WweiL commented on code in PR #40736:
URL: https://github.com/apache/spark/pull/40736#discussion_r1163158951
##########
python/pyspark/sql/connect/proto/catalog_pb2.pyi:
##########
@@ -50,6 +50,7 @@ else:
DESCRIPTOR: google.protobuf.descriptor.FileDescriptor
+@typing_extensions.final
Review Comment:
Any idea why these are added?
##########
python/pyspark/sql/connect/group.py:
##########
@@ -262,8 +264,44 @@ def applyInPandas(
applyInPandas.__doc__ = PySparkGroupedData.applyInPandas.__doc__
- def applyInPandasWithState(self, *args: Any, **kwargs: Any) -> None:
- raise NotImplementedError("applyInPandasWithState() is not
implemented.")
+ def applyInPandasWithState(
+ self,
+ func: "PandasGroupedMapFunctionWithState",
+ outputStructType: Union[StructType, str],
Review Comment:
I see that other part of the code use quoted types, i.e.
`Union["StructType", str]`. Maybe we should also do that for code consistency?
Also doing that seems to [help with forward
consistency](https://stackoverflow.com/questions/46458470/should-you-put-quotes-around-type-annotations-in-python)
##########
python/pyspark/sql/connect/group.py:
##########
@@ -262,8 +264,44 @@ def applyInPandas(
applyInPandas.__doc__ = PySparkGroupedData.applyInPandas.__doc__
- def applyInPandasWithState(self, *args: Any, **kwargs: Any) -> None:
- raise NotImplementedError("applyInPandasWithState() is not
implemented.")
+ def applyInPandasWithState(
+ self,
+ func: "PandasGroupedMapFunctionWithState",
+ outputStructType: Union[StructType, str],
+ stateStructType: Union[StructType, str],
+ outputMode: str,
+ timeoutConf: str
+ ) -> "DataFrame":
+ from pyspark.sql.connect.udf import UserDefinedFunction
+ from pyspark.sql.connect.dataframe import DataFrame
+
+ udf_obj = UserDefinedFunction(
+ func,
+ returnType=outputStructType,
+ evalType=PythonEvalType.SQL_GROUPED_MAP_PANDAS_UDF_WITH_STATE,
+ )
+
+ output_schema: str = outputStructType.json() if
isinstance(outputStructType, StructType)\
Review Comment:
I'm not very sure if we could do like this? In `UserDefinedFunction` above
the `returnRype` is of the same type, and it handles it differently
https://github.com/apache/spark/blob/007a42b8511b005aabdf45955e504d1f71eb1d25/python/pyspark/sql/connect/udf.py#L101
and then
https://github.com/apache/spark/blob/007a42b8511b005aabdf45955e504d1f71eb1d25/python/pyspark/sql/connect/expressions.py#L595-L598
Should we also follow this pattern?
Maybe @ueshin has more context?
--
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]
