Hyukjin Kwon created SPARK-61318:
------------------------------------
Summary: Raise the exception of a Python progress handler from the
execution that called it, as PySpark does
Key: SPARK-61318
URL: https://issues.apache.org/jira/browse/SPARK-61318
Project: Spark
Issue Type: Bug
Components: Connect
Affects Versions: connect-rust-4.3.0
Reporter: Hyukjin Kwon
A Python progress handler (spark.registerProgressHandler) that raises does not
fail the execution in the Python drop-in: its exception is written to
sys.unraisablehook and collect() returns the rows.
Repro: a handler that raises ValueError on each non-done update, and a query
that runs with progress (SELECT id, reflect('java.lang.Thread', 'sleep', 500L)
AS s FROM range(0, 4, 1, 4)). The drop-in returns the rows and prints a
traceback.
PySpark 4.3.0 calls the handlers from the client's own result loop, so
collect(), toPandas() and count() raise the handler's exception itself.
--
This message was sent by Atlassian Jira
(v10.3.26#10030026)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
