it. All of ``database1``'s tokens must remain untouched.
+ """
+ from flask_appbuilder.security.sqla.models import Role, User # noqa: F401
+
+ from superset.models.core import Database, DatabaseUserOAuth2Tokens
+
+ Database.metadata.create_all(session.get_bind()) # pylint:
disable=no-member
+
+ user = User(
+ first_name="Alice",
+ last_name="Doe",
+ email="[email protected]",
+ username="adoe2",
+ )
+ session.add(user)
+ session.flush()
+
+ database1 = Database(database_name="db1_pk_drift",
sqlalchemy_uri="sqlite://")
+ session.add(database1)
+ session.flush()
+
+ # Insert several tokens on database1 so token PKs advance past 1 and drift
+ # away from any single database PK.
+ for i in range(5):
+ session.add(
+ DatabaseUserOAuth2Tokens(
+ user_id=user.id,
+ database_id=database1.id,
+ access_token=f"db1_access_{i}", # noqa: S106
+ access_token_expiration=datetime(2023, 1, 1),
+ refresh_token=f"db1_refresh_{i}", # noqa: S106
+ )
+ )
+ session.flush()
+
+ database2 = Database(database_name="db2_pk_drift",
sqlalchemy_uri="sqlite://")
+ session.add(database2)
+ session.flush()
+
+ assert (
+ session.query(DatabaseUserOAuth2Tokens)
+ .filter_by(database_id=database1.id)
+ .count()
+ == 5
+ )
+ assert (
+ session.query(DatabaseUserOAuth2Tokens)
+ .filter_by(database_id=database2.id)
+ .count()
+ == 0
+ )
+
+ # Purging database2 must not touch database1's tokens, even though one
+ # of database1's token PKs will collide with database2's PK.
+ database2.purge_oauth2_tokens()
+
+ assert (
+ session.query(DatabaseUserOAuth2Tokens)
+ .filter_by(database_id=database1.id)
+ .count()
+ == 5
+ )
+ assert (
+ session.query(DatabaseUserOAuth2Tokens)
+ .filter_by(database_id=database2.id)
+ .count()
+ == 0
+ )
Review Comment:
Good call, added that. Purging database1 now asserts all 5 of its tokens get
deleted, not just one, which is what the old id-based filter would've left
standing. Pushed to the branch.
##########
tests/unit_tests/models/core_test.py:
##########
@@ -1375,6 +1375,85 @@ def test_purge_oauth2_tokens(session: Session) -> None:
assert database.name == "my_oauth2_db"
+def test_purge_oauth2_tokens_scoped_by_database_id(session: Session) -> None:
+ """
+ Ensure `purge_oauth2_tokens` filters by ``database_id``, not by the
+ token-table primary key.
+
+ The existing ``test_purge_oauth2_tokens`` case inserts a single token per
+ database in an empty schema, so token PKs and database PKs align 1:1 by
+ coincidence. This regression test inserts several tokens on ``database1``
+ first so token PKs advance past 1, then creates ``database2`` and purges
+ it. All of ``database1``'s tokens must remain untouched.
+ """
+ from flask_appbuilder.security.sqla.models import Role, User # noqa: F401
+
+ from superset.models.core import Database, DatabaseUserOAuth2Tokens
+
+ Database.metadata.create_all(session.get_bind()) # pylint:
disable=no-member
+
+ user = User(
+ first_name="Alice",
+ last_name="Doe",
+ email="[email protected]",
+ username="adoe2",
+ )
+ session.add(user)
+ session.flush()
+
+ database1 = Database(database_name="db1_pk_drift",
sqlalchemy_uri="sqlite://")
+ session.add(database1)
+ session.flush()
+
+ # Insert several tokens on database1 so token PKs advance past 1 and drift
+ # away from any single database PK.
+ for i in range(5):
+ session.add(
+ DatabaseUserOAuth2Tokens(
+ user_id=user.id,
+ database_id=database1.id,
+ access_token=f"db1_access_{i}", # noqa: S106
+ access_token_expiration=datetime(2023, 1, 1),
+ refresh_token=f"db1_refresh_{i}", # noqa: S106
+ )
+ )
Review Comment:
Fair question, but that's pre-existing behavior separate from this bug.
There's no unique constraint on user_id + database_id today, just an index for
lookups, so nothing stops multiple tokens from piling up. Worth a follow-up if
we want to enforce one token per user per db, but I don't think it should block
this fix.
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
To unsubscribe, e-mail: [email protected]
For queries about this service, please contact Infrastructure at:
[email protected]
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
