[
https://issues.apache.org/jira/browse/SPARK-57809?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Max Gekk updated SPARK-57809:
-----------------------------
Shepherd: Max Gekk
> Support nanosecond-precision timestamps in collect_list/listagg ordering
> safety
> -------------------------------------------------------------------------------
>
> Key: SPARK-57809
> URL: https://issues.apache.org/jira/browse/SPARK-57809
> Project: Spark
> Issue Type: Sub-task
> Components: SQL
> Affects Versions: 4.3.0
> Reporter: Max Gekk
> Priority: Major
>
> This sub-task is part of the umbrella SPARK-56822 (timestamps with nanosecond
> precision).
> h2. Problem
> {{collect.scala}} {{isCastEqualityPreserving}} (~L750-762) marks
> {{TimestampNTZType}} as cast-equality-preserving and {{TimestampType}} (LTZ,
> DST) as unsafe, but has no arm for {{TimestampNTZNanosType}} /
> {{TimestampLTZNanosType}}, so both fall to {{case _ => false}}. NTZ nanos
> should be treated like NTZ micro (safe); LTZ nanos like LTZ micro.
> h2. Goal
> Add nanosecond arms so the ORDER BY-cast optimization applies to NTZ nanos
> and correctly excludes LTZ nanos, matching the microsecond behavior.
> h2. Scope
> Extend {{isCastEqualityPreserving}} (and any sibling checks) for
> {{TimestampNTZNanosType}} and {{TimestampLTZNanosType}}.
> h2. Acceptance criteria
> * Ordered {{collect_list}} / {{listagg}} over NTZ nanos uses the safe path;
> LTZ nanos keeps the same DST caveat as micro LTZ.
> h2. Testing
> {{collect.scala}}-related suites; targeted unit test.
> h2. Dependencies
> None - independent (builds on the resolved widening/cast infrastructure).
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
