[
https://issues.apache.org/jira/browse/CALCITE-4744?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
keaton drebes updated CALCITE-4744:
-----------------------------------
Description:
I have a project that is working with windowed functions and we want to avoid
the expansion of certain aggregate functions (i.e. stddev) in the process of
converting from an SqlNode to a RelNode. We have currently supported this by
copying the entire StandardConvertlet table, and removing the convertlets
responsible for this.
It seems the correct way to resolve this issue is to defer the expansion of
aggregate functions until the planning step, then add a planner rule to
enable/disable it.
was:
I have a project that is working with windowed functions and we want to avoid
the decomposition of certain aggregate functions (i.e. stddev) in the process
of converting from an SqlNode to a RelNode. We have currently supported this by
copying the entire StandardConvertlet table, and removing the convertlets
responsible for this.
It seems reasonable to control this behavior with a configuration for the
StandardConvertlet table. Is this something that people would accept being
merged into Calcite?
I opened a draft PR to show the changes here:
https://github.com/apache/calcite/pull/2496
> Defer windowed aggregation function expansion to the planning step
> -------------------------------------------------------------------
>
> Key: CALCITE-4744
> URL: https://issues.apache.org/jira/browse/CALCITE-4744
> Project: Calcite
> Issue Type: New Feature
> Components: core
> Reporter: keaton drebes
> Priority: Minor
> Labels: pull-request-available
> Time Spent: 20m
> Remaining Estimate: 0h
>
> I have a project that is working with windowed functions and we want to avoid
> the expansion of certain aggregate functions (i.e. stddev) in the process of
> converting from an SqlNode to a RelNode. We have currently supported this by
> copying the entire StandardConvertlet table, and removing the convertlets
> responsible for this.
> It seems the correct way to resolve this issue is to defer the expansion of
> aggregate functions until the planning step, then add a planner rule to
> enable/disable it.
>
--
This message was sent by Atlassian Jira
(v8.3.4#803005)
