[
https://issues.apache.org/jira/browse/SPARK-61655?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Hyukjin Kwon reassigned SPARK-61655:
------------------------------------
Assignee: Kaicheng Zhou
> Let a Rust UDF read the session's Python worker environment
> -----------------------------------------------------------
>
> Key: SPARK-61655
> URL: https://issues.apache.org/jira/browse/SPARK-61655
> Project: Spark
> Issue Type: Improvement
> Components: Connect
> Affects Versions: connect-rust-4.3.0
> Reporter: Kaicheng Zhou
> Assignee: Kaicheng Zhou
> Priority: Major
>
> A session can set environment variables for its Python UDF workers with
> spark.pythonWorkerEnv.<NAME> (SPARK-58752, SPARK-59233, Spark 4.4.0). Python
> UDFs sent by the Rust client already receive them. A Rust UDF does not: it
> runs as a wasm32-unknown-unknown WebAssembly module that the runner
> instantiates without imports, so std::env::var always fails inside it.
> Add spark_connect_build::embed_wasm_udf_for(src, WasmTarget::Wasip1) to build
> a UDF for wasm32-wasip1. When a module imports WASI, the runner gives it a
> snapshot of the worker's environment, without the names Spark reserves for
> its workers and without entries that aren't valid UTF-8. The module gets no
> files or arguments, and its standard output and error are discarded. Modules
> built for wasm32-unknown-unknown behave as before.
--
This message was sent by Atlassian Jira
(v10.3.26#10030026)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]