rusackas commented on code in PR #45507:
URL: https://github.com/apache/superset/pull/45507#discussion_r4321120200
##########
superset/db_engine_specs/oceanbase.py:
##########
@@ -201,3 +203,83 @@ class OceanBaseEngineSpec(MySQLEngineSpec):
{},
),
}
+
+
+def monkeypatch_dialect() -> None:
+ """
+ Monkeypatch oceanbase_py's SQLAlchemy dialect to work under SQLAlchemy 2.0.
+
+ ``OceanBaseDialect.has_table()`` and ``get_columns()`` both pass a raw SQL
+ string straight to ``Connection.execute()`` (e.g.
+ ``connection.execute(f"DESCRIBE {full_name}")``), which SQLAlchemy 2.0
+ rejects outright with ``ObjectNotExecutableError``. Every *other* raw-SQL
+ method in the same dialect module already uses
+ ``connection.exec_driver_sql(...)`` instead, so this looks like an
+ isolated oversight rather than a deliberate choice. ``has_table()`` also
+ never catches the error ``DESCRIBE`` raises for a table that doesn't
+ exist (MySQL/OceanBase error 1146), so even with that call fixed it can
+ only ever return ``True``, raising instead of returning ``False`` for
+ exactly the case ``MetaData.create_all()``'s default ``checkfirst=True``
+ relies on.
+
+ The ``oceanbase_py`` PyPI package has no public source repository to open
+ an issue or PR against, so this is filed upstream by email instead; patch
+ here until (if ever) a fixed release ships.
+ """
+ try:
+ from oceanbase_py.sqlalchemy import datatype
+ from oceanbase_py.sqlalchemy.dialect import OceanBaseDialect
+ except ImportError:
+ return
+
+ def has_table(
+ self: Any,
+ connection: Connection,
+ table_name: str,
+ schema: Optional[str] = None,
+ **kw: Any,
+ ) -> bool:
+ if schema is None:
+ schema = self.default_schema_name
+ quote = self.identifier_preparer.quote_identifier
+ full_name = quote(table_name)
+ if schema:
+ full_name = f"{quote(schema)}.{full_name}"
+ try:
+ res = connection.exec_driver_sql(f"DESCRIBE {full_name}")
+ except ProgrammingError:
+ return False
Review Comment:
Good catch, narrowed it. `has_table` now only returns False for error 1146
(table does not exist) and re-raises any other `ProgrammingError`, so a syntax
or permission error no longer shows up as a bogus `NoSuchTableError`. Added a
test for the re-raise case. Pushed in fb7ca7c.
--
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]
