[
https://issues.apache.org/jira/browse/FLINK-31464?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17702505#comment-17702505
]
Timo Walther commented on FLINK-31464:
--------------------------------------
[~jark] thanks for kicking this off. It will help cleaning the messy code base.
Why does the SqlNode conversion logic actually need a context? I was always
wondering why it needs access to e.g. CatalogManager. Ideally, those converters
should simply convert Calcite parser classes to Flink POJO classes. There
should be no resolution logic involved.
> Move SqlNode conversion logic out from SqlToOperationConverter
> --------------------------------------------------------------
>
> Key: FLINK-31464
> URL: https://issues.apache.org/jira/browse/FLINK-31464
> Project: Flink
> Issue Type: Technical Debt
> Components: Table SQL / Planner
> Reporter: luoyuxia
> Assignee: xuzhiwen
> Priority: Major
>
> Similar to FLINK-31368, the `SqlToOperationConverter` is a bit bloated. We
> can refactor it to avoid the code length for this class grow quickly.
> We can follow the idea proposed in FLINK-31368.
>
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
