houqp commented on a change in pull request #13278:
URL: https://github.com/apache/airflow/pull/13278#discussion_r548450671
##########
File path: airflow/utils/types.py
##########
@@ -16,8 +16,31 @@
# under the License.
import enum
+from sqlalchemy.types import TypeDecorator, String
-class DagRunType(str, enum.Enum):
+
+class EnumString(TypeDecorator):
+ """
+ Declare db column with this type to make the column compatible with string
+ and string based enum values when building the sqlalchemy ORM query. It can
+ be used just like sqlalchemy.types.String, for example:
+
+ ```
+ class Table(Base):
+ __tablename__ = "t"
+ run_type = Column(EnumString(50), nullable=False)
+ ```
+ """
+ impl = String
+
+ def process_bind_param(self, value, dialect):
+ if isinstance(value, enum.Enum):
+ return value.value
+ else:
+ return value
+
+
+class DagRunType(enum.Enum):
Review comment:
The downside of subclassing `str` is it will lead to inconsistent
behavior between db drivers for columns that are not declared as `EnumString`
type. My main concern is developer adding new db columns that are supposed to
interact with python enum types, but declared it as `String` instead of
`EnumString`. In this case, postgres users won't experience any problem, but
mysql users will end up with crashes or corrupted data in database.
What's your main concern around lack of str subclassing? Are you worried
that developer will get confused between use of `DagRunType.SCHEDULED.value`
and `DagRunType.SCHEDULED`?
----------------------------------------------------------------
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.
For queries about this service, please contact Infrastructure at:
[email protected]
