[
https://issues.apache.org/jira/browse/SPARK-51899?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Yihong He updated SPARK-51899:
------------------------------
Description:
There are risks associated with working with partial data, especially when
unaware that some tables are broken. Throwing an exception instead provides a
clear indication that something is wrong.
Instead we can use error handling rules to determine the proper behavior on a
case-by-case basis. SparkThrowable should be sufficient to capture the cases we
want to handle.
was:
Swallowing all non-fatal errors and returning partial results can lead to
inconsistent outcomes, especially in cases involving transient errors (e.g., a
table may return full results in {{{}listTables(){}}}, but then return partial
results in a subsequent call due to a transient error).
Additionally, the {{tableType}} field in the {{Table}} class is not nullable.
While it is technically possible to set {{tableType}} to {{{}null{}}}, doing so
violates the expectations of the {{Table}} class. This may surprise people.
To address these issues, we can implement an error allowlist based on SQL
states for non-transient errors and avoid returning results for those tables,
rather than returning partial results when non-transient occur.
> Implement error handling rules for spark.catalog.listTables()
> -------------------------------------------------------------
>
> Key: SPARK-51899
> URL: https://issues.apache.org/jira/browse/SPARK-51899
> Project: Spark
> Issue Type: Improvement
> Components: Connect
> Affects Versions: 4.1.0
> Reporter: Yihong He
> Priority: Minor
> Labels: pull-request-available
>
> There are risks associated with working with partial data, especially when
> unaware that some tables are broken. Throwing an exception instead provides a
> clear indication that something is wrong.
> Instead we can use error handling rules to determine the proper behavior on a
> case-by-case basis. SparkThrowable should be sufficient to capture the cases
> we want to handle.
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
---------------------------------------------------------------------
To unsubscribe, e-mail: issues-unsubscr...@spark.apache.org
For additional commands, e-mail: issues-h...@spark.apache.org
