Hyukjin Kwon created SPARK-61003:
------------------------------------
Summary: Register UDTFs on the server with spark.udtf.register of
the Python client as PySpark does
Key: SPARK-61003
URL: https://issues.apache.org/jira/browse/SPARK-61003
Project: Spark
Issue Type: Bug
Components: Connect
Affects Versions: connect-rust-4.3.0
Reporter: Hyukjin Kwon
In PySpark Connect, {{spark.udtf.register(name, f)}} registers the UDTF {{f}}
under {{name}} on the server, with a {{register_table_function}} command of its
{{CommonInlineUserDefinedTableFunction}} (its handler, return type, eval type
and determinism), so that SQL can call it: {{spark.sql("SELECT * FROM
name(2)")}}. It raises {{CANNOT_REGISTER_UDTF}} for an {{f}} that is not a
UDTF, and {{INVALID_UDTF_EVAL_TYPE}} for one of another eval type.
The Rust-backed Python client only renames the UDTF object, so the server does
not know the name: for a registered UDTF, {{spark.sql("SELECT * FROM
name(2)")}} raises {{UNRESOLVABLE_TABLE_VALUED_FUNCTION}} where PySpark 4.3.0
returns its rows. It also takes a handler class rather than a UDTF. Register
UDTFs 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]
