[
https://issues.apache.org/jira/browse/FLINK-39583?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Ferenc Csaky closed FLINK-39583.
--------------------------------
Resolution: Fixed
[{{f9d81b2}}|https://github.com/apache/flink/commit/f9d81b2e11e5cff7469b523cc73ce34b1d28e5ab]
in master
> Normalize Calcite correl variables to make subplan reuse by digest more
> effective
> ---------------------------------------------------------------------------------
>
> Key: FLINK-39583
> URL: https://issues.apache.org/jira/browse/FLINK-39583
> Project: Flink
> Issue Type: Improvement
> Components: Table SQL / Planner
> Reporter: Ferenc Csaky
> Assignee: Ferenc Csaky
> Priority: Major
> Labels: pull-request-available
> Fix For: 2.4.0
>
>
> Goal is to improve digest-based subplan reuse (when
> {{table.optimizer.reuse-optimize-block-with-digest-enabled}} is set to true)
> for plans containing correlation variables, such as {{CROSS JOIN UNNEST}} and
> decorrelated subqueries.
> Currently, structurally equivalent subplans can receive different Calcite
> {{CorrelationIds}} during separate view expansions, cause {{$cor}} uses a
> global counter. These ids become part of the relational digest, so otherwise
> identical subplans do not match and cannot be considered for reuse. The
> change introduces correlation-variable normalization before digest-based
> reuse, assigning deterministic sequential correlation ids per subplan so
> equivalent correlated plans produce matching digests.
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
