[ 
https://issues.apache.org/jira/browse/SPARK-61767?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
 ]

L. C. Hsieh updated SPARK-61767:
--------------------------------
    Affects Version/s: 4.4.0
                           (was: 5.0.0)

> Extract the batch export of the in-process Python UDF evaluator into 
> InProcessArrowBatchExporter
> ------------------------------------------------------------------------------------------------
>
>                 Key: SPARK-61767
>                 URL: https://issues.apache.org/jira/browse/SPARK-61767
>             Project: Spark
>          Issue Type: Improvement
>          Components: SQL
>    Affects Versions: 4.4.0
>            Reporter: L. C. Hsieh
>            Priority: Major
>
> InProcessArrowEvalPythonEvaluatorFactory.evaluateBatches (SPARK-59718) 
> interleaves three things in one method of about 300 lines: pulling input rows 
> under the iterator's lock, building and exporting Arrow batches to Python, 
> and releasing both at task completion.
> This moves the second into a class of its own, InProcessArrowBatchExporter, 
> which owns the Arrow writer of the current batch, the imported result 
> vectors, and the registrations of the UDFs. evaluateBatches keeps the 
> orchestration: it pulls rows into the queue and the exporter until the batch 
> is full, evaluates it, and joins the results.
> This is a refactoring without behavior change. The exporter can be tested 
> without a Python interpreter.



--
This message was sent by Atlassian Jira
(v10.3.26#10030026)

---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]

Reply via email to