flink-issues mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Leonard Xu (Jira)" <j...@apache.org>
Subject [jira] [Updated] (FLINK-15595) Entirely implement resolution order as FLIP-68 concept
Date Tue, 17 Nov 2020 02:50:00 GMT

     [ https://issues.apache.org/jira/browse/FLINK-15595?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel

Leonard Xu updated FLINK-15595:
    Fix Version/s:     (was: 1.12.0)

> Entirely implement resolution order as FLIP-68 concept
> ------------------------------------------------------
>                 Key: FLINK-15595
>                 URL: https://issues.apache.org/jira/browse/FLINK-15595
>             Project: Flink
>          Issue Type: Bug
>          Components: Table SQL / API
>            Reporter: Jingsong Lee
>            Priority: Major
>              Labels: pull-request-available
>             Fix For: 1.13.0
>          Time Spent: 20m
>  Remaining Estimate: 0h
> First of all, the implementation is problematic. CoreModule returns BuiltinFunctionDefinition,
which cannot be resolved in FunctionCatalogOperatorTable, so it will fall back to FlinkSqlOperatorTable.
> Second, the function defined by CoreModule is seriously incomplete. You can compare it
with FunctionCatalogOperatorTable, a lot less. This leads to the fact that the priority
of some functions is in CoreModule, and the priority of some functions is behind all modules.
This is confusing, which is not what we want to define in FLIP-68. 
> We should:
>  * We should resolve BuiltinFunctionDefinition correctly in FunctionCatalogOperatorTable.
>  * CoreModule should contains all functions in FlinkSqlOperatorTable, a simple way could
provided calcite wrapper to wrap all functions.
>  * PlannerContext.getBuiltinSqlOperatorTable should not contains FlinkSqlOperatorTable,
we should use one FunctionCatalogOperatorTable.Otherwise, there will be a lot of confusion.

This message was sent by Atlassian Jira

View raw message