Hyukjin Kwon created SPARK-61034:
------------------------------------
Summary: Send the output schema of transformWithState in the
Python client as PySpark does
Key: SPARK-61034
URL: https://issues.apache.org/jira/browse/SPARK-61034
Project: Spark
Issue Type: Bug
Components: Connect
Affects Versions: connect-rust-4.3.0
Reporter: Hyukjin Kwon
In PySpark Connect, {{df.groupBy(...).transformWithState(statefulProcessor,
outputStructType, outputMode, timeMode)}} sends {{outputStructType}} as the
{{output_schema}} of the {{TransformWithStateInfo}} of its {{GroupMap}}, as
{{transformWithStateInPandas}} does. The server reads it as the schema of the
output rows.
The Rust-backed Python client, and the Rust API
{{GroupedData::transform_with_state}}, send no output schema for
{{transformWithState}}. So the server fails to analyze the DataFrame with
{{[CONNECT_INVALID_PLAN.DATA_TYPE_UNSUPPORTED_PROTO_TO_CATALYST] The Spark
Connect plan is invalid. Does not support convert KIND_NOT_SET to catalyst
types.}} Send it as PySpark does.
--
This message was sent by Atlassian Jira
(v10.3.26#10030026)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
