jun-he commented on a change in pull request #4016:
URL: https://github.com/apache/iceberg/pull/4016#discussion_r796326867
##########
File path: python/src/iceberg/types.py
##########
@@ -15,61 +15,84 @@
# specific language governing permissions and limitations
# under the License.
"""Data types used in describing Iceberg schemas
-
This module implements the data types described in the Iceberg specification
for Iceberg schemas. To
describe an Iceberg table schema, these classes can be used in the
construction of a StructType instance.
-
Example:
>>> StructType(
[
NestedField(True, 1, "required_field", StringType()),
NestedField(False, 2, "optional_field", IntegerType()),
]
)
-
Notes:
- https://iceberg.apache.org/#spec/#primitive-types
"""
-from typing import Optional
+from typing import Any, Dict, List, Optional, Tuple
-class Type:
- def __init__(self, type_string: str, repr_string: str, is_primitive=False):
- self._type_string = type_string
- self._repr_string = repr_string
- self._is_primitive = is_primitive
+class IcebergType:
+
+ _implemented: Dict[Tuple[str, Tuple[Any]], "IcebergType"] = {}
+
+ def __new__(cls, *args, **kwargs):
Review comment:
I agree with @rdblue about it. It would be much simpler and easy to
understand and maintain if we do that separately. I gave it a try in
https://github.com/apache/iceberg/pull/3965 and seems quite straightforward.
For the decorator example, while the object created using FixedType will be
singleton, but IIUC, the FixedType itself is a function not a class any more,
right? So we cannot call class methods from it, etc. It is better we don't use
decorator.
--
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]
