Haiyang Sun created SPARK-59364:
-----------------------------------
Summary: Build external UDF payload and Init messages
Key: SPARK-59364
URL: https://issues.apache.org/jira/browse/SPARK-59364
Project: Spark
Issue Type: Sub-task
Components: UDF
Affects Versions: 4.4.0
Reporter: Haiyang Sun
SPARK-55278 introduces a language-neutral UDF worker framework, but it lacks a
generic contract for constructing the per-session Init message required by a
worker.
Add:
- Declarative worker session requirements for environment variables, Spark/SQL
configuration, and the Spark files directory.
- Typed Init fields for resolved session context.
- Invocation identity, logical input schema, and named-argument metadata.
- A versioned payload format for scalar Arrow Python UDFs.
- Generic Init construction that forwards only requested context and refreshes
metadata after Catalyst rewrites.
- Unit coverage for defaults, validation, rewritten inputs, worker
configuration, and payload compatibility.
Switching existing Python UDF execution to the unified worker framework is
outside this ticket.
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
